The CDT and copywaza lab [Archive] - XtremeSystems Forums

massman

10-30-2007, 07:00 AM

OPB's tweak which helped him to do the
Thread on Coolaler forums: http://forum.coolaler.com/showthread.php?p=1889672
Seems like a very good find, but does it work for everyone? Does anyone have the theoretical explanation?

__________________________________________________ ____________________

1st step:
- Make the pagefile exactly 512~512mb under the partition you place super pi 1.5mod.
- Make a new folder named whatever you like , make 3 files which size can be about { 600m(628666368)--->maxmem (I set), and pplus 32m..}x 3

2nd step:
- Use winrar to make the single rar files ...so it will be smaller or equal to 2gb, if you fail during compress....you know your setting ddefinitely haas problem on it.
- Run the window of super pi 32m (or even 1m) set the size you wanna run...and leave the dialogue there.

3rd step:
- Cut this rar file and paste @ C:/ for 3 timess.

** check with task manager for the available memory size...so let the system idle for at least :

20seconds for 1m
45 second for 32m

1m need to be greater than 460m

suggestion from OPB..:
you will need to do above BEFORE you use setfsb or clockgen to OC your FSB and change your memory parameters from memset.

Good luck.
**you also can use cut-paste -copy ...this is what I will try to do the experiment with CDT-V

the file is like this...
632m(about) need 3 of them...no ned to be exactly, but need to be only 3..
cuz not like you stated above..that would do it like 6 or more...that would completely mess the sequential compress action up.

when you run super pi, you need to choose size right?
after you choose , the dialogue will show:

"now start to calculate"
just stop there and do whatever need to do

reason:
cuz at that moment, you system has about to prepare some elements need to running super pi size you set...very important.

you cut the file , your ram is tempoarily store that img some address,
certainly you always cut from D to c..I did say from where you place the super pi folder right?

so after you paste on c...right click your mouse...can you do one more...two again?...
it will ask you "overwrite it..."
ok ..you need one folder.. only one for entire CDT-IV

why ? becuase you need to make it into a rar files..
so only ...ONLY make a folder ..and you will need to find 3 files are about 632m( 600m+32m is the hint for thinking why..but doesn NOT have to be 600+32..(I can have 598+34? right?...)

so the best way is you only make it once for 632mb for instance, then copy it into the folderright? then re-name the original to othername and copy into that folder...so now your folder has how much? 1264mb? so the one remain out side the folder still 632mb right?
so rename again and "cut it and paste into that folder...
now, if you install winrar already, right click that folder, and make a whatever name of the rar file...there we go, after done...you know what to do...

**All this way above is NOT must do like this...it's only one of the 1000 ways...to form you a 1.8g~2.0gb single rar file....

but why I give the hint for 3x(600mb+32mb) to form that single rar?...haha..

**************************

for everyone:
if you wanna integrate Copy-waza, please, after you do Cut and paste , copy from c to D and then you will need to paste @ d over-write for 2 times..., you will get the most surprised result.

here we go for elementry scool dialogue

http://img132.imageshack.us/img132/8940/85693255to3.th.jpg (http://img132.imageshack.us/my.php?image=85693255to3.jpg) http://img222.imageshack.us/img222/6084/80018471bk3.th.jpg (http://img222.imageshack.us/my.php?image=80018471bk3.jpg) http://img222.imageshack.us/img222/6589/82919142nv0.th.jpg (http://img222.imageshack.us/my.php?image=82919142nv0.jpg)

will post real procedure during making this.

during the initial small file make up:

http://img147.imageshack.us/img147/733/54602559yt2.th.jpg (http://img147.imageshack.us/my.php?image=54602559yt2.jpg)

1st step completed; First C.I formed
2nd step initializing:
see that difference between available memory and system cache?

http://img220.imageshack.us/img220/2406/51085556yh9.th.jpg (http://img220.imageshack.us/my.php?image=51085556yh9.jpg)

then...
** Anti-derevitive?lol 3 times from wherever to D
http://img220.imageshack.us/img220/6431/27862219yd7.th.jpg (http://img220.imageshack.us/my.php?image=27862219yd7.jpg)

Finalize the final step of CDT-iv...Cut...(Dereivitive in calculus?)

http://img220.imageshack.us/img220/462/28587147rn4.th.jpg (http://img220.imageshack.us/my.php?image=28587147rn4.jpg)

the final step, idle and let system balance, close the task manager and you're redy to go.

http://img229.imageshack.us/img229/5528/83178884jd6.th.jpg (http://img229.imageshack.us/my.php?image=83178884jd6.jpg)

EDIT: massman your post is a bit confusing, i removed some free spaces and comments in regards to the tweaks author :P
And i hope you dont mind if i replace the pics with thumbnails since they are quite large -Saaya

BeardyMan

10-30-2007, 07:14 AM

My way of doing it.
http://www.xtremesystems.org/forums/attachment.php?attachmentid=66200&stc=1&d=1193402823

saaya

10-30-2007, 07:23 AM

massman you mind want to reword that post of yours and give it some more structure, its quite messy and doesnt have a clear step by step instruction on what to do. Theres just step 1, step 2 and then dozens and dozens of paragraphs and a couple of screenshots :D

massman

10-30-2007, 07:28 AM

saaya

10-30-2007, 07:32 AM

It sure looks a lot cleaner now :)
yeah but you should make it an easy step by step guide, otherwise there will be tons of misunderstandings and everybody will do things diferently... which is what happened in the last thread.
The problem is that while sorting everything you might mix up something or misunderstand something which then results in the tweak not working... thats the problem with kevin not being a native english speaker and not being available to clear those possible misunderstandsing. But imo you should at least give it a try and maybe ask kevin if he can read it through and let you know if you made any mistakes.

kiwi

10-30-2007, 07:34 AM

Yep, clear step by step would be nice, otherwise its quite messy and hard to understand all details

This is what I understood best and was posted by Elmor obviously:

This tweak definitely works!

Pics up soon.

edit:

System specs, just straight boot didn't editing any timings:
http://bandwidth.se/imgs/1/74/sysspecs.jpg

Then just run without cw/cdt:
http://bandwidth.se/imgs/1/74/nocw.jpg

And with the previous OPB CDT I learned from him:
http://bandwidth.se/imgs/1/74/cdt.jpg

Now, how about with this new OPB CDT-IV tweak:
http://bandwidth.se/imgs/1/74/opbcdt.jpg

Wow, how about that? I just shaved of 16 seconds of a SPI 32M round!

I'll explain exactly how I did it as some have some trouble understandning how to do this tweak.

First you will need to set your maxmem to 600 (I used 620 which worked good aswell, though I belive the file 600+32MB is 600MB = avaliable ram and the extra 32MB is because of pi 32M calculation, am I right Kevin?) and also I believe it is very important to have a 512-512MB big pagefile on the SPI-partition (D: in my and Kevins case). Ok, now the instructions:

1. I created 1x 632mb file using the command "fsutil file createnew D:\myfile 647168000" (1000*1024*632) for a file of exactly 632MB.
2. Then I put it in a RAR-archive with the Store Compression method for no compression.
3. After that I renamed it to myfile2 and then myfile3 and after each rename add it to the RAR-archive with the Store option so that you get a 1896MB big RAR-archive.
4. Then I transfered it to my benchdisk with an USB flashdisk (I created the file on my laptop) and put it on SPI-partition (D:\file.rar).
5. Now you start SPI and select 32M calculation and press OK, but don't press OK to start it, let it wait there.
6. After that you copy it three times to C:\file.rar and the two other times you have to overwrite of course.
7. Next you copy D:\file.rar to any folder on D:\, for me it was D:\cw\, three times just like you did to C:\ before.
8. Now, as Kevin says, to balance it you copy the file C:\file.rar to D:\file.rar and overwrite it. I did it three times and I saw the ammount Avaliable memory raise each time.
9. Now you can close the Taskmanager if you had it open for monitoring and all the other windows not related to SPI and now you can start your calculation!

Hope this writeup helps you understand Kevins outstanding tweak better, I can only imagine what it takes to be able to find this tweak out.

BTW, I didn't get the Avaliable memory and System cache balanced so System cache was about 2MB more. If I can get those two balanced maybe even more gain is possible.

http://www.xtremesystems.org/forums/showpost.php?p=2509761&postcount=31

massman

10-30-2007, 07:37 AM

Copy-Waza guide by 3oh6: http://3oh6.com/forum_posted/ocx/jasons_screencast/3oh6_screencast_copywazza-1.html

KTE

10-30-2007, 08:06 AM

First a basic qs guys : :)

Why is that Windows Task Manager in post#2 showing 613932KB installed (total available memory to OS) whilst 1GB would show as +1000000KB?

K404

10-30-2007, 08:23 AM

First a basic qs guys : :)

Why is that Windows Task Manager in post#2 showing 613932KB installed (total available memory to OS) whilst 1GB would show as +1000000KB?

Thats what it would/does look like after applying a MAXMEM of 600MB

MSConfig-> Boot.ini-> Advanced options-> Maxmem

K404

10-30-2007, 08:26 AM

From the testing i've done, using CDT, Available RAM + System cache are improved in comparison to Copy-Waza but it isnt translating into a better SPi time for me.

I think its very possible, seeing as Copy-Waza boosts the Avail + Sys Cache and gives a better SuperPi time, that the tweak works, but im doing something wrong.

After copy-waza:

Available RAM: 459808
Sys Cache: 531288

After CDT:

Available RAM: 550000
Sys Cache: 555596

MAXMEM set to 600 (giving 613932KB in task manager)

Original Copy-Waza file was a copy of the original F.E.A.R. demo (636MB,) and was copied from a 2nd HDD onto Windows partition.

KTE

10-30-2007, 08:39 AM

Thats what it would/does look like after applying a MAXMEM of 600MB

MSConfig-> Boot.ini-> Advanced options-> Maxmem
Excellent. Means my Maxmem is not working for some odd reason. It's a fudged install so new one coming up tonight it seems. :(

MAXMEM set to 600 (giving 613932KB in task manager)Fixed. :)

K404

10-30-2007, 09:12 AM

K....M..... close enough! :p:

;) Cheers

BeardyMan

10-30-2007, 09:41 AM

I tried maxmem 600 instead of 680, gain seems to be similiar.
Same method used as in my first post here. files copied to the same hdd on my desktop.
edit
it seems 680 is better then 600 for 632kb file.

Zeus

10-30-2007, 11:28 AM

BeardyMan

10-30-2007, 12:07 PM

Beardy, i don't think you can compare AMD to Intel, it has a different way of utilizing the memory so gains from a tweak that frees up memory and systemcache will be different i guess.

I remember how i gained a couple of tenth's in 1M with my old socket 939 Opty with CW, for this C2D setup CW does nothing for 1M.

I know, but we are here to share thoughts wheter it works or not for ourselfs :D
I just went true allot of settings on maxmem and i'm quite sure that 680 gives teh lowest time although when setting 735 i get almost 1sec gain, yet it's time is higher then 680.
Strange :D

STEvil

10-30-2007, 06:08 PM

Just a suggestion guys, but maybe whether this tweak works or not could be related to ram timings or sub-timings?

T_M

10-30-2007, 06:38 PM

Im really interested to find out the theory behind the cache/memory tweaks, why does it help SuperPI and what does balancing it and using maxmem actually achieve.
That would be better for me than just blindly testing/following tweaks.

Onepagebook

10-30-2007, 07:51 PM

No Mr. Stevil;) at least not that big.

* I wanna see how duck can sing rock n roll;)

KTE

10-30-2007, 11:18 PM

OK, I've managed to run a session of experimental testing on an older P4 Celeron I have here - which is the only built system lying around here apart from 2 work laptops.

I ran 8M instead of 32M because at 3300-3600MHz the C2D 32M time is equivalent to this chip at 8M (and.. it fracking takes 47m for 32M :)). So you get a good idea of the SAME time period (you can equivocate seconds better i.e., between 10-15minutes the drop in time from a tweak is the most similar overall percentage to everyone else's)

If there is any gain in any method, this will show it purely simple for me.

I tested ALL the tweaks separately to see WHICH one gains or loses and which one gains most.

-Without Maxmem.
-With Pagefile 250-1000MB
-With Maxmem 600MB (working).
-With Pagefile 512-512MB
-Completely stripped OS (XP Pro SP2 32bit 14 processes 118MB startup)
I followed George's method for CopyWazza testing (but left pagefile constant at 512-512MB for accuracy).
I followed Elmors method for CDT testing.
I tried another tweak which maximizes RAM as I normally do separately.

Before this I was totally "speculative" and "skeptic" as anyone unbiased would be.

Let me say one thing clear now:

CDT works :)

mrlobber

10-30-2007, 11:33 PM

Let me say one thing clear now:

CDT works :)

That was known previously as well :) Can you elaborate, how much you gained with CDT vs copywaza?

Zeus

10-31-2007, 12:16 AM

Just a suggestion guys, but maybe whether this tweak works or not could be related to ram timings or sub-timings?

I think it has a lot more to do with one partitioned hdd or two seperate ones, position of the pagefile and things like that.

I even wonder if cut-paste or copy-paste makes a difference.

Is there somebody who can explain why the rar files consist 3x632mb rar file? There must be a reason why it is created like this rather than making a 1.8-2.0GB file out of pics or other files?

saaya

10-31-2007, 12:51 AM

Onepagebook

10-31-2007, 01:16 AM

I think it has a lot more to do with one partitioned hdd or two seperate ones, position of the pagefile and things like that.

I even wonder if cut-paste or copy-paste makes a difference.

Is there somebody who can explain why the rar files consist 3x632mb rar file? There must be a reason why it is created like this rather than making a 1.8-2.0GB file out of pics or other files?

I will let you have all the answer, buddy, YGPM,
looks like someone start picking agin ;)

massman

10-31-2007, 01:26 AM

So what exactly do those tweaks do?
they memory content is pushed to the pagefile, so onto the hdd, which frees up extra system memory to be used for the pi calculations?
As a result less of the calculation results of the benchmarks need to be moved to the pagefile cause the memory is full? How big are the files that the pi benchmark creates during a 32m run? I dont mean how many files overall but how many files and of what size does the benchmark create that need to remain buffered to create the final result?

How exactly does the pi app work?
it creates temp files in memory?
Are they deleted or ready to be deleted once the iteration finished?
Does the pi app overwrite its older temp files or keeps going up and up memory address wise until the mem is full and then even writes temp files to the pagefile on the hdd?

SuperPi will use the HDD as storage if there is not enough memory, so by doing copy-waza, you free up the rams. 632 seems to be related to the maxmem 600 and 32MB (kinda obvious). 632MB is exactly the size of what a 32M runs needs, so by clearing up the matrix of the memory (where the data is stored), you make place for the data of the 32M run.

i'm not sure at all about this :).

By the way, when copying the files, you should notice that when you copy from D -> C it'll be much slower than when you copy back to D. Why? Pagefile?

KTE

10-31-2007, 01:36 AM

OK, yes I have the results saved. :D

I tested exactly as called by others following the testing methods given by BenchZowner/Hipro5 and Elmor/OPB.

*All standard Windows tweak done on all runs (2 or 3 services running max, 13 processes idle, 118MB) but for the first completely untweaked normal Windows run.
*When I tested 512MB-512MB pagefile vs. 384MB-384MB there was absolutely NO difference in 8M-4M-1M runs so I left it to 512-512MB for all tests but first two.

OPB/Elmor's CDT method: (1) (http://www.xtremesystems.org/forums/showpost.php?p=2509761&postcount=31)
Hipro5's CopyWazza method: (1) (http://www.xtremesystems.org/forums/showpost.php?p=2509928&postcount=40) (2) (http://www.xtremesystems.org/forums/showpost.php?p=2519794&postcount=315)

First thing I'd like to mention is that I could not repeat and get 500MB/500MB for memory/cache out of 600MB installed! The maximum mine went up to after CDT was 365MB/363MB. I "think" this depends purely on your hard drive cache (mine was 8MB/2MB).

For CopyWazza, same method; it did NOT raise my system cache/available memory by much at all. This again I think depends on your hard drive cache size.

Test Setup
P4 Celeron 2800MHz D0 @ 3360MHz (stock volts 1.312V)
512MB x 2 (dual channel) OCZ DDR400 2-3-3-6 1T (4:5)
Hitachi Desktar ATA133 7200RPM 80GB 8MB cache HDD
Maxtor ATA133 7200RPM 80GB 2MB cache HDD

Format
Tweaks
=====
Task Manager Available Memory (SPi Available Memory)
Task Manager System Cache
Best 8M Time

Windows bootup Idle memory/cache/processes
http://img140.imageshack.us/img140/6932/tm1bl1.th.png (http://img140.imageshack.us/my.php?image=tm1bl1.png) http://img141.imageshack.us/img141/3194/tm2vj5.th.png (http://img141.imageshack.us/my.php?image=tm2vj5.png)

SuperPi Mod 1.5 / XS 8M

Test 1:
Pagefile 250-1000MB
Not real-time
====
469.7MB (439.23MB)
409.67MB
10m 48.212s
http://img518.imageshack.us/img518/5770/8m1048gs8.th.png (http://img518.imageshack.us/my.php?image=8m1048gs8.png)

Test 2:
Tweaked Windows
Maxmem=600 (Boot.ini)
Pagefile 250MB-1000MB
Not real-time
====
469.23MB (472.83MB)
69.99MB
10m 25.469s (sorry missed the screenshot :( )

Test 3:
Tweaked Windows
Maxmem=600 (Boot.ini)
Pagefile 512MB-512MB
Real-time
=====
472.86MB (481.07MB)
92.90MB
10m 22.955s
http://img150.imageshack.us/img150/7697/8m1022dn3.th.png (http://img150.imageshack.us/my.php?image=8m1022dn3.png)

Test 4:
Tweaked Windows
Maxmem=600 (Boot.ini)
Pagefile 512MB-512MB
Real-time
CopyWaza (C/D, 1st partitions, different drives)
====
366MB (377.04MB)
363MB
10m 20.752s
http://img143.imageshack.us/img143/9549/8m1020db1.th.png (http://img143.imageshack.us/my.php?image=8m1020db1.png)

Test 5:
Tweaked Windows
Maxmem=600 (Boot.ini)
Pagefile 512MB-512MB
Real-time
FreeRAM XP Pro
====
513MB (528.96MB)
58.5MB
10m 19.881s
http://img145.imageshack.us/img145/8674/8m1019ma2.th.png (http://img145.imageshack.us/my.php?image=8m1019ma2.png)

Test 6:
Tweaked Windows
Maxmem=600 (Boot.ini)
Pagefile 512MB-512MB
Real-time
CDT (D/E, 1st/2nd partitions, same drive)
====
490.60MB (504.48MB)
104.75MB
10m 19.861s
http://img150.imageshack.us/img150/4698/8m10198iw0.th.png (http://img150.imageshack.us/my.php?image=8m10198iw0.png)

Test 7:
Tweaked Windows
Maxmem=600 (Boot.ini)
Pagefile 512MB-512MB
Real-time
CDT (C/D, 1st partitions, different drives)
====
365.1MB (375.41MB)
360.1MB
10m 15.696s
http://img147.imageshack.us/img147/9199/8m1015cdtnp9.th.png (http://img147.imageshack.us/my.php?image=8m1015cdtnp9.png)

Lets just say I spent a whole week (before today) trying to beat the CopyWaza time with everything else and it was impossible... until I tried the last technique. ;)

Hope it helps.

massman

10-31-2007, 01:53 AM

Great work !!

I'll read it through when I get back home :)

saaya

10-31-2007, 02:27 AM

saaya

10-31-2007, 02:33 AM

KTE, what last technique?

Your running this on a celeron with little cache, which makes your results not directly comparable but interesting! and your not running 32m so maybe theres a tweak or setting that only works with 32m, but i find that unlikely, if there is no gain in 8m its unlikely that theres a gain in 32m i think.

so you went from 10:48 untweaked to 10:15? thats a nice boost... :D
Are those individual runs? did you make each run only once?
Is this the average or the best results with each technique?

thanks for all the time you spent on this and thanks for sharing :toast:

mrlobber

10-31-2007, 02:43 AM

KTE, top shelf testing stuff :eek: :up:

Too bad you are using a Prescott Celeron, not a C2D... CDT seems to have different impact on various families of processors, just like BeardyMan showed with his A64.

Interesting thing about the hard drive cache impact... I did all my comparisons on a single 20Gb IDE 2Mb drive, time to retest everything with larger cache drives and with 2 different drives :cool: That 4second gain on your 2nd CDT run (between 2 drives, not 2 partitions of the same drive) is especially noteworthy.

I think it has a lot more to do with one partitioned hdd or two seperate ones, position of the pagefile and things like that.

Heh, KTE just showed the opposite... although OPB of course knows what he is saying, which makes things even more interesting :)

I will let you have all the answer, buddy, YGPM,
looks like someone start picking agin ;)

KTE

10-31-2007, 02:56 AM

hey massman... thats weird though, why does it need that much space to calculate pi? it ultimately makes it a hard drive benchmark considering this bench is so old... back then the cache was so slow that the hdd must have played a big role in getting good results. now which uge caches and mem the problem isnt that much the hdd i guess.Yep that was exactly my theoretical thought before running. I thought it 'aint logical for CW/CDT to effect 1M or 32M since SPi is small enough to fit inside most nowadays CPU L2 cache and if not then RAM. What does HDD have to do with it? I can't yet explain my practical results. They are showing HDD cache/buffers and OS/CPU prefetch is affecting SPi result.

KTE, what last technique?Last one was CDT using two different drives rather than two different partitions on one drive. The latter didn't work much at all.

