狂少's CDT-tweak

[massman]

Awesome

[mrlobber]

Guys, this tweak definitely works! I tested it myself and shaved of 19 seconds vs regular run and 15 seconds vs copy waza. I didn't even bother copying 3 times and using maxmem (was just testing around )

Here are my times:

Regular:
13m 21.781

Copy-waza:
13m 17.453

CDT as discribed by elmor but without maxmem and I copied file 3 times (D>C) + 1 time (D>D) + 1 time (C>D)

13m 02.625

Might do even better if I did maxmem + 3 times copy

Kiwi, you gotta come over to me with your HDD and show what you did lol

Again, this seems to be dependent of some Windows property, for instance, a registry entry some people have by default while other's don't, and we should dedicate ourselves finding out where this "key" is to be picked up. The CDT instructions are actually straightforward, at least the way elmor described at the beginning of this thread... but somehow they work for half of guys and the other half is left wondering with their "curve_hands.dll"

[[XC] riptide]

Right AFTER 278 Posts what do the Mods/Admins/Legends and Experienced, Known Oveclockers think? We could cut this thread to a summary before 300 posts it would be great.

[Gautam]

Guys, this tweak definitely works! I tested it myself and shaved of 19 seconds vs regular run and 15 seconds vs copy waza. I didn't even bother copying 3 times and using maxmem (was just testing around )

Here are my times:

Regular:
13m 21.781

Copy-waza:
13m 17.453

CDT as discribed by elmor but without maxmem and I copied file 3 times (D>C) + 1 time (D>D) + 1 time (C>D)

13m 02.625

Might do even better if I did maxmem + 3 times copy


See attached images

What CPU, FSB and memory speeds, and how big is your copy-waza?

[Onepagebook]

Kiwi, you gotta come over to me with your HDD and show what you did lol

Again, this seems to be dependent of some Windows property, for instance, a registry entry some people have by default while other's don't, and we should dedicate ourselves finding out where this "key" is to be picked up. The CDT instructions are actually straightforward, at least the way elmor described at the beginning of this thread... but somehow they work for half of guys and the other half is left wondering with their "curve_hands.dll"

ya, for example, don't set any size of L2 cache when you run pi instead of trying some hex number like 1000, 2000, 3000(yorks)

[mrlobber]

Retested on my 24/7 system (E6400 / DFI Infinity 965 Dark/2x1 GB Teamgroup PC5300 / 8800GTX) on dirty OS with absolutely no tweaks except realtime priority and affinity on 2nd core. No difference between copy waza (600 Mb C:\ -> D:\ and D:\ -> C:\) and CDT

[Eldonko]

Kick ass kiwi. Glad to see this thread back on track.

[hipro5]

Hipro, what memory voltage is that? 2.28v?

No, it's 2.35Vdimm...

[blossa]

Really nice kiwi!! Now I am really have to pull my thumb out of my ass and do some serious testing.

[DEVIL K-ce]

Trick not work for me

My previous time only with waza and some popular tweaks - 4:40 in PI 16M --> ( polish national sport )

With this tweak I have the same time. Settings of hardware the same for one and next run.


Martin

[hipro5]

So maybe kiwi didn't do "well" the Copy Waza!.....

Kiwi, could you try a Copy Waza of ~4GB rar file with Page file 384-384......

[blossa]

No, it's 2.35Vdimm...

Sorry for the OT, but isn't that a bit high...? Here I have ~2.32v:

I have later found out that it is possible with even lower voltage.

_

[hipro5]

Sorry for the OT, but isn't that a bit high...? Here I have ~2.32v:

I have later found out that it is possible with even lower voltage.

_

X38 and Cas 6-5-3-5-13-5-43-Perf.Level=5-15-16-2-6-2-2 AND 1T and Rams at 900MHz(1800MHz DDR) that I work now, is difficult enough plus my rams need that high voltage....

Could you post a WinRAR bench at 3600MHz as your above picture?...

