45nm / DDR3 era Spi discuz.

[VictorWang]

after playing for months ddr3/45nm
i do feel that less and less effective on tweak a 45nm/ddr3 system.
dont know if someone have the same feeling as I do.

here is 445*9=4G pi_1m/32m & 450*10=4.5G pi_1m/32m results.

maxmem=624m, largesystemcache=1, non-pagefile, copy waza 1.36G.


1) 445*9=4G 1m = 11.468sec
http://www.unihw.com/attachments/forumid_27/20071207_22c2e9d5dbeab60bc623VTkwV4c3YGLU.jpg

2) 445*9=4G 32m=11min00.562sec
http://www.unihw.com/attachments/forumid_27/20071207_11b7faa5c482af804a1eh6kDo0o6JqIj.jpg


'coz of FSB limited, so it seems damn hard to break 11min :(
---------------------------------------------------------------------

3) 450*10 1m=10.219sec (can break 10.1sec ?)
http://www.unihw.com/attachments/forumid_27/20071207_bdd937f5d23da7e44168u6Rho97JuW7H.jpg


4) 450*10 32m=9min59.531sec (just break 10min, any possibility for 9min55sec ? )
http://www.unihw.com/attachments/forumid_27/20071207_597316d85c16eb215d77YRos8nGkJU0d.jpg

[Zeus]

Victor,

I cannot speak from own experience cuz i don't have a 45nm cpu nor do i have DDR3.
From observation it looks like the large cache on the 45nm cpu's is responsible for the great performance in SuperPi but the downside is that tweaks like copy-waza take less effect.

Looking back at A64 days with 512mb cache CPU's, copy-waza gave some impressive gains, even in 1M, which reinforces the feeling that more cache means less gain from CW.

For DDR3, it looks like this is only taking a lead over DDR2 when ran at outrages speeds and relative tight timings.
Running more relaxed than 6-7-7 on DDR3 does not give any gain over DDR2 imo.

[Gautam]

Simply put, the effect is the same as what we saw from going from A64 to Conroe. SuperPi is much less tweakable, since most of our tweaks deal with memory. With a larger cache, there is much less memory access.

[BeardyMan]

compare with DDR2 woudl be cool

example at
600 4_4_4 you need DDR3 at etc etc

[dinos22]

yeah i've played a little and found similar things

i have to say that bioses are still in early stages which will improve performance once they become more stable

i've mainly played with GB boards and the dividers are pretty spotty and hard to tweak but when you hit the right ratio speeds are REALLY good >> 45K 1M efficiency non tweaked with memory not maxed out yet at 5GHz....

i am looking forward to another chipset manufacturer introducing their chipsets which will have dividers that can allow RAM to be maxed out at stock FSB :D as well as any other frequency. That is one thing that bugs me about Intel chipsets.....to take advantage of the new RAM you have to hit high FSB and with quads on the way most people will be maxing out around 450 FSB which is just not enough

FSB plays a big role in SuperPi so we really need those E8500 chips to start hitting 600 to see where we're at and whether it's the FSB as well that is lacking

Victor is right though tweaking isn't as effective....i tried mainly 1M stuff but i see 32M is definitely not as good as it used to be with lower cache CPU from you and hipro :(

[erickwok]

after playing for months ddr3/45nm
i do feel that less and less effective on tweak a 45nm/ddr3 system.
dont know if someone have the same feeling as I do.

here is 445*9=4G pi_1m/32m & 450*10=4.5G pi_1m/32m results.

maxmem=624m, largesystemcache=1, non-pagefile, copy waza 1.36G.


1) 445*9=4G 1m = 11.468sec
http://www.unihw.com/attachments/forumid_27/20071207_22c2e9d5dbeab60bc623VTkwV4c3YGLU.jpg

2) 445*9=4G 32m=11min00.562sec
http://www.unihw.com/attachments/forumid_27/20071207_11b7faa5c482af804a1eh6kDo0o6JqIj.jpg


'coz of FSB limited, so it seems damn hard to break 11min :(
---------------------------------------------------------------------

3) 450*10 1m=10.219sec (can break 10.1sec ?)
http://www.unihw.com/attachments/forumid_27/20071207_bdd937f5d23da7e44168u6Rho97JuW7H.jpg


4) 450*10 32m=9min59.531sec (just break 10min, any possibility for 9min55sec ? )
http://www.unihw.com/attachments/forumid_27/20071207_597316d85c16eb215d77YRos8nGkJU0d.jpg




Wa! Super Fast!
whats motherboard?
whats ram? and voltage??:clap:

[VictorWang]

Wa! Super Fast!
whats motherboard?
whats ram? and voltage??:clap:

Asus Max Extreme.
Cellshocks DDR3 1800, 2.2v




yeah i've played a little and found similar things

i have to say that bioses are still in early stages which will improve performance once they become more stable

i've mainly played with GB boards and the dividers are pretty spotty and hard to tweak but when you hit the right ratio speeds are REALLY good >> 45K 1M efficiency non tweaked with memory not maxed out yet at 5GHz....

i am looking forward to another chipset manufacturer introducing their chipsets which will have dividers that can allow RAM to be maxed out at stock FSB :D as well as any other frequency. That is one thing that bugs me about Intel chipsets.....to take advantage of the new RAM you have to hit high FSB and with quads on the way most people will be maxing out around 450 FSB which is just not enough

FSB plays a big role in SuperPi so we really need those E8500 chips to start hitting 600 to see where we're at and whether it's the FSB as well that is lacking

Victor is right though tweaking isn't as effective....i tried mainly 1M stuff but i see 32M is definitely not as good as it used to be with lower cache CPU from you and hipro :(

ratio is a problem.
fsb is a problem especially on 45nm both wolf/yorkies
and bios/spd seems also have some problems :(
maxmem gives only about 1 or 2sec improvement
copzy waza only about 2sec or 3sec. :(






Simply put, the effect is the same as what we saw from going from A64 to Conroe. SuperPi is much less tweakable, since most of our tweaks deal with memory. With a larger cache, there is much less memory access.

we might need to tweak "write" peformance while as copy/cache tweak "Read".
so a Good HDD or I-Ram is needed.
less time and cpu usage when pi_32m "write" data from Ram to HDD might be give another 2 or 3sec faster.






Victor,

I cannot speak from own experience cuz i don't have a 45nm cpu nor do i have DDR3.
From observation it looks like the large cache on the 45nm cpu's is responsible for the great performance in SuperPi but the downside is that tweaks like copy-waza take less effect.

Looking back at A64 days with 512mb cache CPU's, copy-waza gave some impressive gains, even in 1M, which reinforces the feeling that more cache means less gain from CW.

For DDR3, it looks like this is only taking a lead over DDR2 when ran at outrages speeds and relative tight timings.
Running more relaxed than 6-7-7 on DDR3 does not give any gain over DDR2 imo.


i tried lots of different mem_timing combination.
find the tighten tRP and tRAS will help nth but results in un-stable and might be slower.
but tweak sub_timings usually gives a boost (about 2sec).

[mrlobber]

we might need to tweak "write" peformance while as copy/cache tweak "Read".
so a Good HDD or I-Ram is needed.
less time and cpu usage when pi_32m "write" data from Ram to HDD might be give another 2 or 3sec faster.


I-RAM didn't help in Superpi 32M with Conroe already :) What makes you think with Yorkfield/Wolfdale this would be different?

[massman]

These quite fit the thread. Results based on 265 45nm results:

[dinos22]

massman so there are now a fair few sub 45K runs

who are they from

[mrlobber]

These quite fit the thread. Results based on 265 45nm results:

Hoho, quite a nice curve :up: You might want to remove that outlier though (or is it included with an intent since it falls out of all the other pack not by 100%, but by 1000% at least :) ) and might impact the equation of the curve quite significantly.

[Gautam]

How are you getting those graphs massman? Or rather, the data for them.

[BeardyMan]

How are you getting those graphs massman? Or rather, the data for them.

LOL
i guess hwbot my friend :D

[Gautam]

Yeah, I guessed hwbot too. But is there a way to extract all those numbers?

[metro.cl]

Massman in that graph DDR2 and DDR3 is mixed? you could color DDR2 diferently?

[T_M]

You might want to remove that outlier though

Sconded, then there'd be greater visibility of the rest

[massman]

massman so there are now a fair few sub 45K runs

who are they from

Kevin. I have seen very low 45k runs as well with low memory clocks, but I still have to contact that person to ask how exactly he got those numbers.


Hoho, quite a nice curve :up: You might want to remove that outlier though (or is it included with an intent since it falls out of all the other pack not by 100%, but by 1000% at least :) ) and might impact the equation of the curve quite significantly.

I included the 1.2G run to proof that the 'prediction line' (please, what's the correct term in English?) is still quite accurate at very slow speeds.


How are you getting those graphs massman? Or rather, the data for them.

HWBot, Ripping.org, XS, OCX, Coolaler boards


Yeah, I guessed hwbot too. But is there a way to extract all those numbers?

No, manually added in my excel chart.


Massman in that graph DDR2 and DDR3 is mixed? you could color DDR2 diferently?

I'll try, my excel skills are not that good :)

[VictorWang]

great job massman.

have a 5.4G 32m result, not bad :D

http://www.unihw.com/attachments/forumid_27/20071209_8023c8d3f832b798bff97JbkNQ6DpjE1.jpg

[massman]

DDR2 and DDR3 split up (less data as I don't know for every result what kind of ram was used). Can't make any conclusion, but I suspect that low latencies can outperform high bandwidth.