Your running this on a celeron with little cache, which makes your results not directly comparable but interesting!True. I only tested to get a sense of "across the board" effect. :D

and your not running 32m so maybe theres a tweak or setting that only works with 32m, but i find that unlikely, if there is no gain in 8m its unlikely that theres a gain in 32m i think.Honestly, I've only ran 32MB once on a P4 3GHz Northwood in the last year and it took wayyy too painfully long to want to run again 21-30 times on a Celeron to test. :shakes: I could run it for further investigation if anyone wants, but these are just some preliminary results. Still at work here. ;)

so you went from 10:48 untweaked to 10:15? thats a nice boost... :D
Are those individual runs?Yup. Best result in 3 runs for each technique was given above.

thanks for all the time you spent on this and thanks for sharing :toast:
You're welcome mate. :toast:

v|aDy

10-31-2007, 03:00 AM

Without CDT 14.15.125
http://img219.imageshack.us/img219/4766/3600nocdtfc8.png (http://imageshack.us)

With CDT: 14.04.063
http://img100.imageshack.us/img100/77/3600cdtaw5.png (http://imageshack.us)

With CopyWaza by Hipro5 method: 14.05.563
http://img227.imageshack.us/img227/8470/3600copywazamy2.png (http://imageshack.us)

saaya

10-31-2007, 03:38 AM

KTE, was there a lot of fluctuation between the 3 runs you did for each technique? was this a fresh windows install? with the latest service packs?

several people seemed to have a lot of problems to reproduce their results.
talking about this, has anybody heard back from elmor yet? :confused:

Zeus

10-31-2007, 03:55 AM

hey massman... thats weird though, why does it need that much space to calculate pi? it ultimately makes it a hard drive benchmark considering this bench is so old... back then the cache was so slow that the hdd must have played a big role in getting good results. now which uge caches and mem the problem isnt that much the hdd i guess.

again, weird why it needs so much space...
and 632mb... well we are talking about systems with 2gb of mem so why would pi write something to the hdd and not the mem, the mem should have plenty of space for any temp file pile that bench comes up with...

copying what from d to c makes the pi run slower?

It does not need much harddisk space and only uses very little memory in fact.

Here's a quote from the SuperPi help folder:

Memory requirement:
For the maximum calculation speed, 8 MB/1 million decimal digit is favorable, but 2 MB/1 million decimal digit is completely acceptable. Working memory is automatically adjusted by the software. Anyway, available main memory size is crucial for the processing speed!

Disk storage:
As for the working disk storage, 10.5 MB per 1 million decimal digit is needed. Working disk storage is automaticall freed. As for the permanent data storage, 1 MB per 1 million decimal digit is needed. Elapsed time is very keen to the disk access time. In order to short the elapsed time, you are better to equip high speed hard disk drive!

Having much RAM doesn't have an advantage here so it seems, how interesting is the sentence about the harddisk on the other hand? ;)

zbogorgon

10-31-2007, 04:29 AM

IMHO, Spi 32M is much more tweakable than 1M

EDIT: comment removed, please read the first post
-Saaya

KTE

10-31-2007, 04:34 AM

Two things:

Does SPi write to the HDD real-time, in the last iteration or after finishing?
Do CW/CDT help with the prefetch buffers in both the CPU and HDD?

KTE, was there a lot of fluctuation between the 3 runs you did for each technique?
Differences between the 3 runs were in milliseconds not seconds. Each of those times were very difficult to get at those clocks/timings with that setup. A typical fully tweaked run will get you 10m 22.9s without those extra methods, repetitively, no matter how hard you bang your head. :p:

The first run I did, something was wrong with Maxmem flag not working in boot.ini (that would give a CopyWaza 8M time of 10m 25s) so I reinstalled XP Pro SP2 32-bit + all updates. Then just ported over many of the files I needed on that install. I went into Start > Run> msconfig and chose to load the "minimum services/drivers/processes" option (which gives you problems later when restarting them, beware).

several people seemed to have a lot of problems to reproduce their results.
I'll do some re-tests a little later just to clarify if those results are repeatable and to what extent on that setup. Not sure how long that PC will be there before being disassembled. 5 of each this next time around. I'll test that later on with 32M too (bloody fire haah, 50 minutes x30 of waiting :( :( )

talking about this, has anybody heard back from elmor yet? :confused:
Nope. But I would've done the same right thing in his position.

Elmor, (I forgot your name, Robert?) please give us more information here for everyones benefit. And on behalf of myself and others, thank you for defending what you knew and experienced. I don't like weasels or those not man enough to admit what they factually experienced. :)

Admin has laid down the rule of no ad hominems and kiddy crap, so I believe he will stick to admonishing that in this thread (unlike what happened in the previous).

BeardyMan

10-31-2007, 05:52 AM

A64 X2 at 3.4ghz boxed cooled 376mhz ddr2 5-5-5-18
and 5gb maxtor instead of the 160gb with 16mb buffer.
gain = 0.3secs instead of the 0.6secs average
from 25.500 to
http://www.xtremesystems.org/forums/attachment.php?attachmentid=66560&stc=1&d=1193837635

massman

10-31-2007, 07:10 AM

Onepagebook

10-31-2007, 07:23 AM

Yes Peter, well see someone start yelling and still be allow here? what do we say?;)

saaya

10-31-2007, 07:45 AM

- No, the 632MB is used to clean up the memory's matrix from all the other data. It's not about cleaning the HDD, I think. The HDD transfer is just a way to clean the memory.
how does it clean the memory?
If you overwrite 11101010100101 with 0011101010101001 how does that "clean" the memory?

KTE

10-31-2007, 08:49 AM

how does it clean the memory?
If you overwrite 11101010100101 with 0011101010101001 how does that "clean" the memory?I'm not sure how it cleans memory, but here is one possibility:

Here we have loss of time with these methods. The main purpose of a cache in any device is to decrease access time as well. The larger the cache hit ratio, the better the prefetch, the faster the access time, the quicker the application/instruction time. This may hold us some clues. :)

You have some RAM address ranges reserved for hardware I/O mapping in Windows by default and others reserved deemed necessary for kernel level code. On top of this you'll have extra taken up by basic system processes and they tend to accrue and not release from memory even if the address range is no longer needed. Thus you'll have a section of memory unaddressable for no reason.

When you force the "cleansing" by something which will either a) defrag the RAM b) force it empty, such as a bigger process requiring the full RAM, naturally the Windows cache and memory will empty all unneeded extra address ranges which were reserved by other applications (prioritize) and start filling up if you've given Windows priority to real-time Programs. FWIW the HDD pagefile is also a cache managed by Windows kernel level. Then rather than the RAR files still retaining in memory as many applications will, after the copying finishes, they are released immediately and that part of memory becomes freely available to everything subsequently. I "suspect" the type of files matter (RAR files) but I'll try it more thoroughly soon.

That's as far as I can see if memory increase does play a role and how. For me, this has all to do with the various caches and prefetch algorithms as this is their well known function - to increase speed and decrease latency.

Onepagebook

10-31-2007, 09:06 AM

I'm not sure how it cleans memory, but here is one possibility:

Here we have loss of time with these methods. The main purpose of a cache in any device is to decrease access time as well. The larger the cache hit ratio, the better the prefetch, the faster the access time, the quicker the application/instruction time. This may hold us some clues. :)

You have some RAM address ranges reserved for hardware I/O mapping in Windows by default and others reserved deemed necessary for kernel level code. On top of this you'll have extra taken up by basic system processes and they tend to accrue and not release from memory even if the address range is no longer needed. Thus you'll have a section of memory unaddressable for no reason.

When you force the "cleansing" by something which will either a) defrag the RAM b) force it empty, such as a bigger process requiring the full RAM, naturally the Windows cache and memory will empty all unneeded extra address ranges which were reserved by other applications (prioritize) and start filling up if you've given Windows priority to real-time Programs. FWIW the HDD pagefile is also a cache managed by Windows kernel level. Then rather than the RAR files still retaining in memory as many applications will, after the copying finishes, they are released immediately and that part of memory becomes freely available to everything subsequently. I "suspect" the type of files matter (RAR files) but I'll try it more thoroughly soon.

That's as far as I can see if memory increase does play a role and how. For me, this has all to do with the various caches and prefetch algorithms as this is their well known function - to increase speed and decrease latency.

Based on what your statement here, what I would like to add is
THe type of RAR file doesn't matter but its size.;)

BeardyMan

10-31-2007, 09:20 AM

Size is what i'm going to test now :)

around 2Gb vs 632Mb :D

hipro5

10-31-2007, 11:40 AM

Lets just say I spent a whole week (before today) trying to beat the CopyWaza time with everything else and it was impossible... until I tried the last technique. ;)

Hope it helps.

http://www.xtremesystems.org/forums/showpost.php?p=2520131&postcount=327

;) You're "steeling" my techniques....Shame on you.... :D
Now TRY the SAME with a BIG Copy Waza.... ;)

BeardyMan

10-31-2007, 11:43 AM

using a big file seems to be uselles for me :) loss instead of gaining...

hipro5

10-31-2007, 11:50 AM

using a big file seems to be uselles for me :) loss instead of gaining...

With TWO SEPARATE HDs.....One with OS and everything in it and one ONLY with the PAGEFILE (clean) and at the BIGINING of the HD.....
Though BEST is if someone could use a Gigabyte iRAM as SECOND HD with the page file IN IT... ;)

elmor

10-31-2007, 03:27 PM

Nice to see this back on track again, great work saaya.

Elmor, (I forgot your name, Robert?) please give us more information here for everyones benefit. And on behalf of myself and others, thank you for defending what you knew and experienced. I don't like weasels or those not man enough to admit what they factually experienced. :)

Nowhere near Robert, that's crotale :D I do not have so much more info to add but I can sum up what I've experienced so far.

several people seemed to have a lot of problems to reproduce their results.
talking about this, has anybody heard back from elmor yet? :confused:

First time I tried the CDT Tweak I had those crazy results that I posted in the previous thread. However after reformat I was never able to get System cache about the same as Avaliable memory again and hence no boost (that's what I believe gives the boost) except a few seconds which could probably be aquired with hipros method aswell. I started using maxmem = 600 and I actually got worse times with that than with 620 so one could do some experimenting there.

Currently I'm playing with P5E3 WS Pro as I got kinda frustrated over not getting this to work again, but I'll get into it once again when I have some time to spare.

//Jon

massman

10-31-2007, 04:16 PM

Hmm, I think I'm reaching the point where CDT gives a small boost, though, before I try to explain and rewrite OPB's guide, I need to ask a few questions, make a few statements. I hope that Kevin finds the time to respond and point out the wrong ones :).

- CDT basicly is an improved Copy-Waza. It's build on the same fundamentals, but is more elaborate. The average gain depends on what frequency you're running, but at 3.6 - 4.0G, the gain should be +/- 6 to 10 seconds (if copy-waza and cdt are done correctly).

- The superpi run can be split up in several segments, two of them are very recognisable (got some graphs, tomorrow). These two segments seem highly dependant of the harddisk and pagefile. OPB: does increasing the pci freq help? The lower the harddisk's latency, the better, right?

- Everyone who's wondering what file he should use:
1 file which is exactly 3x632MB big. The proces of how to create the files has already been told in the second post. Tomorrow there will be a graph.

- OPB: your theory is based on Server03, right? Just because of the LargeSystemCache tweak, which is standard enabled on 03, or is there another reason?

ELMOR: have you enabled the Largesystemcache reg key? Without that, CDT and Copy-waza are useless. (Beardy: that's the setting we spoke about on MSN. Of course it's obvious :p)

Onepagebook

10-31-2007, 05:33 PM

Pete, I will directly quote the whole thing I re-post and integrated CDT-V for you , that 's the prepare job before you doing anything further, so check your Pm 5 ~10 minutes later:)

KTE

10-31-2007, 07:16 PM

http://www.xtremesystems.org/forums/showpost.php?p=2520131&postcount=327
;) You're "steeling" my techniques....Shame on you.... :D
Now TRY the SAME with a BIG Copy Waza.... ;)
I didn't know about that BUT it's good that works for both of us. :D

I'll re-run if you want doing copywaza again.

But how big a CW file?
I tried 4GB already, two different drives, 1st partitions.

[XC] gomeler

10-31-2007, 07:47 PM

Just curious, has anyone run a few 32M runs with some HDD monitoring applications to see what sort of hits are being addressed to the disk? This will solve all your questions of how important the drive is. You could also artificially cripple the drive by loading up the drive with some throughput tests while running SPi but this would also affect the memory available to SPi, skewing the results. If I had a second machine I'd try to run the HDD tests myself but all I have is my webdev server right now.

dinos22

10-31-2007, 08:22 PM

- CDT basicly is an improved Copy-Waza. It's build on the same fundamentals, but is more elaborate. The average gain depends on what frequency you're running, but at 3.6 - 4.0G, the gain should be +/- 6 to 10 seconds (if copy-waza and cdt are done correctly).

it has to be a bigger gain because OPB did 12m 39s at 3.6GHz with CAS5 even :confused:

hipro5

10-31-2007, 11:00 PM

it has to be a bigger gain because OPB did 12m 39s at 3.6GHz with CAS5 even :confused:

Quoted and I also repeat......
You MUST see ~10 - 15sec LESS than Copy Waza according to the letest OBP's 32M at 3.6GHz...;)

I didn't know about that BUT it's good that works for both of us. :D

I'll re-run if you want doing copywaza again.

But how big a CW file?
I tried 4GB already, two different drives, 1st partitions.

Yes make it ~4GB and cut/paste from one HD to another for 3 times.......Also put a fixed Page File at 384-384.... Thanks....:)

mrlobber

10-31-2007, 11:43 PM

Well, even if kind of "semi-working" CDT gave 6 seconds vs a good copywaza, I'd be satisfied :) Eagerly awaiting your results and explanation massman.

elmor

11-01-2007, 01:01 AM

ELMOR: have you enabled the Largesystemcache reg key? Without that, CDT and Copy-waza are useless. (Beardy: that's the setting we spoke about on MSN. Of course it's obvious :p)

Yes, of course I use that :)

JD1985

11-01-2007, 01:10 AM

Anyone mind posting a link to the copy waza tweak having some troubles finding it :rolleyes: and would be interested in trying it.

BenchZowner

11-01-2007, 01:12 AM

gomeler;2526677']Just curious, has anyone run a few 32M runs with some HDD monitoring applications to see what sort of hits are being addressed to the disk? This will solve all your questions of how important the drive is. You could also artificially cripple the drive by loading up the drive with some throughput tests while running SPi but this would also affect the memory available to SPi, skewing the results. If I had a second machine I'd try to run the HDD tests myself but all I have is my webdev server right now.

http://i16.tinypic.com/4ky15yv.jpg

;)

KTE

11-01-2007, 02:58 AM

I want you to pay special attention to what I'm going to say next guys...

OPB... you FRACKING GENIUS!! :) :eek: :eek: :eek:

OPB... damn what the hell?! :eek: :eek: :eek:

My God heeeeh! :eek: :wth:

I've no words. :slobber:

I've no words. :slobber:

I've finished my testing for today. Let me get some tranquilizers first.

:worship:

hixie

11-01-2007, 03:02 AM

massman

11-01-2007, 03:10 AM

So, like I said, I did a few 32M runs yesterday, made an excel sheet in which I put the time of every loop and the time difference between a loop and the preceding loop. I can share my excel-file if someone wants it ... not that special, though :p: .

During the testing I've switched to a faster sata drive with more cache. I gained a bit more than 1 second.

http://www.xtremesystems.org/forums/attachment.php?attachmentid=66607&stc=1&d=1193913911

This is the pic with the two segments I talked about. The first area will be explained later, the second one is quite important. It shows the difference in time between the loop24/output and loop23/loop24 ("/" is not the symbol of division). Normally, loop24/output should be about 1 second faster than loop23/loop24, in my case I can't do this. My times would be 2 seconds faster.

http://www.xtremesystems.org/forums/attachment.php?attachmentid=66608&stc=1&d=1193913911

The first segment I talked about is very interesting. It shows an almost lineair decrease.
Line3: ran superpi from C, while pagefile on D. Used 3x632 files and copied from D to C first. IDE hdd for page
Line2: ran superpi from D; while pagefile on D. Used 3x632 and copied from D to C first. IDE hdd for page
Line1: ran superpi from D, while pagefile on D. Used 1 file exactly 3x632MB and copied from C to D first. Sata HDD for page.

I need to do some extra testing before I can jump to conclusions. But there are a few things that catched my eye.
- The faster you get that lineair decrease, the better the time.
- Running superpi from the same harddisk as the pagefile is on will be slower if not properly tweaked, but if well-tweaked = faster.
- Pi from page HDD = slower loop24/output.
- The size of the systemcache is important (the higher, the better), BUT available memory must be the same size of the systemcache to get the best results!

NOTE: Initial time was 14m32s, after some testing I went to 13m59s, now I'm at 13m56s.

mrlobber

11-01-2007, 03:20 AM

May i ask whats going on? :shrug:

I think he just got THE TWEAK working full scale :) Cant wait the report.

hixie

11-01-2007, 03:22 AM

I think he just got THE TWEAK working full scale :) Cant wait the report.

Could it be the CDT-V tweak OPB mentioned before? If thats the case i'm looking forwards to CDT-VI and CDT-VII.

I'll see if i can do some tests on a I-RAM, latter this week.

KTE

11-01-2007, 03:31 AM

Copywaza is an old method - well known and I've ran it countless times before. Gets me about 8-5 seconds drop usually (at a Pi test which runs in 15-10 minutes).

May i ask whats going on? :shrug:
I think he just got THE TWEAK working full scale :) Cant wait the report.:yepp:

Its tricky but it ONLY works if you get it just right and boy does it work well. Damn what a tweak, this is more a miracle! ;)

Took me around 15 attempts over two days to learn it right and EVEN still I've not perfected it no where near to the screenshots I saw of OPB. The guy will slaughter anyones Pi times with this tweak at the same clocks, timings, hardware, and settings. Make no mistake about that.

KTE

11-01-2007, 03:52 AM

Nice to see this back on track again, great work saaya.

Nowhere near Robert, that's crotale :D I do not have so much more info to add but I can sum up what I've experienced so far.

First time I tried the CDT Tweak I had those crazy results that I posted in the previous thread. However after reformat I was never able to get System cache about the same as Avaliable memory again and hence no boost (that's what I believe gives the boost) except a few seconds which could probably be aquired with hipros method aswell. I started using maxmem = 600 and I actually got worse times with that than with 620 so one could do some experimenting there.

Currently I'm playing with P5E3 WS Pro as I got kinda frustrated over not getting this to work again, but I'll get into it once again when I have some time to spare.

//Jon
Sorry Jon I get mix up between you two all the time. :D

Try and make sure you get above 500MB system cache and above 410MB available memory before you run Pi. Below that there wasn't much difference for me.

before

11-01-2007, 04:10 AM

Took me around 15 attempts over two days to learn it right and EVEN still I've not perfected it no where near to the screenshots I saw of OPB. The guy will slaughter anyones Pi times with this tweak at the same clocks, timings, hardware, and settings. Make no mistake about that.

lol

So what about teaching us dude? I eventually tried it and not succeeded in beating my cw method... may be you know something else, something we don't know. I'm not sarcastic, honestly if something exists, I really would like to learn it. :)

dinos22

11-01-2007, 04:17 AM

massman

11-01-2007, 04:25 AM

KTE

11-01-2007, 04:28 AM

lol

So what about teaching us dude? I eventually tried it and not succeeded in beating my cw method... may be you know something else, something we don't know. I'm not sarcastic, honestly if something exists, I really would like to learn it. :)I followed the same routine you did from reading here late on. Kev did all the work and my utter thanks to him for putting up with **** when he could do and new what no one else here did. Rather than keeping it hidden he spoke openly and tried to help everyone get and beat his times. Salute to that. :up:

If I get hold of a Phenom straight after release and someone else has one to test as comparison then I'll show you the time difference it can make if done properly. Pretty damn sure about that. :) Just run it as good as you can and try to keep variations out of it as little things can make it fluck. You've seen my 8M results above with a Presy Celeron (similar to your 32M) and I'll show you my new test results for CopyWaza and CDT. Just at work yet. :D

Did you get your system cache and available memory above 450MB or 500MB each before running Pi? You need to do that, read kevs words properly honestly. He 'aint kidding. You need to balance both values and have them as high as possible or it doesn't quite show its true nature. :D

There's no doubt the tweak influences hardware caches, buffers and prefetch. It decreases latency very much (I've only tested on a slow CPU yet, it may give less drops in time on a faster core).

BRB

dinos22

11-01-2007, 04:35 AM

KTE
it would be nice to see your conroe 3.6Ghz times
that's the best indication and comparison points for this

Zeus

11-01-2007, 04:41 AM

Man i hope you're right.

I 'm not sure you can replicate the very same thing with a C2D setup though.

Last night i really tried hard, got 527984 for available memory and 532088 for Systemcache, which to me is close to supreme but i did not see the accelaration in the third group of loops as it should.

I'm not done testing myself as i still have faith in this tweak and hope to get it working properly on my C2D setup.

Reading all OPB's ramblings on this subject makes me feel like a complete n00b, the study he made in SuperPi calculation time is nothing short of awesome.

Just trying to comprehend all this complicated stuff takes some time.

mrlobber

11-01-2007, 04:45 AM

...You need to balance both values ...

That's the goal I've failed to reach with my 3.6Ghz C2D testing so far.

KTE

11-01-2007, 04:54 AM

Sorry dinos22 I don't have a C2 here, can get one probably in 10-12 days (and your tempting me now seriously :D ) but I don't see a need for the hassle and am not planning on getting one soon. Got a load of old parts to sell for now. Maybe get a Penryn late March then to test>sell but for now, waiting for Phenom if all goes right. FWIW that Celeron is also my work system, used to be mine but I sold it to work place (where I still use it) and I only have a personal lappy for now (this isn't my home country either).

Zeus: I doubt it'll make anywhere near the same difference with faster cores (more powerful cores) but it certainly will make a difference over a top CW. I need to run some tests myself on various platforms first. I have a P4 chip lying around somewhere but again its completely in parts, some missing and no conroe either.

BTW the forums timing out for me repeatedly. :(

massman

11-01-2007, 05:14 AM

Did you get your system cache and available memory above 450MB or 500MB each before running Pi? You need to do that, read kevs words properly honestly. He 'aint kidding. You need to balance both values and have them as high as possible or it doesn't quite show its true nature. :D

The higher the better?

I'm running with systemcache over 560MB and Available memory over 550MB.

BeardyMan

11-01-2007, 05:24 AM

The higher the better?

I'm running with systemcache over 560MB and Available memory over 550MB.

no, otherwise you wouldn't been setting 600 or 680 as a max, then you could just leave it like that 2048/4096 etc..