[blossa]

X38 and Cas 6-5-3-5-13-5-43-Perf.Level=5-15-16-2-6-2-2 AND 1T and Rams at 900MHz(1800MHz DDR) that I work now, is difficult enough plus my rams need that high voltage....

Could you post a WinRAR bench at 3600MHz as your above picture?...

I can not do that atm, I will try that later. The reason is that my system seems to like other dividers better( ?? ). Also, strangly enough it seems like 7-5-etc performs very much the same as 6-5-etc with lower voltage. I have to test more to confirm that though. I am a poor bastard so I do not want to risk my RAMs to much yet with high voltage.

To not go to much OT, a winrar-bench thread might be good?

Btw, to run PerfLev=5 I think I have to use 200-strap. I am not sure but I think that screenshot was with 266-strap.

Last time OT for me in this thread. *sorry*

[mrlobber]

So maybe kiwi didn't do "well" the Copy Waza!.....

Copy-waza can net you around 10 secs if done properly, right? (at least that's always been the case for me). Kiwi has a gain of 20 secs - seems too big for copy-waza only...

[andre X_X]

CW can give me arround 20s vs regular, so i think kiwi dont have a good wazza...

[hipro5]

Copy-waza can net you around 10 secs if done properly, right? (at least that's always been the case for me). Kiwi has a gain of 20 secs - seems too big for copy-waza only...

He gained ONLY 4sec with Copy Waza and 19sec with this tweak.....So not good Copy Waza or at least the method of it that he performed....

I can not do that atm, I will try that later. The reason is that my system seems to like other dividers better( ?? ). Also, strangly enough it seems like 7-5-etc performs very much the same as 6-5-etc with lower voltage. I have to test more to confirm that though. I am a poor bastard so I do not want to risk my RAMs to much yet with high voltage.

To not go to much OT, a winrar-bench thread might be good?

Btw, to run PerfLev=5 I think I have to use 200-strap. I am not sure but I think that screenshot was with 266-strap.

Last time OT for me in this thread. *sorry*

No, Performance Level = 5 can be get at 266 strap with Cas 6.......If you put Cas 7, you can't get it....You can get only Perf. Level = 6 out of Cas 7....
....and Cas 7 with the exact same ram timmings as Cas 6, have a difference of about 100MB/s in WinRAR coz in one hand you have Cas 6 AND Perf. Level = 5 and in the other hand you have Cas 7 and Perf. Level = 6......

[dinos22]

That 's all for now, I appreciate Fugger call me lastnight and being very nice to me.

you mean this conversation

Hipro5 what speed do you install the operating system?

makes no difference......this was tested by a few reliable guys and make NO difference at all

Guys, this tweak definitely works! I tested it myself and shaved of 19 seconds vs regular run and 15 seconds vs copy waza. I didn't even bother copying 3 times and using maxmem (was just testing around )

Here are my times:

Regular:
13m 21.781

Copy-waza:
13m 17.453

CDT as discribed by elmor but without maxmem and I copied file 3 times (D>C) + 1 time (D>D) + 1 time (C>D)

13m 02.625

Might do even better if I did maxmem + 3 times copy


See attached images

Please provide full info before saying it works (as per OPB screen...that should be sufficient info). it's hard to judge if we don't know where you're at

i will be testing Kevin's tweak soon as i was shocked to see that time as well on OCX and would certainly like to improve my times by 20 seconds or so

...... so far i have to say that not one person has been able to actually do Kev's tweak properly yet in this thread >> why???? i don't think i would actually consider anything other than 12m40s few seconds up or down as "PROPER" runs........anyone that shows us 10,15,20,30 seconds gains does not make any difference if your times are still in 13m plus category @3.6GHz with same or better RAM MHz and timings if you understand what i am saying. Any guys here which are going to seriously bench these tweaks should have equal or better configs otherwise what is the use of comparing....ideally the settings supplied should be the same to make apple>apple comparisons >>> Elmor sorry mate but you really need to run at equivalent or same settings as OPB if we are to compare your results with his

i've done sub 13m with 617MHz 4-4-4-x on DDR2 RAM so a tweak like this with these settings easy 12m40s-something if i get the tweak to work properly.