I've done some more testing ,i lend a seagate 160GB 16Mb hdd yesterday 7200rpm and compared it to the same disk but 8Mb and a maxtor 5Gb wich has i believe no buffer

maxmem 680 + no pagefile + 632mb file + performance set to systemcache + 160Gb 8Mb = average gain of 0.7secs in SPI 1M on my AMD rig.

maxmem 680 + no pagefile + 632mb file + performance set to systemcache + 160Gb 16Mb = average gain of 0.9secs in SPI 1M on my AMD rig.

maxmem 680 + no pagefile + 632mb file + performance set to systemcache + maxtor 5gb = average gain of 0.2/3secs in SPI 1M on my AMD rig.

the best HDD woudl be this,
150 GB, 10.000 rpm, 4.5ms Zugriffszeit, 16MB Cache,

10k rpm instead of 7200
4.5ms instead of the 8.9/8.5 and ofcourse 16mb cache.

KTE

11-01-2007, 06:40 AM

OK. You saw my old CDT tests right HERE (http://www.xtremesystems.org/forums/showpost.php?p=2524781&postcount=26).

Stock 8M best time was around 10m25s
With best CopyWaza was around 10m20s
With best CDT was around 10m15s

You can't beat those with that setup, doing those things, that way. :p:

New testing:

http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/Setup.png

Changed 2nd HDD-> Now both have same cache/size.
Different monitor :p: and res is smaller.
Same version of OS (a fully working one although this is an older install).
Disconnected all USB, optical, drives etc that are extra, disabled everything extra in BIOS.
Page-file kept consistent at 384-384MB all throughout testing: http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/pagefile.png
You can see my normal services running here: http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/Services.png
You can see my bench setup services running here (1): http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/Services-oc.png
Processes/Memory/Cache at bootup: http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/bootupmain.png

As you can see from the stock results, today this system was faster than ever, even yesterday (by 4-5 seconds). More Windows tweaks. :cool:

Straight jump to the testing. Saaya asked (not verbatim) "Can you reproduce the results?"
With testing a millisecond bench, how many people can get their fastest record results every time they try it? I think we can agree, that you run it time and time again but they are usually one offs a bit better than the rest. But he means to reproduce the gain by a tweak. :) (Spi -1ms makes a lot of difference as loops go on and even how fast you hit Enter makes difference ;) )

I can replicate all of them with each technique.

Still I complained "I can't get above 345MB/345MB". That I need to do to replicate Kevs results with that technique.

Experiment? You have to count everything, even your fudged results. These tests were ran straight one AFTER another -> no extra results missing from here. This is all that I ran. So nothing is hidden.

1st CDT test I ran this morning before work (383MB/380MB :) ): 10m16.266s
http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/8M-10m16s.png

2nd CDT test ran (381MB/381MB): 10m 14.043s
http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/8M-10m14s.png

You have NO idea how chuffed I was at that time. It is impossible with that system without CDT. :D Don't go away yet though...

3rd CDT test ran (386MB/384MB): 10m 10.768s
http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/8M-10m10s.png

:D

4th CDT test ran (386MB/384MB): 10m 9.746s
http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/8M-10m9s.png

:eek: (sorry messed up the window size and it wouldn't expand up later :shakes: )

Crazy gains. 10m25s to 10m9s. At that time I still did not manage to get high cache (etc). This was my best 8M/32M time uptil now (the rest were ran on other days higher clocks): http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/Spi-complete.png

Just wait till what happens next... after a while.

Now to CopyWaza for a brief moment:

I already KNOW what I get with CW. We're not testing that here because its an old tweak. One thing I messed up is I used a 3.74GB file instead of 4GB. I doubt its a huge difference (from previous experience on this setup): http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/winrar4gb.png

3 runs and the slowest was 10m 18s and the fastest was this @ (383MB/153MB): 10m 16.647s: http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/10m16s-copywaza.png

THEN, fed up, I tried CDT again 3 times. I wanted to get 500MB/500MB like Kev...

5th CDT test ran (420MB/524MB): 7m 44.188s
http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/8M-7m44s.png

Holy fork :eek:

It dropped 7 seconds EACH loop consistent! :shocked:

You need to know one thing.. I CANNOT run the first result below 7 seconds in ANY case what-so-sever (Error non convergent hundreds of time) UNLESS I did CDT and the System cache went above 500MB.

I doubted that result a bit :shakes:

Still not perfect like Kevs, so I tried again for a run->> look at the time upper right corner for all shots.

6th CDT test ran (431MB/530MB): 7m 27.533s
http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/8M-7m27s.png

:eek: :eek: :eek:

Simply unbelievable. That's like you testing 100 ES Phenoms being sore clockers at 3GHz max and cold bug -50C and then suddenly, you chose to test another one and it reaches 7GHz no cold bug :D

That is far far below the best. I was just shell shocked completely. What is happening? It dropped 7-10 second each loop!??! (yes I ran a mobile stop watch because I couldn't believe it :D )

Now I NEVER ran CDT again, but straight after the above finished (and I recorded it) I ran 8M again.

7th CDT test ran (411MB/520MB): 7m 27.604s
http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/8M-7m27s6.png

:eek:

These are not the fastest. It can be beat by using this method better although I ran out of time. Someone explain to me what is happening? :confused:

Also, keep your eye on the 32M time (stock of today). Tomorrow morning I'll go in again and test 32M after CDT.

I believe you can possibly gain anywhere from 10-30 seconds with a C2 in 32M if done perfectly. Maybe more, but I've not tried.

BTW the clocks of the CPU/RAM are not highest, they're just what I prefer. They can be thrown higher to 3550 stock volts (which is 2/3 benchable) and 3700/435 2-3-3-6 stock volts on all components, max. The clocks above I used are ALL only altered in BIOS at bootup (no PLL tweaking) and because they're full stable (no errors possible).

Hope you had a good read. :D

mrlobber

11-01-2007, 06:52 AM

WTF :eek: :eek2:

From 10 minutes down to 7? There must be aliens in your system somewhere :shocked:

Edit: it seems for your 5th CDT test, the image shows your 6th run instead.

Another question: as far as I understand, throughout all the testing (those 6 CDT runs and all of the copywaza), you didn't reboot, did you?

JMKS

11-01-2007, 07:11 AM

@KTE: so it seems You absolutely MUST rerun it with C2D.
I will compare it to the Celeron "Conroe-L" - simple CW can produce ~1sec difference in 1M and ~40s in 16M - with C2D the difference is nothing [1M] to ~3sec [16M].
So it depends on many things, but yes, 30% difference is absolutely worth analyzing :up:.

Eldonko

11-01-2007, 07:45 AM

10 minutes to 7, WTF.. That's very strange I think, could you try 32M?

Mord

11-01-2007, 08:13 AM

KTE, mother of god! What did you do, man?! WTF did you do to that poor little app SPi?! OMG, OMG, OMG!!! This is unbelievable, simply jaw-dropping!

Must resist, must not bench, must get life, must keep sanity... Damn so hard... I can't take it - I'm going benching. The force got me! Damn, where's my cigarettes?

hixie

11-01-2007, 08:14 AM

KTE

11-01-2007, 08:20 AM

Edit: it seems for your 5th CDT test, the image shows your 6th run instead.I've updated it with the correct link. Thanks! :)

Another question: as far as I understand, throughout all the testing (those 6 CDT runs and all of the copywaza), you didn't reboot, did you?Nope. No reboot.

If you discount the second set of results for whatever reason (below 10m) then even with the first set I gained from 10m25s to 10m16s with CW and 10m25s to 10m9s with CDT today. Big worthwhile difference to someone competing.

@KTE: so it seems You absolutely MUST rerun it with C2D.The Pentium E2160 is on the core muarch, right? I can get that and a P5B-D today but it 'aint till weekend that I drive over to pick it up at earliest. Still I don't really want to waste cash right now. Plenty of others have one to test.

For now, go ahead testing Core 2 and any other CPUs. I won't be testing the Pres/Cely anymore (unless someone asks for it). The CPU still works fine but I don't want it to be about discussing older CPU results. I just did it to test teh tweak across various platforms really. :D

You guys who've tested CDT and say it failed, what are your best 32M 3600MHz C2D times and at normal voltages/latencies/DDR speeds?
(some config you can repeat today like DDR1000 4-4-4-12)

You have to keep things constant if you want to test "just" a technique.
Same CPU MHz, same DDR speed, same latencies, same ratio, same strap if possible, same OS, same monitor resolution, same all tweaks, same pagefile, same maxmem, same order, as similar as HD cache as possible.

FWIW, while 8M was running at the last loop I moved the mouse 4 cm to the right and added 7 seconds from the normal time it should've been for a loop. That's how freaking sensitive ms benches be. :yepp:

dduckquack

11-01-2007, 08:20 AM

holy milking cow

hixie

11-01-2007, 08:24 AM

I've updated it with the correct link. Thanks! :)
Nope. No reboot.

If you discount the second set of results for whatever reason (below 10m) then even with the first set I gained from 10m25s to 10m16s with CW and 10m25s to 10m9s with CDT today. Big worthwhile difference to someone competing.

The Pentium E2160 is on the core muarch, right? I can get that and a P5B-D today but it 'aint till weekend that I drive over to pick it up at earliest. Still I don't really want to waste cash right now. Plenty of others have one to test.

For now, go ahead testing Core 2 and any other CPUs. I won't be testing the Pres/Cely anymore (unless someone asks for it). The CPU still works fine but I don't want it to be about discussing older CPU results. I just did it to test teh tweak across various platforms really. :D

You guys who've tested CDT and say it failed, what are your best 32M 3600MHz C2D times and at normal voltages/latencies/DDR speeds?
(some config you can repeat today like DDR1000 4-4-4-12)

You have to keep things constant if you want to test "just" a technique.
Same CPU MHz, same DDR speed, same latencies, same ratio, same strap if possible, same OS, same monitor resolution, same all tweaks, same pagefile, same maxmem, same order, as similar as HD cache as possible.

FWIW, while 8M was running at the last loop I moved the mouse 4 cm to the right and gained 7 seconds from the normal time it should've been for a loop. That's how freaking sensitive ms benches be. :yepp:

:rofl: So what did you do different from the original posted CDT. Or did you just do it more carefully? and tried to max system cache and memory?

T_M

11-01-2007, 09:42 AM

Do you know why suddenly your cache etc went up from 380?

Zeus

11-01-2007, 12:56 PM

Come on KTE, get yourself a C2D setup and show the same sort of gains for crying out loud!

Come over to my place with your harddisk and do it on my machine if there's no other option. (i'm in Holland hehe)

Geez, if you can get the same sort of gain on a C2D setup you will be king of the hill for sure!!

kiwi

11-01-2007, 01:55 PM

That's the goal I've failed to reach with my 3.6Ghz C2D testing so far.

If balance means start 32M when system cache = available memory then that is easy to do with CDT, at least I always can :)

dinos22

11-01-2007, 01:59 PM

OK. You saw my old CDT tests right HERE (http://www.xtremesystems.org/forums/showpost.php?p=2524781&postcount=26).

Stock 8M best time was around 10m25s
With best CopyWaza was around 10m20s
With best CDT was around 10m15s
.................................................. ......

Now to CopyWaza for a brief moment:

I already KNOW what I get with CW. We're not testing that here because its an old tweak. One thing I messed up is I used a 3.74GB file instead of 4GB. I doubt its a huge difference (from previous experience on this setup): http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/winrar4gb.png

3 runs and the slowest was 10m 18s and the fastest was this @ (383MB/153MB): 10m 16.647s: http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/10m16s-copywaza.png

THEN, fed up, I tried CDT again 3 times. I wanted to get 500MB/500MB like Kev...

5th CDT test ran (420MB/524MB): 7m 44.188s
http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/8M-7m44s.png

Holy fork :eek:

It dropped 7 seconds EACH loop consistent! :shocked:

You need to know one thing.. I CANNOT run the first result below 7 seconds in ANY case what-so-sever (Error non convergent hundreds of time) UNLESS I did CDT and the System cache went above 500MB.

I doubted that result a bit :shakes:

Still not perfect like Kevs, so I tried again for a run->> look at the time upper right corner for all shots.

6th CDT test ran (431MB/530MB): 7m 27.533s
http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/8M-7m27s.png

:eek: :eek: :eek:

Simply unbelievable. That's like you testing 100 ES Phenoms being sore clockers at 3GHz max and cold bug -50C and then suddenly, you chose to test another one and it reaches 7GHz no cold bug :D

That is far far below the best. I was just shell shocked completely. What is happening? It dropped 7-10 second each loop!??! (yes I ran a mobile stop watch because I couldn't believe it :D )

Now I NEVER ran CDT again, but straight after the above finished (and I recorded it) I ran 8M again.

7th CDT test ran (411MB/520MB): 7m 27.604s
http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/8M-7m27s6.png

:eek:

These are not the fastest. It can be beat by using this method better although I ran out of time. Someone explain to me what is happening? :confused:

Also, keep your eye on the 32M time (stock of today). Tomorrow morning I'll go in again and test 32M after CDT.

I believe you can possibly gain anywhere from 10-30 seconds with a C2 in 32M if done perfectly. Maybe more, but I've not tried.

BTW the clocks of the CPU/RAM are not highest, they're just what I prefer. They can be thrown higher to 3550 stock volts (which is 2/3 benchable) and 3700/435 2-3-3-6 stock volts on all components, max. The clocks above I used are ALL only altered in BIOS at bootup (no PLL tweaking) and because they're full stable (no errors possible).

Hope you had a good read. :D

OMG I JUST FELL OFF THE CHAIR :shocked:

Zeus

11-01-2007, 02:04 PM

mrlobber

11-01-2007, 02:19 PM

Having the available cache matching the system is cache is easy, i accomplished that in each and every copy waza session i did, most often available cache is even exceeding system cache by a small amount.

This NEVER happens for me, never. The only way I can get available memory > system cache is when I rar a big file in superpi's folder right before the 32M run. In all other cases (copywaza, CDT, whatever) I see a pattern like available memory + 20 Mb = system cache. The OS is the default XP SP2 Corporate install with initial "commit charge" after reboot showing between 60 and 70 Mb (I wonder where people get those 100+ Mb figures with "clean install"? :confused:) Right now, testing Windows 2003 server with the same results.

Have tried both IDE as well as SATA HDD's, 2Mb and 8Mb cache as well.

Kasparz

11-01-2007, 02:24 PM

Does anyone tried using usb flash because of fast access times or ramdisk for spi?
Second, does anyone tried writing .bat script that automatically copy files?

mrlobber

11-01-2007, 02:34 PM

Does anyone tried using usb flash because of fast access times or ramdisk for spi?
Second, does anyone tried writing .bat script that automatically copy files?

Running Pi from ramdisks do not make a difference, this has been tested before. However, placing pagefile on ramdisk might give you a chance to optimize your copy waza better, so in this sense it might help.

Bat scripts didn't make a difference for me. Then again, nothing has made it so far (for me) :p:

Kasparz

11-01-2007, 02:38 PM

Well, if you find best filesize/pagefile size then you could make script to automatically copy files as every preparation takes time and if you're running with temporary cooling solutions then you can't control temperature and do alot of tweaking at the same time.

MaSell

11-01-2007, 02:51 PM

Does anyone tried using usb flash

Do it :D . Times like with Pentium 3 :D

KTE

11-01-2007, 07:41 PM

Hi fellas

I've experienced a heavy unexpected grievance so I'm trying to not stay in solitude so I start thinking about it and do something other.

I ran Super Pi 32M like you guys run. Here are the quick results.

Same exact setup as last time
Maxmem=600
Pagefile 384-384
Fully tweaked Windows

I started the day with this as my fastest (no extra tweak): 47m 54.694s
http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/SPi-complete2.png

1st stock run - (375MB/135MB): ~47m 15s

2nd run - Free'd full RAM - (554MB/67MB): 45m 00.613s
http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/32M-45m.png

A screenshot during CDT working in the background: http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/high.png

1st and only run with CDT @ (523MB/533MB): 37m 00.513s
http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/32M-37m.png

Latest figures:

http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/LatestSPi.png

Roughly a 10 minute drop there.

===============

T_M: It usually takes me 3-5 tries to get above 500MB cache running what Jon/Kev run.

dinos22

11-01-2007, 07:44 PM

so are you running LargeSys Cache on slow and fast runs on only fast runs as that will give you a massive boost in times on its own

dduckquack

11-01-2007, 07:48 PM

10mins... the hell!

how much with cw only?

still waiting for the c2d results though

KTE

11-01-2007, 07:59 PM

so are you running LargeSys Cache on slow and fast runs on only fast runs as that will give you a massive boost in times on its ownThe reg tweak? Thats there by default on every run, even on the 47m ones. The only difference between each run is like I stated: no CDT tweak / some other tweak / CDT tweak. Never ran CW this time at all, no time and it doesn't gain me more drop than 47m to 45m, so useless for me.

KTE

11-01-2007, 09:21 PM

Here's a good comparison of that setup. Best 32M times: http://www.bleedinedge.com/forum/showthread.php?t=17725

Look at Intel 32M number 52 and 54. They are closest to my specs.

My system at 3550MHz/DDR424 is the same in PCMark 04 to a P4 2.8GHz Northwood stock. You can see some comparison in the review here too: http://www.sharkyextreme.com/hardware/cpu/article.php/3398581

52. Just Learnin - 34min 30sec 310ms - Pentium 4 3.2Ghz Northwood@3600mhz - P4 3.2 C @ 3.6 GHz, X800 XT AIW
54. Snafu - 42min 11sec 990ms - Pentium 4 3.0Ghz Northwood@3000mhz - P4 3.0C @ 3000

Look at how I was 5 minutes slower than snafu before today and now after trying CDT I'm 5 minutes ahead with a much weaker system. I need 3750MHz on my setup usually to match his 3GHz P4 in performance. That's just beyond belief. I have a feeling this procedure can affect your times in other number benches too.

hipro5

11-01-2007, 11:06 PM

KTE, do you use registry's LargeSystemCache=1 ?.....:)

mrlobber

11-01-2007, 11:09 PM

Spent another 5 hours tonight on my C2D setup, the same as here http://www.xtremesystems.org/forums/showpost.php?p=2511079&postcount=117, except I switched for a 160Gb 8Mb cache HDD.

Still no difference between copywaza and CDT for me lol :shrug:

The only thing I could feel having different from KTE's testing is, of course, the hardware, as well as I'm using a single drive. Then again, OPB is talking about CDT on a single drive with 2 partitions, at least that's the info I have so that shouldn't make a difference between a successful / unsuccessful tweak.

Hipro, it seems KTE is running LSC=1, that's what he said from the very beginning :)

Too bad I don't have any Netburst cpu to test, would be very encouraging if I could repeat KTE's performance :cool:

hipro5

11-01-2007, 11:35 PM

Come on you guys.....Noone with C2D to test at 3600MHz?....Were have you gone?... :( :D

dduckquack

11-01-2007, 11:48 PM

no available hardware atm :(

T_M

11-02-2007, 12:50 AM

KTE, do you use registry's LargeSystemCache=1 ?.....:)

The reg tweak? Thats there by default on every run, even on the 47m ones.

see above.

As for testing it, im too busy atm trying to catch your Ultra 3D'06 score to worry about Pi George :P

KTE

11-02-2007, 12:53 AM

Yep. These are the standard registry settings I use on that OS (pagefile has been changed back again though).

http://img64.imageshack.us/img64/125/registryvv6.th.png (http://img64.imageshack.us/my.php?image=registryvv6.png)

I'm in redundant mode for at least a week now. Will see what I can do after that.

What's a very good (full tweaked/copywaza) 32M time for C2D 3600MHz? (specify FSB/board/RAM timings/divider/speed as well please)

hipro5

11-02-2007, 12:56 AM

see above.

As for testing it, im too busy atm trying to catch your Ultra 3D'06 score to worry about Pi George :P

Which one?.....the LOW 214xx one or my "other one"...... ?...... :D :p:

mrlobber

11-02-2007, 01:04 AM

What's a very good (full tweaked/copywaza) 32M time for C2D 3600MHz? (specify FSB/board/RAM timings/divider/speed as well please)

You can look in the Superpi low clock challenge thread, but this thread @ OCX (http://www.ocxtreme.org/forumenus/showthread.php?t=961) imho is much better structured and contains the Holy Grail - teh uber OPB 3.6Ghz CDT run (12m39s) as well :)

T_M

11-02-2007, 01:05 AM

I had suspected you have an "other one", but for the time being your 21.4 will do me.
At least if i do catch it let me have a WR for 24hrs please ;)

hipro5

11-02-2007, 01:09 AM

I had suspected you have an "other one", but for the time being your 21.4 will do me.
At least if i do catch it let me have a WR for 24hrs please ;)

OK.....NO worries......I'm bluffing.....I got no "other" one.... ;) :D

KTE

11-02-2007, 02:12 AM

You can look in the Superpi low clock challenge thread, but this thread @ OCX (http://www.ocxtreme.org/forumenus/showthread.php?t=961) imho is much better structured and contains the Holy Grail - teh uber OPB 3.6Ghz CDT run (12m39s) as well :)Thanks.

That's more of a $$$ hardware competition than Super Pi tweak contest. Different clocks, different latencies, different straps, different boards, different CPUs, different RAM, different volts even different DDR - everythings just different apart from CPU clock speed. I'm not going to waste cash on buying hardware like that now with Phenom around the corner, having sold my old, that's for sure. But no promises, I may have a C2D build by sometime next week at the earliest.

Guys will need to run their C2D at 3600, ≤500 RAM speed, Y latencies, normal 24/7 usable volts, which you can reproduce at least 3 times, the best time you get and others can then have a crack at that as well to compare with different tweaks what times they are capable of. Then we can test the tweak and not just another number competition because I won't be doing that, simply put, no matter what the outcome. Otherwise I find nothing to compare to yet but the absolute fastest numbers which I won't be doing.

A tweak can get gains at ANY speed and latency over a no-tweak. Once you can find you get gains at low GHz you can increase it for your comps and get more .. later.

Zeus

11-02-2007, 02:24 AM

Having a C2D running 32M with the memory running <500Mhz will give dull results, system will be bottlenecked and tweaks might not take full effect as the system is starved for bandwidth imho.

I think 9x514 at 1:1 with 4-4-4 timings on a P35 platform won't be that hard, bare quadcores perhaps.

Even better result will show with 5:6 or 4:5 divider but that's something not everybody can do at 4-4-4 timings but at 5-5-5 again it will not be xtremely hard if you have some half decent memory.

mrlobber

11-02-2007, 02:24 AM

You can look at the middle of the pack - 514x7 with 1:1 ratio on ram is easily done on air and just about on any E6600 and non-value DDR2 configuration - that's what I'm also using on my so far unsuccessful CDT C2D testing.

400x9 1:1 would be an even better choice, of course (for "everyones availability").

KTE

11-02-2007, 03:05 AM

Having a C2D running 32M with the memory running <500Mhz will give dull results, system will be bottlenecked and tweaks might not take full effect as the system is starved for bandwidth imho.I understand where you're coming from, but if C2D gives a "dull result" at 3600/1000, then you can say the same thing about the Celeron bro. Its more than crap at 3360/400. Do you think I would get more CDT gains if I went higher up in clock speeds? My honest opinion is no, the extra would be clock speed gains.

Let me re-iterate what I'm saying: I didn't say "I'm going to get the best time" with that Celeron because I was not going to run the volts/latencies/speeds others who chose to compete in a competition will - others at higher clocks, lower latencies, better chipset, higher volts, same hardware will beat those times.
However, I did say "with that setup and those timings/clocks/volts, you won't even come close to the best time on that hardware until you do CDT".

Testing a tweak has absolutely nothing to do with the maximum best time with your maximum best clocks, thats for you guys like before, Onepagebook (etc) in their competing. Testing a tweak is about how much the tweak makes a difference in SPi. I 'aint running a horse race here but only trying to replicate "what effect CDT has on SPI". Only testing I need to do is:

At the same clocks, same everything, no CDT vs CDT > what is the time difference?
At the same clocks, same everything, no CW vs CW > what is the time difference?

I ran an old Celeron, it made gains unforeseen previously. I also have a P4 that I can run later. But now even I'm curious to see if C2D gains anything. But please try to understand, spending $800 "at impulse" as a uni student on a scholarship in a different country is not something we do, especially when we can't make our money back. :)

And BTW. Maybe someone didn't understand what I hinted to before.
Yesterday evening around 6pm I found out that the lady I loved for 7 years, who was everything to me my friends and family will tell you, fiancee for 9 months, marriage was set to be in December.. was killed in a car crash on the 29th in the US. So I'm not in the best of moods, honestly. Its a bit hard to put a smiley up. I just don't want you to think I'm ignoring your responses.

hipro5

11-02-2007, 03:23 AM

Keep in mind that different tweaks that used with old hardware and they WORKED GREAT, now with the C2D CPUs, ain't THAT good anymore or they do nothing.....Keep this in your mind for start...... ;)

EDIT: No shi(f)t..!......I JUST read the last sentences KTE...... :(
BAD....BAD things happening..... :(

mrlobber

11-02-2007, 03:45 AM

KTE, sorry for you man :(

Zeus

11-02-2007, 03:55 AM

First of all, my condelances, that's more than just sad.

Without trying to be rude, wheter you're out on getting the best time or not is not the issue here, we're still talking the effect of this tweak. ;)

I'm pointing out that if a system is starved for bandwidth, a tweak that gives a boost in data throughput for example will be executed with the handbrake on if you know what i mean.

Running 32M at 3.6GHz with 400FSB does well on a 965 chipset board (dividers are a must of course) but try this on a P35 based board and you'll be limited due to the "low" FSB speed.

Once limited on FSB, tightening the memory timings and subtimings does nothing untill one starts raising the FSB, gains wil reappear, from my own experience at least.

This is just an example and i don't know if this is relevant for the CDT tweak but i'd be rather safe than sorry.

That's why i suggested 514x7 1:1 which should be doable with 4-4-4 timings for 90% of the C2D owners around here.

Like Hipro said, some tweaks yield better results on older or other hardware, this C2D stuff might have a different way to handle system memory, sytem cache and pagefile than older stuff, hence different result outputs.

Not saying this is a fact but it's a possibilty one have to take into consideration.

KTE

11-02-2007, 04:05 AM

True. Its always a likely possibility. We'll have to see.

Already got three parts (2xsame drives/2x1GB RAM/PSU). Probably get the next two as soon as I get a phone call from the main man himself. Been told its an E6850 and P5K-E WiFi but I don't know yet for sure.

It's OK. But I tell ya what... love hurts -> deeper than deeep itself. Man I would've sold myself for her ah! :(

andre X_X

11-02-2007, 04:14 AM

Presscot and Core Technology is totally different. thus, the same tweak will have diffrent result....

KTE

11-02-2007, 05:24 AM

Perfect.

Two boards I'm offered; P5K-E WiFi and P5K Premium Black Pearl.

Just checked and both boards don't have the VGA port I need for the work CRT I'll be using neither the PS/2 keyboard/mouse there. Doesn't look like I can get another board. Only other option is GA-965P-DS3.

Presscot and Core Technology is totally different. thus, the same tweak will have diffrent result....We know they're both totally different but it didn't make a difference when using CopyWaza, did it? Inconclusive is the more accurate term here. Maybe a few seconds drop is still possible even though it is not as good but I expect a 4-15 second drop even still. That's for you C2D owners to test.

mrlobber

11-02-2007, 06:07 AM

Presscot and Core Technology is totally different. thus, the same tweak will have diffrent result....

Well, that's the obvious thing we all know so let's keep focusing on the discussion of results (if we have them) :)

Monstru

11-02-2007, 07:38 AM

First of all my condelances KTE, there are moments in a man's life when you are really hurt and stripped of all you thought you will never loose. Keep your head up man, even if that is one of the worst things that can happen to a man :(

Second, congratulations for your test, they are really inspiring and when I get the free time I will try your tests on a C2D and see if it works for me. If it does, that that's a great deal my friend. The only thing is that I cannot get hipro's words out of my mind, regardind tweaks on Yorkfield :(

The last thing.... I saw that you guys are talking about the HDD's used and I see that you are not using the best HDD possible to maximize and level your results. I did a test about the impact the HDD has in SuperPi here (http://www.crazypc.ro/forum/showthread.php?t=7348) and it was confirmed to my what everybody knows allready, that the HDD's speed and cache really matters. It also revealed something weird for me, the fact that RamDisk doesn't work better than a Raptor and that RamDisk works different depending on what HDD you are using (and I find this discovery to be very weird).

Dfi Infinity P965-S
E6600 400x8
2x1gb Mushkin XP6400 4-4-3-10 @ 800 4-4-4-10 2T
WinXP SP2 with unnecessary services, let's say basic tweaked. No CW, no CDT, no "special" tweak at all.

Seagate Barracuda 80Gb | 7200.7 ST380011A ATA133 2Mb Buffer

Seagate Barracuda 250Gb | 7200.10 ST3250410AS Sata II 16Mb Buffer

Western Digital Raptor 74Gb | WD740ADFD-00NLR1 Sata 16Mb Buffer 10000rpm

I ran Spi 1M and Spi 4M 5 times on each HDD and RamDisk to get an ideea about how a medium result looks. I only posted the best 3 times.

Seagate ATA133:

Spi1M : 14.203, 14.187, 14.188
Spi1M : 14.187, 14.172, 14.188 - RamDisk
Spi4M : 1.20.125, 1.20.125, 1.20.125
Spi4M : 1.20.047, 1.20.078, 1.20.078 - RamDisk

Seagate Sata II

Spi1M : 14.187, 14.157, 14.156
Spi1M : 14.157, 14.172, 14.156 - RamDisk
Spi4M : 1.19.985, 1.19.906, 1.19.969
Spi4M : 1.20.063, 1.20.015, 1.20.047 - RamDisk

WD Raptor Sata

Spi1M : 14.156, 14.156, 14.156
Spi1M : 14.157, 14.172, 14.156 - RamDisk
Spi4M : 1.19.984, 1.19.906, 1.19.969
Spi4M : 1.19.953, 1.19.968, 1.19.969 - RamDisk

What I got was the Raptor's might in Spi 4M and continuity in it's results. Maybe that is the HDD to use for testing tweaks, giving very similar result most of the time.
I hope I am not off-topic, I thought that you might find this little piece of information useful in your testing. If it is not please delete :)

Good luck with your testing, I am sure that there are lots of us waiting to see where this story is going to.

Zeus

11-02-2007, 07:54 AM

KTE, what do you mean by VGA port?

If you're looking for a high-end LGA 775 board with onboard videocard you'll be hard pressed i reckon, just the cheapest PCI-E videocard will do in your case.

Monstru, thanks for your test results but the only suprise is that ramdisk isn't any faster than any harddisk.

Raptor has the lowest acces time, probably due to being a 10K rpm disk, every single read and write action SuperPi will do to the harddisk will be delayed to the lowest possible level by using a harddisk with low accestime.

Remember the temp files SuperPi is creating in the SuperPi folder after each calculation? ;)

Monstru

11-02-2007, 08:04 AM

Raptor has the lowest acces time, probably due to being a 10K rpm disk, every single read and write action SuperPi will do to the harddisk will be delayed to the lowest possible level by using a harddisk with low accestime.

Sorry, I don't understand correctly, you are saying that Raptor gets the worst or the best time in Spi???

Zeus

11-02-2007, 08:18 AM

hipro5

11-02-2007, 09:03 AM

Most likely the best result due to low accestime.

The higher the accestime , the higher the delay --> slower calculationtime.

Low accestime means less calculationtime wasted by having to wait to acces the harddisk.

Raptor has a very low accestime, probably one of the lowest of all Sata harddisks around atm.

There MUST be something else because how do you explain that with Gigabyte iRAM with a access latency = 0 :D , the 32M is WORSE than a regular HD....?..... :confused: :D

Zeus

11-02-2007, 09:30 AM

There MUST be something else because how do you explain that with Gigabyte iRAM with a access latency = 0 :D , the 32M is WORSE than a regular HD....?..... :confused: :D

Do you have the SuperPi folder on your iRAM?

How much worse is it? Within error margin or really a couple of seconds or more off?

hipro5

11-02-2007, 09:34 AM

Do you have the SuperPi folder on your iRAM?

How much worse is it? Within error margin or really a couple of seconds or more off?

I had tested it into a iRAM and then again windows instalation and all into iRAM, but it was worse than a "regular" HD...... :) .....Yes ~3 - 6sec loss with the iRAM..... ;)

Zeus

11-02-2007, 09:44 AM

Erm....let me think about that.

Theoratically zero latency (or close to zero, zero latency is impossible) should yield better results.

Was the pagefile on the iRAM as well?

hipro5

11-02-2007, 11:03 AM

Erm....let me think about that.

Theoratically zero latency (or close to zero, zero latency is impossible) should yield better results.

Was the pagefile on the iRAM as well?

It was in and it was out in different testings.... ;)

andre X_X

11-02-2007, 06:42 PM

i've tested the CDT lastnight and it was nothing diffrent with CW
E6600 @ 7x514 on Biostar P35 1:1 CL4-4-4-4

No tweak = 13m 18s
CW = 13m 11s
CDT= 13m 11s

G H Z

11-02-2007, 09:30 PM

.....Yes ~3 - 6sec loss with the iRAM..... ;)

This does work (PF & Pi.exe in i-Ram) on a P4...it's not good on C2D.

Dumo

11-02-2007, 09:46 PM

My deepest symphaty KTE.

dinos22

11-02-2007, 09:56 PM

omg KTE i feel your pain man :(
the best thing to do is not sit around thinking about
THINKING=BAD atm
just get yourself busy for now and all the best. Remember you are not the only one hurting in the world....there are and have been many harder things people have endured....believe me i have :up: you'll be ok :)

dduckquack

11-02-2007, 10:19 PM

sheesh just read about the accident, sorry about your loss man, never easy, i know

Zeus

11-03-2007, 12:33 AM

Can you guys get this done?

andre X_X

11-03-2007, 12:40 AM

i've tested it before, i have similar amount between system cache and available memory.. no gain from CDT IV with copy wazza. both of them have similar result.

G H Z

11-03-2007, 01:30 AM

Damn sorry to hear that KTE...

spainis

11-03-2007, 02:14 AM

Sorry KTE :(

But anyone know why KTE at startup gets that explorer commits more than 40Mb?:confused:

:)

KTE

11-03-2007, 02:40 AM

It's OK. Thanks everyone, very appreciated and comforting when a mountain suddenly falls on you. Nice to know we're humans at the end of the day and someone you didn't know but interact with over a network still retains some humanity and care. :)

No C2D. In fact what I'll most likely be getting won't be till next Sunday now (I have to go away 4 days (funeral) then come back to an aunts for 2 days then drive to meet my uncle 7 hours away from there (who'll give me the parts as he has them)). It's looking like an E6750 now, Zalman CNPS9700, with Gigabyte P965 DS3, a Samsung Syncmaster TFT, Corsair XMS2 PC-6400 C4 dual channel kit (black heatspreaders), an X800 GT or a 2600 XT, 2x Western Digital 80GB Caviars and a Seasonic S12 650W. They won't be mine though, I'm not paying a dime, they're his and I have to return them soon after in working condition.

But anyone know why KTE at startup gets that explorer commits more than 40Mb?That's something I'm stuck on for a long time with Microsoft. Their explorer.exe gives me over a million page faults over an hour or two of running. :(

I did some testing with 1M on the P4 Celeron. I expected a second or two at most. Findings are very good and consistent. I've noticed the pattern too and tested to see how it works. Look at my above old 1M best time. I'll post my new ones soon. :)

My Prediction:
I "predict" the guys who are getting ~500/500 (balanced) on a C2D won't see any more gains. Not higher than 500, not lower, balanced. That's the sweetspot with the P4Celeron anyway. Any higher gets higher time, any lower gets higher time too. But I'll test it soon to see for myself.

JMKS

11-03-2007, 03:05 AM

It's looking like an E6750 now (...)with Gigabyte P965 DS3
Not that good... E6750 has a x8 max multiplier [and wall ~500 probably, so no 514x7] and Giga i965 has "strap" >400 or so, with 450x8 the result will be absolutely non-comparable to other [like 9x400 with any board or 514x7 with P35 / X38 / QuadGT / DFI Inf 965] 3,6GHz runs... SetFSB may help [boot @ 400, in Windows => 450, x8] but I still will be 450 1:1. But it will of course let us see a real gain with Conroe.

mrlobber

11-03-2007, 03:07 AM

Can you guys get this done?

Standard situation for me in copywaza and/or CDT testing.

KeZzZu

11-03-2007, 04:44 AM

Hi my experimets without copywaza and cdt, later doing with them.

Celeron 440@2.3ghz
Abit ip35-E
OCZ Reaper 6400 2x1g
40gb sata hardisk
Vista 32 bit without any updates on it.

Untweaked vista
http://img208.imageshack.us/img208/8328/sp1muntweakedtq9.jpg

Tweaked somemore on normal mode.
http://img68.imageshack.us/img68/8346/tweak1mzc3.jpg

Safe Mode some tweaks but no mem tweaks becose it wont launch memset on safemode, also it doesnt launch cpu-z got errors only.
http://img217.imageshack.us/img217/2176/nimetnyr6.jpg

Well another thing: Got 0.050 secs off from that safemode run but vista self booted just before screenshot :mad:

KTE

11-03-2007, 07:02 AM

Not that good... E6750 has a x8 max multiplier [and wall ~500 probably, so no 514x7] and Giga i965 has "strap" >400 or so, with 450x8 the result will be absolutely non-comparable to other [like 9x400 with any board or 514x7 with P35 / X38 / QuadGT / DFI Inf 965] 3,6GHz runs... SetFSB may help but I still will be 450 1:1. But it will of course let us see a real gain with Conroe.I realize JMKS. I have no other option TBH, other than that I can't test a C2D setup now and someone else would have to do it instead. If you don't want me to run tests on that for whatever reason then I'd like to be told now please rather than finding out later. :)

While you have a factual point for "absolute time comparisons across platforms" here you are making the one same mistake. Let me make it clear again:

I am not testing absolute performance or my best time! We are testing the gain CDT can provide. People wanted to see that. P4C showed gains beyond reach. SO now they want to see a C2D do it because they compete with that. Pretty simple really. All you need for that is:

A. Stock best time (same all tweaks with same OS/hardware settings)
B. CW best time (same all tweaks with same OS/hardware settings AND CW done)
C. CDT best time (same all tweaks with same OS/hardware settings AND CDT done)

A - B = CW effect

A - C = CDT effect

I am only helping to see if I can replicate the gain I can get on a P4C. I am NOT a SPi competitor, you should know that by counting how many times I've submitted a SPi result or even tried for one and showed it off on XS before 2-3 days ago. :)

*I was after the E6850 but he said its needed elsewhere so the second best option was a retail E6750 (unopened yet).
*Can still get a X38/P35/680i/650i board but I see no point. Then it'd start becoming a competition which is not the reason I'm involved for neither do I have hours to spend messing and tweaking with it. I don't spend much time indoors, I have sporting commitments and work 7 days a week and study 5 days for a PhD too.
*I can get an E6600 B2 as well but it doesn't do plus 400FSB air and thats just booting. 1M/32M stable would be less.
*The board can do 520FSB air and they've used it for 4 days.
*If cooling becomes a problem and I get a really sucky chip I'll throw everything into a -20C freezer (a tried and tested method :D ).
Hope its not a dud though.

[B]Quick SPi 1M

Ever do a really fast run and you're getting happy that it's broken your PB but then you get

NOT EXACT ROUND ?

http://img512.imageshack.us/img512/408/35531merrorws2.th.png (http://img512.imageshack.us/my.php?image=35531merrorws2.png) http://img256.imageshack.us/img256/9431/3533erroryy4.th.png (http://img256.imageshack.us/my.php?image=3533erroryy4.png)

I think we all have had it some time or another.
I get it on the P4C at or above 3553MHz (needs volts which I can't change), definitely if the initial value is at or below 0.681s. There's hardly a run complete at that speed unless its a little slower than the best time. Keep this in mind very firmly. :)

Results

Best times till yesterday:

http://i222.photobucket.com/albums/dd28/KTE__/Prescott_Celeron/LatestSPi.png

Screenshot of that 1M time: (http://img528.imageshack.us/my.php?image=1m35534232335recordvv0.png) 3554MHz - 2x 512MB - DDR423 - 2-3-3-5 (No Maxmem-3x 4.13GB CW-384-384 PF-fully tweaked hardware/OS-813MB memory-768MB cache) = 47.518s (http://img166.imageshack.us/my.php?image=settingstj6.png)

Other best times:
1800MHz - 1x 512MB - DDR215 - 2-2-2-5 = 1m 56.547s (http://img205.imageshack.us/img205/5028/1m18002152225512mbuv8.png)
2800MHz - 1x 512MB - DDR333 - 2-2-2-5 = 1m 10.431s (http://img101.imageshack.us/img101/2949/1m2800333512mbld3.png)
3472MHz - 2x 512MB - DDR413 - 2-3-3-5 (No Maxmem-large PF-normal Windows run) = 54.338s (http://img211.imageshack.us/img211/7718/normal1m3472ue5.png)
3472MHz - 2x 512MB - DDR333 - 2-2-2-5 (No Maxmem-384-384 PF-normal Windows run) = 53.958s (http://img150.imageshack.us/img150/5329/1m34723332225nw8.png)
3472MHz - 2x 512MB - DDR413 - 2-3-3-5 (No Maxmem-384-384 PF-normal Windows run) = 53.547s (http://img128.imageshack.us/img128/3077/1m34722335gu9.png)
3528MHz - 2x 512MB - DDR420 - 2-3-3-5 (No Maxmem-large PF-normal Windows run) = 52.806s (http://img211.imageshack.us/img211/43/1m3528cq8.png)
3360MHz - 2x 512MB - DDR400 - 2-3-3-5 (No Maxmem-384-384 PF-stock run in Safe Mode with Networking) = 52.385s (http://img221.imageshack.us/img221/8935/1m3360400233652sap6.png)
3416.5MHz - 2x 512MB - DDR406 - 2-3-3-5 (600 Maxmem-4.8GB CW-384-384 PF-fully tweaked hardware/OS-475MB memory-649MB cache) = 50.352s (http://img512.imageshack.us/img512/4730/1m34165406233550sob9.png)
3416.5MHz - 2x 512MB - DDR406 - 2-3-3-5 (600 Maxmem-4.8GB CW-384-384 PF-fully tweaked hardware/OS-476MB memory-630MB cache) = 49.992s (http://img401.imageshack.us/img401/8532/1m341654062335499scc5.png)
3461MHz - 2x 512MB - DDR412 - 2-3-3-5 (600 Maxmem-4.8GB CW-384-384 PF-fully tweaked hardware/OS-478MB memory-607MB cache) = 49.531s (http://img440.imageshack.us/img440/4483/1m34614122335495ga4.png)
3528MHz - 2x 512MB - DDR420 - 2-3-3-5 (No Maxmem-384-384 PF-fully tweaked hardware/OS-820MB memory-105MB cache) = 49.411s (http://img212.imageshack.us/img212/5077/1m3528420233549stm0.png)
3495MHz - 2x 512MB - DDR416 - 2-3-3-5 (No Maxmem-4.8GB CW-PF?-fully tweaked hardware/OS-724MB memory-402MB cache) = 48.519s (http://img411.imageshack.us/img411/5408/1mnotweakrealtimece8.png)
3495MHz - 2x 512MB - DDR416 - 2-3-3-5 (No Maxmem-4.8GB CW-384-384 PF-fully tweaked hardware/OS-731MB memory-806MB cache) = 48.360s (http://img252.imageshack.us/img252/4696/1m34954162335after48gcwqh2.png)
3495MHz - 2x 512MB - DDR416 - 2-3-3-5 (No Maxmem-4.8GB CW-384-384 PF-fully tweaked hardware/OS-724MB memory-450MB cache) = 48.320s (http://img411.imageshack.us/img411/5408/1mnotweakrealtimece8.png)
3540MHz - 2x 512MB - DDR421 - 2-3-3-5 (600 Maxmem-4.8GB CW-384-384 PF-fully tweaked hardware/OS-484MB memory-689MB cache) = 48.099s (http://img214.imageshack.us/img214/5372/1m3540421233548sfe8.png)
3503MHz - 2x 512MB - DDR417 - 2-3-2-5 (No Maxmem-4.8GB CW-384-384 PF-fully tweaked hardware/OS-724MB memory-447MB cache) = 47.849s (http://img264.imageshack.us/img264/1364/1m3503417232547849sxi3.jpg)
3540MHz - 2x 512MB - DDR421 - 2-3-3-5 (600 Maxmem-4.8GB CW-384-384 PF-fully tweaked hardware/OS-513MB memory-638MB cache) = 47.699s (http://img261.imageshack.us/img261/254/1m354042123354769sqn9.png)