Kevin, is there anything else you may have accidentally left out of description such as some other OS setups or ANYTHING else.

would it also be possible for you to run 514MHz*7 1:1 4-4-4-x runs with plain and CDT tweak on a fresh install of standard OS with tweaks as explained and full disclosure of all timings and what was done. i think buttertoast asked if it was possible to make a video of this....that would really make it crystal clear i guess. >>>> you get the idea >>> completely same settings for everyone and then just run the tweak >> this would actually be good if Kevin could show up for CES and have fugger or someone prepare two identical systems and then do the runs........that way we will put this one to bed forever and stop this age old debate......

that 32M time is amazing but i think what is even more amazing is the 1M time and this is the one i really want to nail. I think it was really the 1M time that really shocked me....sub 13.50s at 3.6Ghz is just on DDR2 CAS5..........i have never been able to do anything significant with wazza and 1M so this one is looking especially tough so it is really interesting that it works....fingers crossed .

Awesome, i look forward to seeing pigs fly

T_M what sort of a brain fart was that man >>> why even bother posting
why not try the tweak and tell us how you actually did. That would probably be more productive

[dinos22]

CW can give me arround 20s vs regular, so i think kiwi dont have a good wazza...

i get around 8-9 if done properly but lately it's been worse as well lol

[Zeus]

ya, for example, don't set any size of L2 cache when you run pi instead of trying some hex number like 1000, 2000, 3000(yorks)

You mean in registry for "SecondLevelDataCache"?

[andre X_X]

lol, maybe i have the most dirty windows .... but 20s is real

[Misko]

dinos22 is right, none of guys having spi time above 13min@3.6ghz on ddr2 are not comparative and dinos22 ram timing do not have some influence on 1M, like hipro5 said you can set 5-5-5-5-15 with only one stick, and 4-4-2-4-4 with BOTH (dual channel) and you can't still make not even close difference like OPB has, so it is not having anything with ram timings or eprom programming, just win tweaking.

[hipro5]

You mean in registry for "SecondLevelDataCache"?

Yes.....

[KTE]

Copy wazza, Maxmem, Defrag RAM nor CDT have never made a difference with my testing... nor increasing pagefile, nor keeping it at 'X' size (for SPi).

Windows OS tweaks and RAM speed/timings have.

Because all those above methods work on behind the mark is decreasing system cache and maximizing more RAM by allocating more RAM marked off by other threads so freeing them for a new software to use - and I already have 820MB/1GB free to start with when benching using a standard RAM free-er, so pointless trying for me.. you gain nothing more.

They would help with someone who has too much RAM being used up elsewhere and/or an untweaked system - not fully tweaked/stripped down. They should also help your EVEREST Memory benches if anything. (test them)

If I use the CW, CDT, Maxmem separately each time isolating them, then Maxmem and CDT show the same "little" gain to me in 1M (less than 0.4s) but there is none for CW.

All this seems to depend HIGHLY on your choice of system and setup for some reason I can't lay my finger on and I understand much of how Windows works, but you cannot say "I didn't get it and so its untrue" because then so would CW and Maxmem be. So far you are each just going off individual experience and truthness.

Thanks for sharing the tweak OPB. I don't doubt that you experienced what you presented and shared with everyone else openly without call. Computers can be funny at times and not completely understood at all.

[s7e9h3n]

Dinos, take that stupid pic of me down! I don't think I coulda looked/felt any more cracked out that night while @ 2AM watching Charles and OPB chatting away on my phone for ~30minutes. Add to that the fact that not only did I just drive 45 minutes to meet up with you guys, I had gotten pulled over for speeding and given a sobriety test

FWIW, even I've run a 32m with a C2Q @ 9x400, 600mhz DDR2 4-4-4-4 with absolutely no windows tweaks whatsoever -> ~13m15s. So I have a feeling that although 12m40s is an outrageous time @ that speed, it's within the realm of possibility.....with the right tweaks of course