Before you look further... where's my favorite emoticon gone? :eek: http://img320.imageshack.us/img320/1429/k0158qw.gif

STOCK - Standard Windows - 1GB pagefile - 1GB RAM - +35 processes running in background - +25 services running - 1x CDT quick run

3360MHz - 2x 512MB - DDR400 - 2-3-3-6 (660MB memory-670MB cache) = 44.504s (http://img148.imageshack.us/img148/756/1m3360400233644scdtmd4.png)
3495MHz - 2x 512MB - DDR416 - 2-3-3-6 (645MB memory-667MB cache) = 42.671s (http://img261.imageshack.us/img261/1300/1m3495416233642scdthu4.png)

CDT gives the system an extra 500MHz. :confused:

Can I break 40s with OS tweaks? http://www.smiley-channel.de/grafiken/smiley/geschlecht/smiley-channel.de_geschlecht003.gif

Onepagebook

11-04-2007, 09:05 AM

T_M

11-04-2007, 09:35 AM

*puts on lab coat*

Here's some testing i spent all day doing:

1. Set up OS maxmem=600, pagefile=515~512 and registry large system cache etc. Set WIndows to run in Diagnostic Mode.

http://img.photobucket.com/albums/v310/T_M/CDT-IVcpuZ.jpg

http://img.photobucket.com/albums/v310/T_M/CDT-IVregistry.jpg

2. Set up CDT-IV folders and files. I created the files using 1024x1024x632=662700030. Didnt have enough room on the OS partition, so ran all test copying the files between the SuperPI partition and a data partition. Rar the 3 CDT files into a single uncompressed .rar of size 1.85Gb.

http://img.photobucket.com/albums/v310/T_M/CDT-IVcmd.jpg

http://img.photobucket.com/albums/v310/T_M/CDT-IVfile.jpg

http://img.photobucket.com/albums/v310/T_M/CDT-IVrar.jpg

3. Reboot, let idle a while and check taskmanager (picture below). Then perform the CDT copying of .rar file from SuperPI partition to other partition 3 times (rest inbetween each until available memory settles), each time renaming the .rar so i dont replace files. After CDT check task manager again and see 5xxxxx / 5xxxxx .

http://img.photobucket.com/albums/v310/T_M/CDT-IVtaskmanager.jpg

http://img.photobucket.com/albums/v310/T_M/CDT-IVcopy.jpg

4. Run 3 CDT tweaked SuperPI (reboot between each) and compare to 3 pre-tweaked runs i did earlier (no reboot between each)

Untweaked:
1. 14m 37.078s
2. 14m 36.188s
3. 14m 35.797s

CDT-IV:
1. 14m 27.985s (think this run was not done properly)
2. 14m 27.891s
3. 14m 27.922s

Compare then to Hipro5's 4Gb "CW" (i dont know how to do proper CW so just did a 4Gb.rar file from a partition to the SuperPI partition.

1. 14m 28.094s
2. 14m 28.109s
3. 14m 27.859s fastest run of the day

6. Graph all results and analyse loop times (time that each loop took). Find that both CDT and CW gave exactly the same characteristics.

http://img.photobucket.com/albums/v310/T_M/CDT-IVvsCW.jpg

OPB, can you please give me some pointers on anything in the above that would cause CDT to not give a better boost than just copying across a single large random file.
I even tried "balance method" where after doing CDT out of the SuperPI folder, i then returned all 3 files back into the SuperPI folder and my time was 1s slower, although both my Available and System Cache were higher (but balanced).

KTE

11-04-2007, 09:58 AM

KTE, thanks for your hard work, much appreciate
if you have time, try this,
use the lower full speed of your CPU is going to run
get the pre-cdt job done(while you reach the available/S.C. balanced)
either use setfsb or clockgen, pull all the way to the speed that you desire to run(ex: from 400x9 to 425x9) and then start the super pi; wait for 2nd or 3rd loop finish, directly press the x to close spi windows and start the super pi, run twice of 1m and start 32m, then check the 32m 's available...you will know what is gonna happened.:)Thanks Kev. I'll try it with the P4/P4C/C2D setup soon. A little far away from a computer at the moment (using laptop). :)

2. Set up CDT-IV folders and files. I created the files using 1024x1024x632=662700030. Didnt have enough room on the OS partition, so ran all test copying the files between the SuperPI partition and a data partition. Rar the 3 CDT files into a single uncompressed .rar of size 1.85Gb.
3. Reboot, let idle a while and check taskmanager (picture below). Then perform the CDT copying of .rar file from SuperPI partition to other partition 3 times (rest inbetween each until available memory settles), each time renaming the .rar so i dont replace files. After CDT check task manager again and see 5xxxxx / 5xxxxx .Good testing T_M.
-When you copy the same RAR 3 times to a partition (x3), you can overwrite it.
-Did you copy the RAR file from i) SPi partition->Another partition ii) SPi partition->another folder on that partition and from iii) Another partition in i) to SPi folder partition?

Your system cache was really low and unbalanced. Those won't be your best results.

T_M

11-04-2007, 10:05 AM

The screenshot of my task manager was before CDT. After CDT it was 5xxxxx / 5xxxxx always. I have photos of them all but cant be bother downloading them.
I was copying from SuperPI partition to another partition on same HDD.

All of the above was already written in my first post .....

Onepagebook

11-04-2007, 11:43 AM

kiwi

11-04-2007, 03:07 PM

your way to do CDT is completely WRONG;

OK, since you are the author of tweak maybe you will be willing to answer a few questions :)

1. Is CDT described here correct?
2. Is it full tweak or only 60% as someone mentioned here?
3. Should we see better results than CW if it is only 60%?
4. Should it be working on XP or is this specific to 2k3?
5. Any special registry tweaks without which CDT won't work as good?

T_M

11-04-2007, 06:40 PM

your way to do CDT is completely WRONG;

Thanks, but that doesnt really help me.
I have tried my best to understand the instructions from the broken english post in the first page.
Can you help me more constructively by saying what exactly is wrong with it?
I have put in a lot of time yesterday trying to get this working.

Maybe i should put some of my doubts down as quesitons in point form and you could indulge me with answers:

1. Does the Rar'ing of the files need to be done prior to the SPI run as part of the tweak process (rather than pre-preparing a Rar file).
2. Does the Rar need to be always copied from the SuperPI partition to the OS perition (or will any other parition do as was my case).
3. to be continued...

Onepagebook

11-04-2007, 06:49 PM

Thanks, but that doesnt really help me.
I have tried my best to understand the instructions from the broken english post in the first page.
Can you help me more constructively by saying what exactly is wrong with it?
I have put in a lot of time yesterday trying to get this working.

Maybe i should put some of my doubts down as quesitons in point form and you could indulge me with answers:

1. Does the Rar'ing of the files need to be done prior to the SPI run as part of the tweak process (rather than pre-preparing a Rar file).
2. Does the Rar need to be always copied from the SuperPI partition to the OS perition (or will any other parition do as was my case).
3. to be continued...

thanks but I thought what you wanna see is how pig can fly?;)

T_M

11-04-2007, 07:11 PM

Onepagebook

11-04-2007, 07:33 PM

Pigs flying comment was not to do with CDT process or whether it works or not.
It was with regards to your allowing someone independent into your benching room to monitor some SPI runs, since many people cried cheat.

oh so you can just forget about how it works and go back to what you think;)because everyone cried that and you can just follow;) why bother:)

T_M

11-04-2007, 07:48 PM

I think this a good case of "lost in translation".
Taking it to PMs.

Kain XS

11-04-2007, 08:51 PM

OK, since you are the author of tweak maybe you will be willing to answer a few questions :)

1. Is CDT described here correct?
2. Is it full tweak or only 60% as someone mentioned here?
3. Should we see better results than CW if it is only 60%?
4. Should it be working on XP or is this specific to 2k3?
5. Any special registry tweaks without which CDT won't work as good?

1.- There are now a lot of varians from CDT (as well as for Copy wazaa) that its difficult to really know if thats the whole point. Theory still works if you can apply it. :)

2.- I think Kev said this is 60% of the tweak, Actually, it is, but also its not.
In practice, what you see here (the tweak and how to do it) is 100% complete. Even Kevin added some extra comments on pre-job and how to get benching OS ready. In theory, thats were I can say its about 60% or probably less. Still, what its posted already, works, but notneccesarily its the whole theory and arguments behind it.

3.- Probably Yes, since its 100% practically for almost every guy, but you also can get better results with a lot of variations out there.

4.- Yes.

5.- Probably the ones mentioned on Pre benching and all the tweaks you know are neccesary for manage it.

KTE

11-04-2007, 11:30 PM

Chill out guys. You'll look low if you wage war over a virtual paper you can't even touch which runs an automated calc we do not control... :D Let's just standardize and perfect the method, then test and give appropriate feedbacks and try to collaboratively explain what we can.

I can say that SCSI U160 has slower SPi times than an ATA. I was going to test C2D with that too, but I've given up. Its because of the XP check FLAGS when writing to the SCSI that add latency. W2k doesn't suffer from this AFAIK.

hipro5

11-04-2007, 11:32 PM

I STIL see bulls coming and going around here........ :rolleyes:

IF a user asks questions about the "tweak" and "someone" responds with pigs flying and things like that, HOW is it possible to continue talking?.... How the user could understand what the other user wants to say?.... :rolleyes:

IF we aren't getting answers to our questions here, I see THREE things is gonna happen:

1. ONLY OBP has the BEST time in 32M at 3.6GHz in the whole PLANET :rolleyes: because of his "CDT Tweak"..... and because of what?..... :rolleyes:
What is this that gives him THAT MUCH boost to get ~10 - 15sec LESS than Copy Waza and other tweaks?...... and we are talking about a 20 - 25sec LESS time of a run without Copy Waza and things.... :rolleyes:

2. Users are gonna start dissapointing more and more of the:
a. not hepling them with more infos IF ANY and
b. having OPB talking bad to them so they will STILL OR JUST STARTING now not to like him nor his additude after all and NO MATTER how much they're trying to be friendly to him and "understand" his point of view, they will finally give it up.

3. QUIT this silly conversation and everyone goes back to his work and OPB to his reviews that he does quite well....... :)

Someone please come up with the WHOLE "Tweak" IF ANY, so we start testing seriusly and cut the craps..... :cool:

Onepagebook

11-04-2007, 11:50 PM

Bsically, I stand my point is correct, if there anyone is see me as pig or cheater, please don't bother to try it and make the whole CDT just look really bad, because you did making it wrong with whole lots of work, then it's still wrong.Then Why bother, and that's the point to keep in mind that if YOU are gonna just accuse me as bulls or cheater again, and I believe I will be with you and have the war contiuned, cuz I don't care you wanna make me a friend or not.

Master Hipro5 say it very correct, basically as long as I come here and post again, I don't even keep those childish in mind, but at least speak like him, I would highly appreciate that. Honestly we are discuss a theory to help people around the world, not argue who is better tweaker in OC.

KTE

11-04-2007, 11:51 PM

Not a reply to your post George but something in general I'd like to say that your post reminded me is the emoticon :rolleyes: I am definitely not fond of for the various meanings from one extreme to another it can show where users are friendly to one another (it causes misunderstandings) i.e. a) mock the thing->b) mock the person->c) condescend the person and it can exasperate things pretty damn fast without logical reason.

I can do some real-life XS testing on that if you'd like... :p:

T_M

11-05-2007, 12:46 AM

1.- There are now a lot of varians from CDT (as well as for Copy wazaa) that its difficult to really know if thats the whole point. Theory still works if you can apply it. :)

2.- I think Kev said this is 60% of the tweak, Actually, it is, but also its not.
In practice, what you see here (the tweak and how to do it) is 100% complete. Even Kevin added some extra comments on pre-job and how to get benching OS ready. In theory, thats were I can say its about 60% or probably less. Still, what its posted already, works, but notneccesarily its the whole theory and arguments behind it.

3.- Probably Yes, since its 100% practically for almost every guy, but you also can get better results with a lot of variations out there.

4.- Yes.

5.- Probably the ones mentioned on Pre benching and all the tweaks you know are neccesary for manage it.

Thanks kain for taking the time to answer some questions. :clap:
Would it be possible to take some of your time to go thru the method i used yesterday and you can correct me on the glaring mistakes i made to cause my CDT effort to fall way short?

Kain XS

11-05-2007, 06:51 PM

Thanks kain for taking the time to answer some questions. :clap:
Would it be possible to take some of your time to go thru the method i used yesterday and you can correct me on the glaring mistakes i made to cause my CDT effort to fall way short?

I´ve been very busy those days (still until wednesday :shakes: ) Even that, can you post step by step what youve done to manage Cache diversing?
I can help you in that way I think, but sorry if I get too slow because I have a lot of work.

youngpro

11-05-2007, 07:11 PM

i think thats what we need to focus on, copywaza is basically designed to push your available system cache as close to maxmem as possible, whereas cdt is about balancing your cache diversity, ive been playing with this alot, and im consistently getting 2-3 seconds faster with the methods opb described than my normal copywaza method!

T_M

11-05-2007, 07:17 PM

No worries Kain :D
Here's step by step what i did:

1. Reasonably clean Windows XP SP2 with all updates. Set up OS maxmem=600, pagefile=515~512 in SuperPI partition, and registry large system cache / clear pagefile on shutdown. Set Windows to run in Diagnostic Mode.

2. Set up CDT-IV folders and 3 x 632Mb files. I created the files using 1024x1024x632=662700030.

3. RAR the 3 CDT files into a single uncompressed .rar of size 1.85Gb.

4. Run OPB Cleaner

5. Reboot

6. Open SuperPI and prepare to run 32M, pause at 'Click OK to begin'.

7. Copy RAR file from SuperPI partition to other partition (note it was not to my C:\) - (1)

8. Rest

9. Copy RAR file from SuperPI partition to other partition - (2)

10. Rest

11. Copy RAR file from SuperPI partition to other partition - (3)

12. Rest

13. Check task manager and see 53xxxx / 54xxxx, or 55xxxx / 55xxxx, or 54xxxx / 53xxxx, or even 50xxxx / 54xxxx .

14. Run SuperPI

If you can please give me some pointers on anything in the above that would cause CDT to not give a better boost than just copying across a single large random file.
I even tried "balance method" where after doing CDT (3 copies) out of the SuperPI folder, i then returned all 3 files back into the SuperPI folder and my time was 1s slower, although both my Available and System Cache were higher (but balanced).

Kain XS

11-05-2007, 07:38 PM

Well, your metho is completely different from what has been described bro.
pagefile 512-512 in some way will work with your system available memory and cache. How many partitions are you using? Which is your OS partition? (or why did you moved to another than D: ?

Did you check then your HD was defragmented?

T_M

11-05-2007, 09:58 PM

completely different

I wouldnt say that, given that i have the same pagefile, set on the SuperPI partition, i have the same maxmem, the same file sizes, same RAR size, copy 3 times the same, find that my task manager gives generally the right numbers the same.

The trouble is i was not totally sure what has been described, as its been re-written about 15 times in different languages and by different people.

This post here: http://www.xtremesystems.org/forums/showpost.php?p=2522825&postcount=2 says to use 512~512 pagefile, supposedly an instruction from OPB himself.

The disk I was using has about 8 partitions of 6Gb (for OS's) and another of about 20Gb and another of 400+Gb.
The reason i couldnt copy the RAR between SuperPI partition and OS was because the 3 x RAR files are too big for my OS partition. If i could just overwrite the file each time then i could do it to the OS partition, but OPB's instructions were not to overwrite.

Partitions were only defragged using Windows XP defrag, and not necessarily prior to the run. Is that important too?

You mention "moved to D:", but the instructions say from from D:\ to C:\, is that what you meant?

KTE

11-06-2007, 12:28 AM

4. Run OPB Cleaner

5. Reboot

6. Open SuperPI and prepare to run 32M, pause at 'Click OK to begin'.

7. Copy RAR file from SuperPI partition to other partition (note it was not to my C:\) - (1)

8. Rest

9. Copy RAR file from SuperPI partition to other partition - (2)

10. Rest

11. Copy RAR file from SuperPI partition to other partition - (3)
I'll be honest with you, Elmor made it pretty crisp clear when he wrote his method and his subsequent replies. After that I find it hard to believe why people are still faulting to follow basic instructions he wrote down one by one and so have others many times. The only reason I could think is, is there another post somewhere I've not read where all these differences are coming from? Or some behind the scenes conversations where people are building on their misunderstandings with other than the guys who have ran it successfully? Because you're not the first and I doubt the last even, but your method was different to what was stated T_M. :yepp:

FWIW I started reading the CDT thread very late on and followed the last post by OPB/Elmor and was able to do it (albeit not perfectly). If I errored repeatedly, complained to make it clearer and that I don't understand it many times over with how George wrote down to do the CW step by step, the answer I'd get is " <rolleyes> dude go back and read it again, its all there" or just be ignored and mocked. People would get fed up of me. I'm being genuinely frank with you, you can't make it much more simpler, there is no hidden magic or conspiracy trash, what you need to do is all written there and by people adding, detracting and saying stuff about it which Kev did not say or affirm, all this confusion comes about. This is the full method: http://www.xtremesystems.org/forums/showpost.php?p=2509761&postcount=31

Kain is right you know. Your method is different to what OPB/Elmor stated and different to what I followed and understood by what they wrote too. Still v.good that you're trying genuinely. I'm not speaking on your qs because its best since you asked Kain, that he advises you instead so no more confusion comes about. Its bad enough as it is. :)

If you can please give me some pointers on anything in the above that would cause CDT to not give a better boost than just copying across a single large random file.
I even tried "balance method" where after doing CDT (3 copies) out of the SuperPI folder, i then returned all 3 files back into the SuperPI folder and my time was 1s slower, although both my Available and System Cache were higher (but balanced).
-Copying a single random file across does not give me any gains whatsoever (tried).
-As I listed before, you have to copy the RAR 3) to the SPi drive and 2) to a folder on the SPi drive too. This is intricate to the method.
-Why did you reboot?
-Why did you not overwrite the 1.85GB files each time?

CW does not make a difference in my case many times. Could someone explain that and give me pointers as to why not and how to get a boost everytime? Most likely not because I follow their method step by step. Such idiosyncrasies are there to stay IMO, just like you can pour LN2 in a tube and get 57xx clocks as your highest while another guy on a different day using the same hardware may pour it the same way and be able to achieve 58xx. We don't try and critique that or its underlying principles, we just know what we observed and experienced because its out of our control how a processor will behave. It would be awfully ignorant and jealousy of the first person to say "lol dude thats flat out fake even a pigeon can see that, you cheated, I ran the same damn thing and the processor does NOT go above 57xx, NO WAY!!" :D

T_M

11-06-2007, 12:42 AM

Elmor made it pretty crisp clear when he wrote his method and his subsequent replies. After that I find it hard to believe why people are still faulting to follow basic instructions he wrote down one by one and so have others many times.

Because i was trying to follow OPB's (the originator) method, in the second post of this thread.

The only reason I could think is, is there another post somewhere I've not read where all these differences are coming from?

Yes.

Kain is right you know. Your method is different to what OPB/Elmor stated and different to what I followed and understood by what they wrote too.

Im pretty sure i knew that when the tweak was not working for me, hence my repeated requests to ask someone who can get it to work where am i going wrong (and i dont want to hear another "its just wrong"!)

-Copying a single random file across does not give me any gains whatsoever (tried).
-As I listed before, you have to copy the RAR 3) to the SPi drive and 2) to a folder on the SPi drive too. This is intricate to the method.
-Why did you reboot?
-Why did you not overwrite the 1.85GB files each time?

- Definately gives me gains, in the order of 10s. Which would likely be due to the much increased system cache value (look at my task manager i posted after bootup which only has 1xxxxxx)
- I rebooted because nobody told me not to
- I did not overwrite the 1.85Gb file each time because OPB's instructions say not to (second post in this thread)

It would be awfully ignorant and jealousy of the first person to say "lol dude thats flat out fake even a pigeon can see that, you cheated, I ran the same damn thing and the processor does NOT go above 57xx, NO WAY!!" :D

Correct, and i never did.

Anyhow, looks like i'll need to go back to the ol' drawing board and see if i can better follow the post you just linked me to.
BTW, where did you get the info from OPB on 'balancing' by copying back to SuperPI partition?

KTE

11-06-2007, 12:58 AM

- I rebooted because nobody told me not to
That's bad...
You're adding stuff to instructions now. :D

- I did not overwrite the 1.85Gb file each time because OPB's instructions say not to (second post in this thread):p: Kev and Jon said don't overwrite the 632MB file in the RAR (rename them). But to overwrite the ~1.85GB file each time when copying over it (each RAR file is copied 3 times to the same place, and the copying is done from place->to place in 3 separate places as well - so you have 9 file moves altogether). Hope that's a bit clearer.

BTW, where did you get the info from OPB on 'balancing' by copying back to SuperPI partition?

Jon and Kev both mentioned it. You have to move back the RAR from the first place you copied it to back where you copied it from the first time around. ;)

T_M

11-06-2007, 01:29 AM

you cut the file , your ram is tempoarily store that img some address,
certainly you always cut from D to c..I did say from where you place the super pi folder right?

so after you paste on c...right click your mouse...can you do one more...two again?...
it will ask you "overwrite it..."

This is what confused me on that overwriting stuff.

So, in summary it would be:

1. Setup OS with maxmem, pagefile, etc, OPB Cleaner and then reboot
2. Open SuperPI and prepare to start 32M
3. Create 3 x 632Mb files in a folder (location important?), then RAR them into 1 .rar located in the SuperPI partition
4. Copy that .rar file to C:\ 3 times
5. Copy that .rar file to xxxxxxxxxxxx(?) 3 times from yyyyyyyyy(?)
6. Copy that .rar file back to SuperPI partition 3 times from C:\
7. Wait for taskmanager to balance
8. Run PI

Never heard of the copying it 9 times, so can you advise me on step 5 above?

T_M

11-06-2007, 01:41 AM

That's bad...
You're adding stuff to instructions now. :D

Another question, you linked me to elmor's method, which was to RAR the file on a seperate laptop, then transfer the RAR onto his computer for use.
That would indicate to me that the reboot i did should have no effect. Correct?

KTE

11-06-2007, 01:52 AM

It is 9 times total like I mentioned.

So, in summary it would be:

1. Setup OS with maxmem, pagefile, etc, OPB Cleaner and then reboot

2. Open SuperPI and prepare to start 32M
3. Create 3 x 632Mb files in a folder (location important?), then RAR them into 1 .rar located in the SuperPI partition
4. Copy that .rar file to C:\ 3 times
5. Copy that .rar file to xxxxxxxxxxxx(?) 3 times from yyyyyyyyy(?)
6. Copy that .rar file back to SuperPI partition 3 times from C:\
7. Wait for taskmanager to balance
8. Run PI
1. Why reboot though? I didn't reboot and nor did Kev so there was no reboot in the method he mentioned. What happens if you don't reboot, any change?

3. Doubt it. I always have one 3x632MB RAR made already in the SPi partition.

4. From SPi partition to another partition, yup.

5. Copy it from SPi partition to a folder on that partition, i.e. if SPi partition is D:\ then copy the RAR you have on D:\ to D:\CDT\

The rest is right as they mentioned and what I ran. :up:

T_M

11-06-2007, 06:57 PM

Great, thanks for a pointwise response :up:
You have mentioned "another partition", not C:\ so i presume then i can use any other partition and not the one that my OS is installed on?
Will do some tests tonight and report back with the differences to my previous runs :yepp:

T_M

11-07-2007, 08:40 PM

After all the good advice from KTE, i thought i might finally see some better results, so i tested again last night (albeit a bit faster) on QX6850 @ 3600MHz (8 x 450), DDR3 running 12xx 6-8-6-15-xxxxx:

1. Create 3 x 632Mb files then RAR them into 1 .rar located in the SuperPI partition. Setup OS with maxmem, pagefile, diagnostic mode, OPB Cleaner and then reboot
2. Open SuperPI and prepare to start 32M
3. Copy that .rar file from SuperPI partition to OS partition 3 times (overwriting, with pause in between to rest taskmanager)
4. Copy that .rar file from SuperPI partition to SuperPI:\CDT folder 3 times (overwriting, with pause in between to rest taskmanager)
5. Copy that .rar file from OS partition to SuperPI partition 3 times (overwriting, with pause in between to rest taskmanager)
6. Wait for taskmanager to balance at 56xxxxx / 56xxxxx
7. Run PI

Tested using 8M first:

No tweak: 2m 58.547s
CDT tweak: 2m 57.000s
Single 4Gb file copy: 2m 56.859s

Tested second with 32M:

No tweak: 14m 11.188s
CDT tweak: 14m 01.969s

So, my results are still no better than copying a single large file from one unrelated partition to another.

Am i going wrong by pausing in between each file copy?

dinos22

11-07-2007, 08:43 PM

Tim how long do you pause

make sure your available and system cache don't drop to standard idle figures otherwise you will probably see no benefit

probably a good idea to have Windows Task Manager Performance Tab open so you know where you're at

T_M

11-07-2007, 08:45 PM

In between each file copy i pause until the available and system cache both stop moving, and are generally 5xxxxx / 5xxxxx.
Each copy has task manager running to show me.

hipro5

11-08-2007, 12:50 AM

In between each file copy i pause until the available and system cache both stop moving, and are generally 5xxxxx / 5xxxxx.
Each copy has task manager running to show me.

Keep on trying.........IF you see any good resault out of it, let us know..... :D

mrlobber

11-08-2007, 12:55 AM

Just went back from a 3day business trip to UK, so I see a lot of things still going on about CDT here, good :D

Btw, has anyone noticed in OPB's screenies about the CDT tweak, he is actually doing the stuff a slightly different way:

(look carefully at that little explorer copying window, if copying is done from folder, the path has like drive letter + a chinese letter after it) ;)

1. copy from D:\folder to C:\ (elmor did D:\ to C:\)

http://img220.imageshack.us/my.php?image=51085556yh9.jpg

2. copy from D:\folder to D:\ (elmor did D:\ to D:\folder)

http://img220.imageshack.us/my.php?image=27862219yd7.jpg

3. copy from C:\folder to D:\ (OPB has even circled it, while writing the C:->D: stuff, interesting). Elmor did C:\->D:\

http://img220.imageshack.us/my.php?image=28587147rn4.jpg

... which kinda contradicts elmor's way of doing CDT... as well as KTE's? :cool:

Any thoughts about it?

hipro5

11-08-2007, 05:07 AM

Any thoughts about it?

Yes........For start.......Screenies don't work for me so I can't see anything.... :D :p:

mrlobber

11-08-2007, 05:15 AM

Yes........For start.......Screenies don't work for me so I can't see anything.... :D :p:

Ok, transformed them to hyperlinks :D Btw, they're from the original post of this thread, nothing new supposedly, just that little detail to turn the attention to.

before

11-08-2007, 05:42 AM

Ok, transformed them to hyperlinks :D Btw, they're from the original post of this thread, nothing new supposedly, just that little detail to turn the attention to.

Really? Humm not sure... details make the difference. Not sure they really make it (empirically), but we all have to take care of them. Rigourously.

TBH, folder or not folder, X partition or Y partition, etc... I don't think that could make a difference between these processes (CDT, CW, etc, different names for an equivalent aim); but if you wanna discriminate them, you have to take care of them.

So, thank you for showing us these inconsistencies.

I'm afraid, actually, we don't know what is precisly the whole CDT method. Additionnally, what's CW?

Stupid question?... not really ;) Each of us seems to have his own CW method making useless the comparison between CW and CDT...

We all need now an official CDT method and an official CW method. Continuing without this requisite will never draw an end. Simple as that.
:)

mrlobber

11-08-2007, 06:57 AM

Really? Humm not sure... details make the difference. Not sure they really make it (empirically), but we all have to take care of them. Rigourously.

Absolutely agree. This difference might be something, and might not - but it again proves the fact that nobody has confirmed we have been testing the "correct CDT", and it hasn't been clearly denied either except for a couple of "completely wrong" remarks without explaining what exactly has been done wrong :)

We all need now an official CDT method and an official CW method. :)

As to CW methods... although Hipro has posted his CW method with clear details how to do it, it's also true we most of the time are doing our own methods (and I have a personal CW which on my C2D 3600 Mhz setup is working 1second better in 32M than hipro's). However, as long as people can't get CDT working better than their copywaza, it doesn't matter which CW they're using because we are looking for the difference between the two methods to analyze in our "laboratory"... and if there isn't any, there is nothing to analyze :)

About the inconsistencies I have pointed out, I've been in process testing both the "written CDT" (original / elmor / KTE) as well as the "pictured CDT"... While testing, I have observed different things going around with both the amount of available memory and system cache as well as the ratio between them. I have done also various copy numbers of that 1.85Mb big file (from 1x to 5x) and found so far that the more you copy /replace the file around according to CDT rules the more available memory / system cache rise... and the more closer they get.

However, atm I've not finalized the testing yet and thus can't provide exact numbers... and regardless of numbers I haven't seen the revolution in my 32M times so far, so any hint from the CDT experts would be appreciated anyway :) (especially about that "written CDT" vs "pictured CDT" thing).

I'll try to post more details this evening.

Also, massman was posting previously in this thread he was able to get some 6secs better time with CDT, but also mentioned he has not finished his testing... I wonder, when would he post his final results? :)

http://www.xtremesystems.org/forums/showpost.php?p=2526281&postcount=48
http://www.xtremesystems.org/forums/showpost.php?p=2527077&postcount=60

massman

11-08-2007, 07:10 AM

Haven't had the time to finish my testings, REALLY busy atm. Sorry guys.

KTE

11-13-2007, 02:19 PM

I built a setup as promised, although I'm honestly not feeling too well these days and haven't had the slightest of extra time. I also have not messed with a new chipset since the 680i SLI, bear in mind please. ;)

E6750 L726 SLA9V 1.350VID / Zalman CNPS9500
Gigabyte P35 DS4 rev 1.1 (F4 and F8 BIOS)
2x 1GB Corsair XMS2 PC2-6400 CAS4 rev 2.1
Sapphire HD 2600 XT
WD Caviar SE 80GB SATA II 16MB cache
Seasonic S12 550W
Two extra fans
Idle @ 2.67GHz 1.2V DDR2-800 1.8V : 123W VAC
Load (orthos) @ same : 253W VAC
Idle in BIOS draw: http://i222.photobucket.com/albums/dd28/KTE__/E6750-726/13112007196.jpg

Board suffers from vdrops everywhere. :(

A quick few stock 1M runs, completely no tweaks at all, new Windows fully loaded, auto "high" volts/latencies, just frequencies are higher up to compare with later. All set in BIOS.
CPU MHz / FSB / Divider / DDR MHz / tCL-tRCD-tRP-tRAS: 1M Time

2667 / 333 / 5:6 / 800 / 4-4-4-12: 19.281s http://img161.imageshack.us/img161/8787/1m1strun266780044412rc9.png
3290 / 470 / 5:6 / 1128 / 5-7-7-25: 15.875s http://img50.imageshack.us/img50/3741/32894701128sj5.png
*3600 / 450 / 4:5 / 1125 / 5-5-5-18: 14.406s http://img223.imageshack.us/img223/9113/3600465uu8.png
3720 / 465 / 1:1 / 930 / 5-4-4-12: 14.188s http://img408.imageshack.us/img408/5274/37465gn4.png
3720 / 465 / 1:1 / 930 / 5-5-5-18 (:confused:): 14.157s http://img85.imageshack.us/img85/6244/372046555518mh0.png
3720 / 465 / 5:6 / 1116 / 5-7-7-26: 14.141s http://img159.imageshack.us/img159/6955/372055857728gr7.png

As you can see by the 3.6GHz run, RAM can do 1125 5-5-5-18 (can do 15 but not stable, 16 is stable) at 2.05V. I was about to try for some real runs before trying CDT with better clocks and so never really saved anything but out of the blue the system does not boot anymore. :( If the motherboard wasn't dead, at least the fans would spin up. Nope, the chip and everything else should be fine, it's the board that stopped booting completely. I had 5 hours off and on, 2 hours spent on Windows/programs, 1 hour spent on overclock, and the last 2 hours just to get it to boot, even froze components to subzero in a -28C freezer but no go. The DS4 is not booting up at all. I can't figure out what or why since the VDIMM/VCore were left untouched around 2.35V/1.472V at the last higher clocks and all temps were very low. :shakes:

Some random cell phone pics when setting up yesterday night taken with a Nokia N95-2 (the phone pics are not digital standard anyway but it really messed up far worse later because of JPEG conversion and decreasing resolution and size from many megabytes to a few kilobytes to upload):
http://i222.photobucket.com/albums/dd28/KTE__/E6750-726/13112007188.jpg
http://i222.photobucket.com/albums/dd28/KTE__/E6750-726/13112007179.jpg
http://i222.photobucket.com/albums/dd28/KTE__/E6750-726/13112007169.jpg
http://i222.photobucket.com/albums/dd28/KTE__/E6750-726/13112007170.jpg
http://i222.photobucket.com/albums/dd28/KTE__/E6750-726/13112007195.jpg
http://i222.photobucket.com/albums/dd28/KTE__/E6750-726/13112007194.jpg
http://i222.photobucket.com/albums/dd28/KTE__/E6750-726/13112007189.jpg
http://i222.photobucket.com/albums/dd28/KTE__/E6750-726/13112007191.jpg

Below is my blog for today for any that's interested -> to expand a little... :D

As promised, I did get the setup yesterday morning, but the problem was space in my flat. There is no room free anywhere at all. The E6750 SLA9V L726 1.350VID and from the choice of motherboards I chose a Gigabyte P35 DS4 rev1.1 F4/F8 BIOS for its known consistent high FSB and an Abit IP35-E Offlimits that was lying around unused I picked up too. Everything 'aint mine so I have to return working to my uncle who it belongs to. Two other systems I picked were older P4s and an AMD that Ill test soon. All standard air. One of the P4s died near the start and I don't know how or why since it was fully stock. I'm still thinking...
My god I hate long journeys for nothing. :mad: The CRT monitor I picked up was so bloody heavy walking up 6 flights of stairs to my flat I have a crazy back ache that made me want to throw everything back out again!!

Anyway, I ran into hundreds of problems, the first of which was the monitor and its weird colors. I don't like the DS4 BIOS at all. Its too simple and lacking in a lot. I have no way of finding out what the NB/FSB voltages actually are, only what is being added on to them, nor any PLL voltage control. The setup wasted so much of my time because for 3 hours it wouldn't boot since the board set the VDIMM at 1.8V stock which I couldn't tell until I got into windows one lucky time. 1.85V was good for ~464MHz 4-4-4-12 and the RAM I wanted was taken by someone else so I'll have to go back this weekend or so to get it (Crucial Ballistix Tracer 8500). I picked up some PC2-Corsair XMS2 6400 CAS4 that was lying around v2.1 which are ProMos ICs I believe. Not supposed to be any decent but mine ran 2.05V DDR2-1160 5-5-5-16 500% (1600MB) Memset clear without error. They booted in Windows at 1262 at 2.3V which is +0.55V in BIOS for us to set, :rolleyes: I've still not tweaked them at all, just raised the voltages pretty high on most components to see where the limit on each is, the temps and the ability. Left timings to "auto" mostly.

This DS4 cannot do 500FSB. It can ONLY do 475 tops from BIOS at max FSB/NB voltages or even if lowered down many notches. No voltage added would change that situation and nothing is even warm to the touch so I don't know what the hindrance is or if I just got a dud board. It 'aint the CPU because I tested with x6 multi and with x8 I got 475x8 into Windows at 1.40V BIOS volts (never tried lower). It's unstable because the board is unstable at such high FSBs.

So I can't do runs of 600FSB 1:1 or 500FSB or even 475FSB 32M like many guys are doing, nor can I do 3600/1200. Max I can do at 3600 is 1125 (all in BIOS staying below DDR2-1200) and latencies are high. tCAS I can't even change to below 5 (not an option) nor change Command Rate and that doesn't matter on the temps or volts. I even tried +0.66V which is 2.46V real VDIMM at 1262 which was running at 2.3V before but no go. Not tried raising clocks within Windows using Clockgen/SetFSB yet.

First thing I did is run stock TAT/Orthos/OCCT testing. TAT was hardest on my system, topping out at 253W VAC stock 2.67GHz volts was 1.2V, ambient was 10C, and the setup ran very cool with a Zalman CNPS9500, a 40mm 6000RPM fan on the NB, open bench and a 12 inch powerful fan cooling the DIMMS/heatpipes all the way from Asia :D EVEREST and Speedfan report the real temps and voltages accurately (I tested without a heatsink just to make sure) but Core Temp and TAT read 15C lower than the real temp at any time. Here's is TAT after 1 minute 100% load at stock. LOL its just so you get a rough idea of stock temps, I aint keeping it for use: http://img239.imageshack.us/img239/5472/tempsww8.png

Stock temps, volts and EVEREST mem/cache: http://img134.imageshack.us/img134/7919/evegpustockkk2.png

T_M

11-13-2007, 08:25 PM

Did you check your video card mount?
In the photos its not in the socket correctly, but i presume that you would have setup the system properly, just that it was like that in the photo.

KTE

11-14-2007, 04:14 PM

Ye, I packed it in yesterday and put in back in box thinking its dead and threw the board in the freezer while I went to sleep. Talk about condensation, but when I came back and thought "let's give it a try" as I did with one of my overclocked phones which stopped booting, it worked right out first time. ;)

Have been messing a little more but the OS got corrupt and won't pick up IDE DVDRW anymore from which I need drivers urgently. RAM is bad, as in I took it to 2.85V and 2.90V for 2 hours and it did exactly the same latencies/MHz as it did at 2.35V. Nothing more. :( I just wanted it to get up to 3600MHz/450/1350 or be able to drop some latencies lower clock as they were v.high.

My testing showed me this:
tRFC 25-50 made no difference in 1M
tRAS also did not make too much difference lower than 5.
Performance Level also did nothing below 5 apart from increase instability.
Don't recall the rest as I don't have access to the drive here. :p:

Will post some new testing scores soon. Need to find the limit before I can apply to test a tweak or I might just say the tweak worked whereas it hadn't and was just the hardware capability I hadn't raked.

KTE

11-14-2007, 09:33 PM

Just came back and started up again. Starting to tweak a little now from stock to see a 1M pattern on this board/CPU... no extra tweaks applied, just the timings you see. These ProMos chips absolutely gain nothing from voltage. 2.35V was tops that made a difference. I tried up to 2.90V, they were cool, but it gained zilch. CAS only changeable from BIOS and can't change Command Rate anywhere.

http://img117.imageshack.us/img117/2606/benchot7.th.jpg (http://img117.imageshack.us/my.php?image=benchot7.jpg)

Stock:
2667/400/4-4-4-12: http://valid.x86-secret.com/show_oc.php?id=268244 [screenshot + 1M = 19.047s (http://img340.imageshack.us/img340/5060/stockallrd0.png)]
(the CAS5 is because I couldn't be bothered to reboot and manually change it since BIOS sets higher than SPD)

Bit tweaked:
3200/400/3-2-2-5: http://valid.x86-secret.com/show_oc.php?id=268231 [screenshot (http://img340.imageshack.us/img340/9841/3225sq5.jpg)]
3200/400/3-2-2-2: http://valid.x86-secret.com/show_oc.php?id=268249 [screenshot + 1M (http://img292.imageshack.us/img292/5451/3222uj2.jpg)]
3200/400/3-2-2-1: http://valid.x86-secret.com/show_oc.php?id=268235 [screenshot + 1M (http://img340.imageshack.us/img340/4743/3221ad6.jpg)]

3200/500/4-4-4-9: http://valid.x86-secret.com/show_oc.php?id=268228
3600/540/4-4-4-8: http://valid.x86-secret.com/show_oc.php?id=268257

DDR MHz limit @ 2.35V:
650/5-5-5-15: http://valid.x86-secret.com/show_oc.php?id=268262 [screenshot (http://img101.imageshack.us/img101/9393/65055515vh1.jpg)]
656/5-5-5-15: http://valid.x86-secret.com/show_oc.php?id=268232
658/5-7-7-25: http://valid.x86-secret.com/show_oc.php?id=268233
660/5-7-7-25: http://valid.x86-secret.com/show_oc.php?id=268240
Not tried lower latencies. The latter should do 5-5-5-17 at same volts.

EVEREST showing VDIMM @ 649 5-5-5-18: Screenshot 2.35V (http://img208.imageshack.us/img208/2585/6495551814609evrbc3.jpg)

Various clock/latency 1M tests. Timings kept loose, VDIMM 2.1-2.35V 1M.

3483/653/5-5-5-18 1M = 14.609s (http://img529.imageshack.us/img529/6498/65355514609zu5.jpg):
5-5-5-15 made no difference.

3594.7/561.6/5-6-6-25 1M = 14.187s (http://img524.imageshack.us/img524/9373/14187qm5.jpg)
3600/450/4-5-5-13 1M = 14.172s (http://img222.imageshack.us/img222/7891/14172jh7.jpg)
3600/450/4-4-4-15 1M = 14.156s (http://img208.imageshack.us/img208/4432/14156hx1.jpg)
3594.7/561.6/5-5-5-15 1M = 14.141s (http://img211.imageshack.us/img211/2129/14141co7.jpg)
3600/450/4-4-4-12 1M = 14.125s (http://img520.imageshack.us/img520/9968/14125ja2.jpg)
3600/562.5/5-5-5-18 1M = 14.016s (http://img521.imageshack.us/img521/8150/14016yb8.png):

[3600 Idle TAT & Core Temp (http://img206.imageshack.us/img206/8273/3600450tempscc8.jpg)] [3600 14min loaded TAT and Core Temp] (http://img528.imageshack.us/img528/4816/3600tempsloaded30minuteeo7.jpg)

Talk about a strange bench. :D More coming soon...

dinos22

11-14-2007, 09:55 PM

what settings are you selecting in bios

make sure you use Turbo memory setting at least

KTE

11-15-2007, 04:50 AM

what settings are you selecting in bios

make sure you use Turbo memory setting at least
Chose 'Extreme' for all runs I posted and the 4-5 other values beneath that I left pretty much stock which was quite low. Can't access the drive here to check what they were but there were more runs. Was missing too many tools yesterday but I've got them today and will try some benches at night or tomorrow morn. :)

dinos22

11-15-2007, 04:51 AM

fine
also look at memset as well and tighten it up

your times are pretty slow :confused:

mrlobber

11-15-2007, 05:35 AM

your times are pretty slow :confused:

Well, a 515x7 1:1 4-4-4-4 run on Asus P5K-E at PL=7 does ~13.92 secs 1M and with a clean OS around 13m20s to 13m24s 32M:

Taskmanager before the run:
http://img221.imageshack.us/img221/8994/5stockoj0.jpg

32M:
http://img403.imageshack.us/img403/3050/632mstockub9.jpg

Edit: should have mentioned, those timings I used aren't the tightest either :)

Anyway, KTE, can't wait your 32M CDT C2D testing ;)

KTE

11-15-2007, 08:12 AM

fine
also look at memset as well and tighten it up

your times are pretty slow :confused:Yeah there is a lot to be tightened up as I left them loose, just as I mentioned. But even though the times weren't supposed to be quick when I tested dropping EACH timing one and two notches at the same clock, the difference was negligible. I'm talking 0.02s-0.001s. I didn't even have Maxmem on those though and a 2GB pagefile and no LargeSystemCache. Still its a good base to test a tweak for me.

Looking around if I compare my 3483/653/5-5-5-18 14.609s (http://img529.imageshack.us/img529/6498/65355514609zu5.jpg) rough run to a tweaked 3520/660/5-5-5-15 14.265s (http://www.pctunerup.com/up/results/_200710/20071027142429_Crucial6605-5-5.jpg) I don't see much efficiency difference. :confused: But I'm stuck with these timings until I go back to get some Crucial.

DJSUB

11-15-2007, 08:29 AM

here's my fastest run @ 2.4ghz 2x512mb @ ddr2 1200 cas5 :)
tweaks :
LSC=1
Pagefile=512/512
CopyWazza=744MB
Realtime Priority
Processor Affinity=1 core
Maxmem=600
ERAM=412000

http://img155.imageshack.us/img155/2923/cdtpostdn3.png

will try the CDT & Hipro5 tweaks as soon as i get a new drive :D

KTE

11-15-2007, 12:00 PM

Nope. I just tried lower timngs at 3600/562.5 5-5-5-5 Perf.Level 6, it booted and the 1M time was 14.078s but I've lost my OS again for the 2nd time and it deleted the latest programs installed and files saved. C:\WINDOWS\system32\config\system gets corrupt each time and I have to reinstall and setup everything back again which is a PITA!! :mad:

It's obvious that there is a hardware limitation here (RAM) so the only way I can compare properly is if people all run their best without CDT at the same fixed MHz/timings/sub-timings and make it quite doable. Or I have nothing to compare against simply put.
But before I can move on, I had a feeling timings are not doing anything more than 0.1s to 1M here so I tested the supposed gain I can expect with RAM timings on P35/C2D, low stable MHz...

LSC/Maxmem=600/PF=384=384/RealTime/SATA II 16MB

Stock: 2667/400/4-4-4-12 19.047s: http://img340.imageshack.us/img340/5060/stockallrd0.png
2667/400/5-7-7-25 19.094s: http://img217.imageshack.us/img217/6673/clipboard04pn7.png
2667/400/5-6-6-20 19.063s: http://img218.imageshack.us/img218/3929/clipboard03ls7.png
2667/400/5-5-5-18 19.047s: http://img105.imageshack.us/img105/2910/clipboard02ku2.png
2667/400/5-4-4-4 19.094s w/ CDT (look at cache/mem): http://img222.imageshack.us/img222/5523/clipboard01kw4.png
2667/400/5-2-2-2 19.015s: http://img218.imageshack.us/img218/8751/clipboard05as9.png

Is this what everyone is getting?

Someone please explain to me why there is hardly any gain (0.079s) from 5-7-7-25 to 5-2-2-2 in 1M :confused: especially considering people posting 0.1s lower time at the same clocks/latencies. Somethings wrong here. Also look at that CDT run, its the slowest, yet it has lowest tRFC and WTP.

massman

11-15-2007, 12:05 PM

Because 5-7-7-25 is way to slow and 5-2-2-2 ain't faster.

Try 5-3-3-11/13

mrlobber

11-15-2007, 12:44 PM

Hmm, what's the exact problem here? :confused:

KTE, it seems you can't run 3600mhz 514x7 1:1, can you?

If not, what about 450x8 1:1? Both your board as well as RAM should be up to this task, and it's not that much slower than higher fsb runs which we have plenty to compare with. Moreover, you need to do the stock/copywaza/cdt comparison inside your system first, and only then we could start doing a comparison with runs other people have made.

KTE

11-15-2007, 12:47 PM

Because 5-7-7-25 is way to slow and 5-2-2-2 ain't faster.I don't understand what you're saying massman. :confused:
They weren't runs to show fastest time possible but to see effect of tRCD-tRP-tRAS on 1M. Let me explain how clear this is:

I only changed major timings (tRCD-tRP-tRAS) that have a large supposed effect, no subtimings changed. That shows what effect each timing is having on 1M clearly. CAS had a bigger effect than all so I left it to 5 for all. Findings: Change in timings from 5-7-7-25 to 5-2-2-2 made not even 0.08s difference in 1M. Look at my results before at higher clocks, they are showing a very similar trend.
So take for example my 3600/562/5-5-5-18 14.016s run... if I ran same MHz/subtimings but 5-5-5-5 according to this trend on my hardware I'm spotting the difference would be ~ -0.03s to 14.013s, but not much more. Either something is wrong, it's board difference or I'm missing something.

Try 5-3-3-11/13
Already did it. It was slower than 5-2-2-2, which is fastest (all things kept constant).

KTE

11-15-2007, 12:49 PM

KTE, it seems you can't run 3600mhz 514x7 1:1, can you?Nope, 475 is max FSB. :(

If not, what about 450x8 1:1? Both your board as well as RAM should be up to this task, and it's not that much slower than higher fsb runs which we have plenty to compare with.Yep they can do that, but post the timings and sub-timings to try please. So I a can begin a comparison. :)

mrlobber

11-15-2007, 01:03 PM

They're not that important, the relative tweak comparison on your system is.

But you can try these, just for example (Memset 3.3): http://img403.imageshack.us/img403/3050/632mstockub9.jpg

KTE

11-15-2007, 01:18 PM

They're not that important, the relative tweak comparison on your system is.

But you can try these, just for example (Memset 3.3): http://img403.imageshack.us/img403/3050/632mstockub9.jpg

I "think" the RAM maybe doable at high volts but not FSB/divider, no way. Closest this setup can do to that is 450x8/900 4-4-4-4 or 540 5-5-5-5 which I don't think is even 4M stable. :shakes: Plus the hard drive is getting corrupt very quickly now, fitness already dropped to 92% and had over 3000 fragments. I'll try maybe tomorrow after reinstalling an OS. No OS yet.

massman

11-15-2007, 01:27 PM

I only changed major timings (tRCD-tRP-tRAS) that have a large supposed effect, no subtimings changed. That shows what effect each timing is having on 1M clearly. CAS had a bigger effect than all so I left it to 5 for all. Findings: Change in timings from 5-7-7-25 to 5-2-2-2 made not even 0.08s difference in 1M. Look at my results before at higher clocks, they are showing a very similar trend.
So take for example my 3600/562/5-5-5-18 14.016s run... if I ran same MHz/subtimings but 5-5-5-5 according to this trend on my hardware I'm spotting the difference would be ~ -0.03s to 14.013s, but not much more. Either something is wrong, it's board difference or I'm missing something.

Already did it. It was slower than 5-2-2-2, which is fastest (all things kept constant).

You're doing something wrong I think. All times tested on effects in 1M bench, keeping every other setting than the variable constant.

http://www.thetechrepository.com/attachment.php?attachmentid=617&d=1193158209
http://www.thetechrepository.com/attachment.php?attachmentid=618&d=1193158209
http://www.thetechrepository.com/attachment.php?attachmentid=619&d=1193158209

KTE

11-15-2007, 01:40 PM

Yeah, that's why I posted the initial results so someone can notice that the timing/sub-timing change is not giving me any difference in results. Negligble difference which is not what I expected and not what others I saw were getting, incl yourself. There's nothing wrong I'm doing other than what everyone does, it's a standard simple procedure, all things are constant, but I think it's hardware related...

Massman what CPU MHz/DDR MHz/latencies did you run those tests at? (so I can try them) I don't get those differences at all, no where near. :shakes:

KTE

11-15-2007, 05:04 PM

I knew the parts were bad but trying got me no where. The system does not oc anymore, any clock/latency change from BIOS results in no change. It defaults to this now no matter what I do: http://valid.x86-secret.com/show_oc.php?id=268661

SetFSB gives me this now:
http://img57.imageshack.us/img57/8018/setle4.png

Sorry fellas. :(

This is what I meant by no or hardly any 1M change with major timings:

3600/540/5-4-4-5 39-7-13-12: http://img214.imageshack.us/img214/9457/clipboard06cp9.png
3600/540/5-4-4-4 39-7-13-12: http://img211.imageshack.us/img211/1668/clipboard07xu1.png
3600/540/4-4-4-4 42-6-16-12: http://img210.imageshack.us/img210/700/clipboard08qu6.png / http://valid.x86-secret.com/show_oc.php?id=268660

All 3 times were 14.094s. Many others I saved are lost during system messup I just found out now when searching, even ran 4-4-4-3 and tRFC 35 and got the same time. Got a 14.000s with 450x8 1:1 4-4-4-4 funnily enough.

T_M

11-15-2007, 06:50 PM

WHen did this turn into a RAM timings thread?
As ML said, just choose some settings that work on your system, then compare no tweak/CDT/CW which is what we are all interested in.

Anyhow, ive had Gigabyte boards doing that 'locked FSB' before, just pull out the battery and clear CMOS and you'll be back up and running.

massman

11-15-2007, 10:04 PM

KTE

11-15-2007, 10:04 PM

WHen did this turn into a RAM timings thread?
When dinos said the times were quite slow and I showed I don't know how on earth guys say 4-4-4-4 PL6 gives you better times than 5-5-5-5 PF7 w/P35 because it doesn't on my setup, not even a millisecond.

As ML said, just choose some settings that work on your system, then compare no tweak/CDT/CW which is what we are all interested in.

Anyhow, ive had Gigabyte boards doing that 'locked FSB' before, just pull out the battery and clear CMOS and you'll be back up and running.
Tried it long ago mate. No go. I think you can carry on the testings yourselves now.

KTE

11-17-2007, 01:32 PM

Can someone post a 32M Pi score ran on XP Pro at 3550-3600/500-550 with or without LSC/Maxmem but definitely without any Copywaza or CDT please. Make it as good as you can without the above two tweaks if possible. It should be easy to replicate these high subtimings exactly.

http://img216.imageshack.us/img216/2833/settingsfk7.png

Just so that I know if its my software optimizations that are screwed or the tweak working.
BTW, ATi flacking drivers suck like crap! 35 too many BSODs has finally screwed my RAM over on the last BSOD:
http://img221.imageshack.us/img221/8259/clipboard015cr9.png
http://img211.imageshack.us/img211/2440/memtest2ms1.jpg

After the last of ATi BSOD, memeory now gives massive errors. 4 days and new memory is dead. Worst of all its not even mine. :shakes: :censored: :cussing:

STEvil

11-17-2007, 03:17 PM

driver BSOD wont kill your memory.

Dead memory will be due to too much voltage or poor filter/regulation hardware on the board itself causing spikes on power-on/load(transient).

KTE

11-17-2007, 10:20 PM

Power variations won't be the problem, the Seasonic is stable as hell 195W/550W max draw, DC primary and secondary filtering is excellent and neither will DIMM voltage since I ran 1.85V for 400/450, 2.1 for 500/550 and 2.3-2.35V for 600+ and the heatsink temp was 8C on average which is more than 15C cooler than stock in a case. I run Memtest every day at the end of use, and only on the day after repeated ATi CCC caused BSODs at 500/550 2.1V did it start giving memory/32M errors. Replacement will take a while now but funnily everything else but 32M Pi/Memtest works fine, even stress tests and intensive gaming.

hipro5

11-18-2007, 01:06 AM

Well......After ALL these pages AGAIN, do we have any news on this or not yet?...... :rolleyes:

KTE

11-18-2007, 03:11 AM

Well......After ALL these pages AGAIN, do we have any news on this or not yet?...... :rolleyes:
Read post #205, run that, but just for you, DO your best copywaza/tweaked time at those settings and post the 32M time. I'll reply to show if I have some CDT gains over copywaza or not. Otherwise there's a billion other who can be testing instead.

hipro5

11-18-2007, 03:38 AM

Read post #205, run that, but just for you, DO your best copywaza/tweaked time at those settings and post the 32M time. I'll reply to show if I have some CDT gains over copywaza or not. Otherwise there's a billion other who can be testing instead.

I'm NOT talking about YOU in personal if you took it that way...... :D
I'm talking in general about this thread..... ;) :)

massman

11-18-2007, 03:44 AM

Well......After ALL these pages AGAIN, do we have any news on this or not yet?...... :rolleyes:

Nope, not yet :rolleyes:

saaya

11-18-2007, 04:38 AM

hey guys, as i said i didnt have much time to follow this thread.
im really glad at how this thread stayed on topic, at least 90% of the time :)

i hope somebody can get the tweak working, atm it looks like only massmann, kte and t_m are trying to reproduce copy waza. so kte you got copy waza working but on a much older system? since your having problems with your new system, why dont you use skype or a webcam chat to guido somebody else through the cdt process?

Rather than keeping it hidden he spoke openly and tried to help everyone
i didnt see that yet but i hope we are getting there :)

mrlobber

11-18-2007, 06:16 AM

Well, I don't think we need much more unsuccessful reports as there have been already plenty of them :( ... but if ya want, here's another one which I've compiled after literally days spent on all this... still can't wait for detailed reports from those who have been at least partially successful with CDT and compare them with my experience :)

System:

CPU: E6600 B2 stepping
Mobo: Asus P5K-E unmodded
RAM: 2x1Gb Corsair 6400C3 (spd 3-4-3-9)
Video: Asus X850XT (no drivers installed :D)
HDD: WD 20Gb IDE for system, Samsung T166 500Gb for pagefile/Superpi
Cooling: all air
OS: WinXP SP2

Tweaks:

OPB registry stuff (LSC, Prefetcher, Throttledualcore etc)
Maxmem=600
Pagefile=512Mb fixed
no ERAM
Services: all disabled/manual except RPC, Plug&Play & Themes (for Luna Silver)
Priority=high from task manager
Copywaza=CDT file (3x632Mb rared, have this from CDT testing :)) 3x times copy on itself in Superpi folder (you'll see it later in the pictures)...

The system layout is 2 partitions: C:\ (system) and F:\, pagefile of 512Mb exactly on F:\, Superpi folder on F:\ and copywaza (CDT) folder on F:\ as well, nothing more there.

Most of the services disabled:

http://img401.imageshack.us/img401/8393/1servicesqh2.th.jpg (http://img401.imageshack.us/my.php?image=1servicesqh2.jpg)

Taskmanager at fresh boot:

http://img98.imageshack.us/img98/6544/2taskmgrfreshbootsj2.th.jpg (http://img98.imageshack.us/my.php?image=2taskmgrfreshbootsj2.jpg)

Taskmanager processes on fresh boot:

http://img232.imageshack.us/img232/4742/3taskmgrfreshbootprocesac8.th.jpg (http://img232.imageshack.us/my.php?image=3taskmgrfreshbootprocesac8.jpg)

How the CDT file does look like (exactly 3x632Mb):

http://img239.imageshack.us/img239/4749/4cdtfileyg7.th.jpg (http://img239.imageshack.us/my.php?image=4cdtfileyg7.jpg)

Before running stock 32M:

http://img221.imageshack.us/img221/8994/5stockoj0.th.jpg (http://img221.imageshack.us/my.php?image=5stockoj0.jpg)

Stock 32M : 13m20.047s (default boot subtimings :) ):

http://img403.imageshack.us/img403/3050/632mstockub9.th.jpg (http://img403.imageshack.us/my.php?image=632mstockub9.jpg)

Copywaza my way (copy of the big cdt file on itself in superpi folder, you can see the *.bat file used for that)

Taskmanager before the run:

http://img405.imageshack.us/img405/6453/7copywaza1na0.th.jpg (http://img405.imageshack.us/my.php?image=7copywaza1na0.jpg)

Result: 13m12.312s

http://img211.imageshack.us/img211/7482/932mcwvn2.th.jpg (http://img211.imageshack.us/my.php?image=932mcwvn2.jpg)

Now on to the CDT how it was described (like elmor & KTE @ XS were doing, again, I'm using a *.bat file for that, check it to see the copy sequence):
How the taskmanager looks after the tweak:

http://img231.imageshack.us/img231/7004/10cdt1wg5.th.jpg (http://img231.imageshack.us/my.php?image=10cdt1wg5.jpg)

The result, though, is slower than the copywaza I did :( 13m13.859s

http://img252.imageshack.us/img252/3553/1232mcdtoriginalgk7.th.jpg (http://img252.imageshack.us/my.php?image=1232mcdtoriginalgk7.jpg)

Now I tried a little different thing as well (from the pictures of applying the CDT tweak which were posted at the beginning of this thread), and managed to achieve very close balance :

http://img142.imageshack.us/img142/4257/13cdtexperimentwf4.th.jpg (http://img142.imageshack.us/my.php?image=13cdtexperimentwf4.jpg)

The result is half a second better than previously (which indicates that memory balance has to do something with 32M times, indeed ;) ), however, it still is a second slower than my copywaza run 13m13.484s :

http://img152.imageshack.us/img152/4753/1432mcdtpictured1ms2.th.jpg (http://img152.imageshack.us/my.php?image=1432mcdtpictured1ms2.jpg)

What can I add... running the original CDT 5times in a row raises the available memory and system cache by additional 10 Mb for me (can even achieve 569/571 balance in my case), however, for me this never translated to better 32M times.

KTE, I just switched my bench HDD to a faster one, hope I'll be able to give you a run with the timings you asked soon :)

massman

11-18-2007, 06:23 AM

I'll continue my testings when I've finished the P35 roundup review :)

KTE

11-18-2007, 06:37 AM

I'm NOT talking about YOU in personal if you took it that way...... :D
I'm talking in general about this thread..... ;) :)
I understood George. :D Just asking for some comparison times. CW does work on my C2D though. I tried it many times but hardware error and limits so can't do more yet.

so kte you got copy waza working but on a much older system?Can get it working on the E6750, pretty well.

since your having problems with your new system, why dont you use skype or a webcam chat to guido somebody else through the cdt process?Don't use them where I'm staying. I hardly ever use MSN bro, let alone those. You won't find me that free to, I usually post from work after finishing all my tasks for the day, time is a major issue. Try doing two years of a science degree in one year and working 7 days a week along with it and you'll know why :p: :shakes:

KTE, I just switched my bench HDD to a faster one, hope I'll be able to give you a run with the timings you asked soon :)Thank you mate. Look forward to it. :up:

Your CW/CDT difference looks the same as mine (mine was at lower clocks though) but that was a fudged run by me. If you don't have it done well through my experience you will get a 32M time 1-3 seconds slower than a 4GB copywazza. However, if you did it well, then it will be quicker than a very good CW by at least 2 seconds by my C2D experience.

T_M

11-18-2007, 08:00 PM

ML, maybe you require some extensive pauses between the file copy processes in your .bat?
Like maybe 1 minute to be sure.

What would the .bat code be to do a pause?

mrlobber

11-19-2007, 02:11 AM

ML, maybe you require some extensive pauses between the file copy processes in your .bat?
Like maybe 1 minute to be sure.

I don't think so... the CDT process states you need to wait those around 45 seconds only after all the CDT copying has been done, not between the steps. As far as I've gone with my testing, letting the system to be idle between the CDT steps turns the CDT into an ordinary not-so-good copywaza. If I'm not correct on this, feel free to turn my attention to it :)

Btw, yesterday it just became more mysterious... I switched to 2 identical HDD's (Seagate 160Gb SATA 8Mb, previously it was 20Gb IDE for system and 500Gb 2nd HDD), and guess what... the CDT 32M time improved by ~0.6 secs (first time I got below 13m13s with CDT and the configuration I described above) while my copywaza time - only by 0.2 seconds :shrug: This means additional testing again :cool:

KTE, I did a run with your timings, but haven't taken the screenshot with me. The stock time (with tweaked Windows, no copywaza, no CDT) was, if I remember properly, around 13m33s.

KTE

11-19-2007, 02:26 AM

Btw, yesterday it just became more mysterious... I switched to 2 identical HDD's (Seagate 160Gb SATA 8Mb, previously it was 20Gb IDE for system and 500Gb 2nd HDD), and guess what... the CDT 32M time improved by ~0.6 secs (first time I got below 13m13s with CDT and the configuration I described above) while my copywaza time - only by 0.2 seconds :shrug: This means additional testing again :cool:Having two HDDs, new, same ones made a gain to my results too! :)

KTE, I did a run with your timings, but haven't taken the screenshot with me. The stock time (with tweaked Windows, no copywaza, no CDT) was, if I remember properly, around 13m33s.Cheers. Try doing a CW run now please and let me know the time and frequencies used. :up:

mrlobber: forgot to throw in one major edit. Use 8 multi please because my setup can't do that FSB at 1:1. :)

kiwi

11-19-2007, 05:07 PM

KTE: I can post that time but please be more specific. I can run for example LSC=1, 9x400, PL=6 or 7 if you wish, RAM @ 500 and your timings

KTE

11-19-2007, 06:37 PM

KTE: I can post that time but please be more specific. I can run for example LSC=1, 9x400, PL=6 or 7 if you wish, RAM @ 500 and your timingsYep, these (http://img216.imageshack.us/img216/2833/settingsfk7.png) settings exactly. I would prefer if you could replicate 450x8 DDR:540 if you can and you can use Maxmem/LSC/CW and so on... but not CDT, so that I can differentiate. Thanks. :)

hipro5

11-19-2007, 11:19 PM

I don't think so... the CDT process states you need to wait those around 45 seconds only after all the CDT copying has been done, not between the steps. As far as I've gone with my testing, letting the system to be idle between the CDT steps turns the CDT into an ordinary not-so-good copywaza. If I'm not correct on this, feel free to turn my attention to it :)

Btw, yesterday it just became more mysterious... I switched to 2 identical HDD's (Seagate 160Gb SATA 8Mb, previously it was 20Gb IDE for system and 500Gb 2nd HDD), and guess what... the CDT 32M time improved by ~0.6 secs (first time I got below 13m13s with CDT and the configuration I described above) while my copywaza time - only by 0.2 seconds :shrug: This means additional testing again :cool:

KTE, I did a run with your timings, but haven't taken the screenshot with me. The stock time (with tweaked Windows, no copywaza, no CDT) was, if I remember properly, around 13m33s.

M8......ACCORDING to this "CDT Tweak" you HAVE to see MORE than 10sec dropping out of it.....;):p:

I mean that IF you have - for example - 13.13m with the Copy Waza, when you apply the "CDT Tweak", you MUST SEE 13.03m......THAT'S the time difference you SHOULD see according to OPB..... ;)
.......and you are talking about 0.6sec.....!........Give me a break...... :D :p:

mrlobber

11-19-2007, 11:28 PM

M8......ACCORDING to this "CDT Tweak" you HAVE to see MORE than 10sec dropping out of it.....;):p:

People can't get upstairs without taking the first steps, so let's consider these small differences being those steps ;) :D

hipro5

11-19-2007, 11:32 PM

People can't get upstairs without taking the first steps, so let's consider these small differences being those steps ;) :D

OK......At least give us ONE step at a time....for example ~ 3sec difference and then the rest 10sec difference.....
DON'T just move your feet a bit upwards coz you'll never reach the next step..... :D

T_M

11-19-2007, 11:51 PM

There's certainly more to this CDT tweaking than has been described in this thread. Thats for certain after having spent time on MSN with OPB.

filmbot

11-20-2007, 12:09 AM

OK......At least give us ONE step at a time....for example ~ 3sec difference and then the rest 10sec difference.....
DON'T just move your feet a bit upwards coz you'll never reach the next step..... :D

Now what happens if you or any other members here dont get a 10sec improvement?

hipro5

11-20-2007, 12:23 AM

Now what happens if you or any other members here dont get a 10sec improvement?

Nothing..... :) I personally don't care any more...... ;)

IF someone get's the "gain of 10sec", then he'll be happy I think..... :)

filmbot

11-20-2007, 12:30 AM

Then why keep harping? Or even why start in the first place?

If nobody saw the gain, what was that suppose to prove?

hipro5

11-20-2007, 12:32 AM

Then why keep harping? Or even why start in the first place?

If nobody saw the gain, what was that suppose to prove?

I have forgot what was the purpose of this thread..... :(

I think it's about that this "CDT Tweak" works and that OPB is not a cheater.....Yes.....I think that's all about......10sec lower than Copy Waza and OPB is not a cheater..... :)

kiwi

11-20-2007, 02:21 AM

Then why keep harping? Or even why start in the first place?

If nobody saw the gain, what was that suppose to prove?

I think he is saying that 0.xx seconds is within margin of error for 32M time.

blossa

11-20-2007, 03:09 AM

I think it's about that this "CDT Tweak" works and that OPB is not a cheater.....Yes.....I think that's all about......10sec lower than Copy Waza and OPB is not a cheater..... :)
Actually, I am sorry, I think you are a bit out of line here... It seems like some people HAVE gotten better result with this tweak. Maybe there is something to it, maybe a bug, I don't know. Let us find out. :)

By (atleast extremly closed to) calling OPB a cheater for presenting a tweak that has worked for him, you are VERY close to point out the few people that also has gotten better result with this tweak as cheaters, total n00bs or something like that.

If you think OPB is/was a cheater for other reason, then that is not what the thread is about. ;)

As a reminder (not only to hipro) here is a copy of the topic:

The CDT and copywaza lab

Regards / blossa - who is reading this thread to MAYBE find a good tweak.

KTE

11-20-2007, 03:55 AM

There's certainly more to this CDT tweaking than has been described in this thread. Thats for certain after having spent time on MSN with OPB.Then we, all of us alike, want you to write out what we are missing please, because that is as far as I know and others here know so far. :)

hipro5

11-20-2007, 04:07 AM

.....It seems like some people HAVE gotten better result with this tweak.

Hmmmmm.......Name ONE (except OPB) please with C2D.......ONLY ONE.... :)

massman

11-20-2007, 04:19 AM

I have forgot what was the purpose of this thread..... :(

I think it's about that this "CDT Tweak" works and that OPB is not a cheater.....Yes.....I think that's all about......10sec lower than Copy Waza and OPB is not a cheater..... :)

Then you're wrong ;).

Like I've asked in the previous thread, keep the OPB trash talk out of this, please :)

blossa

11-20-2007, 04:38 AM

Hmmmmm.......Name ONE (except OPB) please with C2D.......ONLY ONE.... :)
Me?

I think the 'problem' is that you think of this as 'OPB CDT-tweak'. If you looked at it more like 'maybe someone has come close to find a better CW than I use today' I think it would be easier. ;)

I also gained a few seconds trying to do this. If I actually did the CDT or a kind of CW, I do not know. ;) I know it is better than the cw I learned a year ago or so, that is good enought for me.

Last post from me in this thread. Back to topic.

KTE

11-20-2007, 04:52 AM

The only real thing left for me to establish is what stock/CW/other tweak times people here are getting at settings I posted. If I know that I can judge my own stock/CW/CDT results :
a) If they are purely the tweaks working
OR
b} If they are my own inefficient scores which could've been improved without the tweak being tested.

I.E. you can get low and high scores at the same settings by just pure efficiency.

Conclusion 1: So I might get 14m for a setting at stock and then run copywaza and get 13m50s and then run CDT and get 13m40s and then conclude, "hey I got a 10s faster time with just CDT over CW!!" -

Conclusion 2: While if I had compared the stock/CW run to others here first then I could've known that others were getting 13m39s just stock without even copywaza ... :(
So my results would be wrong and in that case would show me getting no gain by the tweak tested at all, just an improvement in stock run efficiency.

This is why I'm asking for similar test setup results from the C2D guys; moderate clocks and fair high timings (repeatable by all with a 8x multi). Because I can compare and know if indeed a time is decent or bad at chosen setting at stock, with CW, and with CDT.

I only have 2 days left now before returning the RAM and trying to get some more better RAM he had around. I just need a spare day to pick it up first. But I do have some results waiting since 3 days for someone to replicate the settings before I can comment.

BenchZowner

11-20-2007, 04:55 AM

I'm gonna bother sometime during the weekend ( reviews come first these days sorry ) and fiddle with it.
CW usually gives me about 16 seconds with C2D, will CDT give me more ? will CW + CDT work together and show gains over CW only ? Will CDT give me anything in a fully tweaked run ?

We'll see.

What bothers people the most is the "secret" SuperPi 1M C2D tweak...SuperPi 32M is tweakable and various stuff can help...even silly things that we never thought that they'd help.

hipro5

11-20-2007, 05:20 AM

Me?

I think the 'problem' is that you think of this as 'OPB CDT-tweak'. If you looked at it more like 'maybe someone has come close to find a better CW than I use today' I think it would be easier. ;)

I also gained a few seconds trying to do this. If I actually did the CDT or a kind of CW, I do not know. ;) I know it is better than the cw I learned a year ago or so, that is good enought for me.

Last post from me in this thread. Back to topic.

FIRSTLY: DON'T count me......count only other's oppinions/tests..... :)

Now.....You claim that you ACTUALLY see "the gain" out of it ONE YEAR AGO....Right?.....SO you KNOW the "whole CDT Tweak" and NOT "half of it".....
People here trying to find out/be told the "WHOLE CDT Tweak" (if any!) MANY DAYS (more than a month I think) and NONE is saying (knowing!).......
SO IF YOU claim that it has worked for you A YEAR AGO and you know the FULL "CDT Tweak", then WHY don't you say it to the other guys who are testing day by day spending their time/days doing nothing?..... :cool:

ALSO IF people here don't know the "FULL CDT tweak", why they are STILL TESTING something that it'll give them shi(f)t?....

In GENERAL now and to whom is conserned: SPREAD out the "FULL CDT Tweek" (if any) OR BUST..... :rolleyes:

I DON'T see any point for people trying and trying and trying and spending their precius time/hours/days, only to make runs of a "half CDT Tweak" and do nothing..... :confused:

Do something NOW OR close this thread coz it's useless...... ;)

.

kiwi

11-20-2007, 05:25 AM

Yep, these (http://img216.imageshack.us/img216/2833/settingsfk7.png) settings exactly. I would prefer if you could replicate 450x8 DDR:540 if you can and you can use Maxmem/LSC/CW and so on... but not CDT, so that I can differentiate. Thanks. :)

Here you go:

http://img513.imageshack.us/img513/6377/8x450ox6.jpg

KTE

11-20-2007, 05:32 AM

Thank you kiwi. :bows:

Can you mention any major tweaks used ther please. Will post my preliminary result as soon as I get a break from here.

kiwi

11-20-2007, 05:36 AM

* More or less freshly installed XP SP2
* Maxmem=620
* Largesystemcache=1
* Pagefile 512MB
* copy waza 4GB from SP partition to system and back
* SP high priority

Nothing else

hipro5

11-20-2007, 05:51 AM

Thank you kiwi. :bows:

Can you mention any major tweaks used ther please. Will post my preliminary result as soon as I get a break from here.

May I tell you something..... :)
You don't need to see other's benchies at the same frequencies as yours, etc......
You NEED to run a 32M by yourself (in general) with:

EVERYTHING with Windows tweaks which will be STANDARD for ALL of your runs (plus the tweaks needed with every method).....

a. A run WITHOUT Copy Waza
b. A run WITH Copy Waza
c. A run WITH "CDT Tweak"

THAT SIMPLE and we are talking and keep talking ALL THESE DAYS.....:)

massman

11-20-2007, 06:20 AM

FIRSTLY: DON'T count me......count only other's oppinions/tests..... :)

Now.....You claim that you ACTUALLY see "the gain" out of it ONE YEAR AGO....Right?.....SO you KNOW the "whole CDT Tweak" and NOT "half of it".....
People here trying to find out/be told the "WHOLE CDT Tweak" (if any!) MANY DAYS (more than a month I think) and NONE is saying (knowing!).......
SO IF YOU claim that it has worked for you A YEAR AGO and you know the FULL "CDT Tweak", then WHY don't you say it to the other guys who are testing day by day spending their time/days doing nothing?..... :cool:

ALSO IF people here don't know the "FULL CDT tweak", why they are STILL TESTING something that it'll give them shi(f)t?....

In GENERAL now and to whom is conserned: SPREAD out the "FULL CDT Tweek" (if any) OR BUST..... :rolleyes:

I DON'T see any point for people trying and trying and trying and spending their precius time/hours/days, only to make runs of a "half CDT Tweak" and do nothing..... :confused:

Do something NOW OR close this thread coz it's useless...... ;)

.

Useless to you !

C'mon, George, there's more in benching than just applying the tweaks everyone knows and bench ahead. Some people are willing to test every theory regardless if it's a true tweak or not. As far as I know, this tweaks was developped in five(!) years time. Why do you expect results in only a month? On the one hand, i'd like to know the tweak entirely, on the other hand, I just want to find it out together with the people who are contributing to this thread. Kevin may or may not post the entire tweak, purely as a bencher, I don't mind. Giving the hint is more important than explaning the full tweak, to me.

I know this thread is not your favourite one and I fully respect that. But please, please respect the enthousiasm of the users who just want to spend time on this :).

hipro5

11-20-2007, 06:28 AM

OK M8.......Do what you have to do..... :)

I won't post anything else.......I'll ONLY "watch" this thread over the upcoming years to see if anything new is going on....

Have a nice day/evening.... :D

KTE

11-20-2007, 12:41 PM

OK George, you have what you wanted. Gather around to hear my story... :D

I am sorry I couldn't replicate my timings I did 3 days back on the RAM anymore (ones kiwi did) -> it would error first loop after any tweak (its dying). However, I had to increase the timings and I could get it to run at the lowest possible timings for me at those clocks.

At kiwi's timings, with LSC, Maxmem=600, pagefile=512-512 but nothing else, I got a stock time of 13m 58s: http://img229.imageshack.us/img229/3116/360054055518jr1.png
(which kinda seems slow but RAM errors at those timings so not sure)

Pagefile makes ZERO difference in my case between 384 and 512, so I always choose 384MB-384MB for all the tests. I rebooted after any "tweak" test and started fresh on a new tweak run.

Stock Setup at bootup: http://img236.imageshack.us/img236/4751/setuphq8.png
New XP install OS setup for all the runs: http://img239.imageshack.us/img239/598/setupmaincj8.png

(450 x 8) 3600 / 540 / 5-6-6-15 (see screenshots for more)

#Stock->no tweak best I could manage: http://img124.imageshack.us/img124/6194/14m12sstockik0.png

14m 12.078s

#Copywaza 3.8GB file best: http://img229.imageshack.us/img229/4001/14m07scwqs0.png

14m 07.172s

#Copywaza 4.5GB file best (tried 5GB made no difference): http://img239.imageshack.us/img239/7695/14m06scwhd2.png

14m 06.890s

#CDT (with same pagefile though): http://img124.imageshack.us/img124/5132/14m01scdttk1.png

14m 01.843s

Memory/Cache was ~550/553 for Copywaza runs and ~535/550 for CDT runs.

Each of those times I cannot do unless I do those tweaks, hardware limit. I think its working for me, but I "think" it can do better if RAM wasn't bad. Will be getting Crucial Ballistix Tracer PC2-8500 soon to test. :)

Cheers.

kiwi

11-20-2007, 04:36 PM

How did you do CW?

Lets say system partition is C: and SP partition is D: (might be 2 different hard drives)

With SP window ready to start 32M, copy from D->C then from C -> D to another location without overwriting (overwriting as well as vice versa copy is another option)

Then watch task manager till available memory stops rising rapidly and is close to system cache. Then start SP

T_M

11-20-2007, 06:22 PM

Then we, all of us alike, want you to write out what we are missing please, because that is as far as I know and others here know so far. :)

I'd love to, however most of what i was told was "its only about 30%" and then the rest was just hints at pieces of information which i didnt really understand.

Some hints i think were:
Totally clean install of XP using nlite with all hotfixes and OPB Cleaner.
Create a file that once it is RAR'd is exactly 632Mb.
Must have balance inside and outside of the CDT folder on D:\.

KTE

11-20-2007, 09:45 PM

Since I'm not a SPi competitor, unlike others I don't care to "hide" any detail or "secrets" which they reckon makes them get a better time than someone else.

How did you do CW?

Lets say system partition is C: and SP partition is D: (might be 2 different hard drives)

With SP window ready to start 32M, copy from D->C then from C -> D to another location without overwriting (overwriting as well as vice versa copy is another option)

Then watch task manager till available memory stops rising rapidly and is close to system cache. Then start SP
I did this exactly, which is what I usually do for CW: http://www.xtremesystems.org/forums/showpost.php?p=2519794&postcount=315

On mine Drive 1 = F:\ and Drive 2 = C:\ (SPi folder)

Just that I only have one Explorer windows open as it occupies less memory. My Windows used MB was reading 56MB after CW, just before starting. No matter what size CW or tweak I do, I can't get any different times to that, if I switch all the services off but 2-3, the time is 2-3 seconds slower than leaving Windows theme running, and the fastest theme for me is the one you were seeing.

I did some runs 3 days earlier which saw good gains for both CW and CDT at higher clocks and lower timings but non of the runs completed, they failed on loop 6-7 because of memory problems. They will complete if memory is OK. One run was particularly decent.

I'd love to, however most of what i was told was "its only about 30%" and then the rest was just hints at pieces of information which i didnt really understand.

Some hints i think were:
Totally clean install of XP using nlite with all hotfixes and OPB Cleaner.
Create a file that once it is RAR'd is exactly 632Mb.
Must have balance inside and outside of the CDT folder on D:\.
OK. I'm setting off now to drop faulty RAM off, so may not have access to system for a while. I have 2 full AMD builds here untested (which I'll test on soon) but no RAM for them unfortunately and RAM 'aint cheap here, double US price.

mrlobber

11-21-2007, 03:11 AM

KTE

11-21-2007, 12:24 PM

KTE, you did that again :D That 5sec difference is all I'm looking for. Will try to replicate your runs as soon as I can... although Asus boards mostly are faster clock-for-clock than Gigabyte (and I wonder, could that make a difference somewhere in CDT as well?)I had a chance of getting P5K Black Pearl twice but I turned it down in favor of higher FSB on this board (usually). But it seems my CPU is walled at 480FSB so P5K would've been better for "absolute" Pi times. I wish I would've know that earlier.

Just to 100% clarify the things how we are doing those 30% CDT :p:, could you write the CDT copy sequence you are doing step-by-step, including the paths involved, whether any pauses between the copying of the files and literally everything you're doing while applying the information we've got about CDT

*I chose the hardware/OS settings I have already talked about at bootup in BIOS and in the previous Windows session.
*Boot up.
*Opened Task Manager to check memory/cache.

-My SPi folder is Drive C: which is the 1st partition on one SATA drive.
-My second folder is C:\CDT which is on the 1st partition of the same SATA drive.
-My third place of movement is Drive F: which is 1st partition on another SATA drive.

-Two identical SATA II drives.

**(On them runs) I started SPi, in Task Manager set the process priority of superpi to "Real-Time".
**I then ran 16k three times, 128k once, 1M twice, 32M upto 6th loop once and then choose 32M until the final window prompt to start it appears. I do nothing further with it yet.

Additional Info: I don't use the mouse (I/O). In case you don't know, I/O takes up CPU consumption, that's why they say "don't move the mouse during a bench". It will lower your scores. Moving the mouse fast does this worst. If you've captured screens/saved CPUZ validation files at very unstable clocks, you will know this very well. :yepp: You move the mouse as slow as possible (don't take years please, it won't help :p: ) and smoothly to where you need and leave it there.
I use Alt and scroll keyboard keys and hit the Enter key very fast 3 times successively for my runs. It DOES make a difference many times.... with me anyway. :D

What I did for the above CDT runs?

Step 1:

1: I opened Windows Explorer (folder side view) and right-clicked a 632MBx3 (1.85GB) RAR file which I have already made and left in location Drive C:\ called CDT.RAR.
2: I choose "Copy" and then right-click location Drive F:\ in the left hand-side folder view and choose "Paste".
3: I wait until it finishes copying. As soon as it finishes (no delay) I repeat the above two steps twice more for a total of 3 times file moving, rewriting over the previous file each time.

Step 2:

1: I then right-click the CDT.RAR file in Drive C:\ again and choose "Copy".
2: Then I right-click folder C:\CDT and choose "Paste".
3: I wait until it finishes copying. As soon as it finishes (no delay) I repeat the above two steps twice more for a total of 3 times file moving to this place, rewriting over the previous file each time.

Step 3:

1: I click on Drive F: and right-click the same folder CDT.RAR in it, and chose "Copy".
2: I right-click Drive C: and choose "Paste".
3: I wait until it finishes copying. As soon as it finishes (no delay) I repeat the above two steps twice more for a total of 3 times file moving to this place, rewriting over the previous file each time.
4: on the last copying over, I initiate the copying and close Windows Explorer.

Final Step:

1: I look at the Task Manager to see when the memory value reaches its max before starting to drop -> most of the time this takes 10-15 seconds delay with my experience.
2: As soon as I see it losing memory, I close Task Manager and wait 5 seconds more.
3: I then hit Enter and let 32M start.

Then... if it turns out a slow run, you sit back and cry "Why me!" ... LOL! :D

Are you sure you didn't swap these numbers? I mean, 550/553 for CDT and 535/550 for CW...Nope. CW gave me bigger memory/cache but while that DOES improve times very much from stock, it isn't the same thing as CDT -> based on my experience.

CW is the best SPi method of improvement that I knew before this=> there is NO way I can get CW times out of my hardware/setup without running CW=> practically impossible as all I do is hit keys. This is the same for many people that I know including most of you dear geeked out readers. :p:

From what I've experimented with over 4-5 methods of improving memory/cache and getting it to from 580-490/620-500, improving memory/system cache is not all that is happening here with CDT. That's what CW does best.
CDT is improving the memory/cache like CW but also there is something do with the method => something happens to your hardware/OS prefetch/cache by running this method which you do not get by running CW or anything other that I've tried and it is this that improves the time over a CW.

Also a few more personal comments into understanding this based on the above notes:

I don't know anyone who understand computers completely, not even Intel or AMD. Ask the lab engineers when they design a core if its always like they expected-> never.

You can run ANY tweak and get better times at the same setting for everything compared to your own PB. This is because on a different day and time the OS can behave and prioritize slightly more efficiently, and this "sightly" can easily make a second of a difference over a 13 minute bench. Run SPi on Linux and you'll see how its quicker, meaning the way the OS behaves matters crucially.

Some hardware is quicker than others at the same settings. This is beyond our user control.

You can run a CDT perfectly according to the procedure Elmor/OPB/I or anyone does but still not see any gain over CW. I've had this happen to me... but out of the blue, by that I mean beyond your control or understanding you may run it once more time and it will give you a faster time than whats possible otherwise with CW. All you do is, sit back and smile and know that it was a more efficient run, like you would say about a better CW. :)

My last runs were quite unfair. I did CW 4 times, and the two best times you see there. But I did CDT only once. Mainly because I didn't want to install GFX and so on again if it crashed and it takes so long. FWIW doing CDT is very hard on RAM, I noticed when I had looped 32M perfectly at higher clocks/lower timings with CW, but when I tried doing it after CDT, it failed repeatedly on the first loop on those settings, and even at stock clocks after CDT: (link place reserved because I can't access the drive yet)

Not even Orthos or OCCT caught those errors for over 2 hours, only Memtest, P95 and CDT-32M did. So now its become my way of memory stability testing too. :)

CDT leaves an "aftermath" effect of your system. Your times improve not just right after the method but maybe for all of the day while you have the system running. I've seen this with all 3 of my tests, on Celeron D, Pentium 4 and Core 2.

I kind of wish I had better hardware now, then I could really test CDT by running the same clocks/latencies as others who fail to understand or believe, but I did this on purpose. Getting better hardware was quite easy for me, although time and interest is an issue. People asked for "C2D" and so I tested one, but my testing hasn't ended yet, its preliminary. In one major way it is good not to have done it because that would end up with users here behaving the same way as what they do with OPB, which would make you see the "not-so-polite" side of me where I'm as sour as any of you can be and I don't want to be banned for no fault of mine. Wherever you have "lobbies" that always happens, and in this case it would be those who have ran SuperPi day and night more than the seconds I've used a computer for lobbying saying "they know better". I had a Duron earlier this year that was giving wrong readings and high benches of 4GHz+ and the daft of the onliner visitors reacted the same petty way, regardless of me saying "theres definitely something wrong". This is where this emoticon is best used => :rolleyes:

BTW, I'll decrease that CDT time further yet, how much I'm not sure, maybe even only a ms but I can drop it a little more at those settings I feel.