[KTE]

3 of the DDR2 runs and 2 of the DDR3 runs are a cut above the rest. By quite a large margin for something that simple, quick and so often run.

Thanks for the charts. How long do you spend on such things massman? :p:
Man you guys have some time or what. Whose got the furthest left DDR2 run?

[massman]

Long enough to realise that I do not have a 45nm chip to test my theories with :p:.

Furthest left: Rolco 11,265 - 4001 - 600 - 45071,265

[KTE]

Thanks mate. :)

[VictorWang]

45nm 4Ghz 1m < 11.4sec is damn damn fast.
any idea how to achieve that ? :(

[Rol-Co]

45nm 4Ghz 1m < 11.4sec is damn damn fast.
any idea how to achieve that ? :(

Yep..... 11.234 @4g now
:D

[T_M]

Whats ur secret ? :P

[Rol-Co]

Found someting with asus maximus and yorkfield, i will share later....;)
maybe it works with wolfdale too, i want too know first but my chips have a big hold up.

There will be more people in a few day's who can do this, just wait and see.

[massman]

Has SP3 something to do with it?

[before]

DDR2 and P965 aren't dead lol :D

http://img144.imageshack.us/img144/1524/qx9650pi32m9m17094fj7.png

[metro.cl]

Has SP3 something to do with it?

Dont think so i have SP3 and performance is the same.

[KTE]

MSI was advising me to get SP3 for better perf.

[dinos22]

what does MSI know about performance
they have :banana::banana::banana::banana: motherboards lol
unless they were doing their testing at stock OCs :rolleyes:

[Rol-Co]

@ Massman
Sp3 give's a little advantage though , but that's a difference between 11.265 and 11.250.....very little.
The thing i'm talking about give you a 0.3xx difference on the maximus. at every speed, nothing illegall/cheat or something like that.

[eva2000]

Found someting with asus maximus and yorkfield, i will share later....;)
maybe it works with wolfdale too, i want too know first but my chips have a big hold up.

There will be more people in a few day's who can do this, just wait and see.
Ai Clock Twister and Skew ???

[T_M]

Whoa! subscribed

[dinos22]

@ Massman
Sp3 give's a little advantage though , but that's a difference between 11.265 and 11.250.....very little.
The thing i'm talking about give you a 0.3xx difference on the maximus. at every speed, nothing illegall/cheat or something like that.

that's a pretty big difference for SP3 actually wow

looks like i'll have to whack it on heh

sooooo what is this other setting :shocked: :D

[Gautam]

Betcha its static read control.

[dinos22]

Betcha its static read control.

lol you'd think most people benching Pi on Asus would know about it though heh

[Gautam]

It seemed to take people a while to figure it out on the 965 boards...:p:

I say keep it quiet for now, this is a mystery I'd rather solve myself. ;)

[VictorWang]

ya, i'm wondering how to gain 0.3sec difference on pi_1m.

somebody plz share it, lol

[mrlobber]

somebody plz share it, lol

+1 :D

[Rol-Co]

It seemed to take people a while to figure it out on the 965 boards...:p:

I say keep it quiet for now, this is a mystery I'd rather solve myself. ;)

Respect Gautam...:clap:

another two examples.

just win xp pro sp3 with no eram or other tweaks only some windows performance settings thats it.

at high memspeed though

http://82.173.172.10/Rol-Co/stock-os.JPG

but even at 1:1 cl3 its good.

http://82.173.172.10/Rol-Co/cl3.JPG

imagine if Gautam runs this with all his tweaks..:shocked:

[KTE]

The last two tests are pretty commonly known here I reckon. :)

I hardly ever run SuperPi and found the same thing on two runs. 676MHz 5-5-5-15 CR2 was slower than 450 4-4-4-4 CR2 at same CPU MHz.

[Rol-Co]

The last two tests are pretty commonly known here I reckon. :)

I hardly ever run SuperPi and found the same thing on two runs. 676MHz 5-5-5-15 CR2 was slower than 450 4-4-4-4 CR2 at same CPU MHz.

different timings and settings doesn't matter that much, it runs all about the same time, some little difference though at higher memspeed.

but the two screens is all about the efficienty.... show me a non tweaked run with the efficienty of the above firtst screen... 45047 ;)

[massman]

Yeah, you're the third I've seen under 45k ... impressive :D

Like I said, i'm soo looking forward to know what the explanation is :D

[Rol-Co]

Yeah, you're the third I've seen under 45k ... impressive :D

Like I said, i'm soo looking forward to know what the explanation is :D

I've seen a few too but not at a maximus formula board i think...;)

explanation will come ,you seen the video already ?

[VictorWang]

loosen tRFC/ tWTR / tWR ?

[Rol-Co]

loosen tRFC/ tWTR / tWR ?

timings everything on auto at bios vicktor, booted up and ran it .....no memset used....just to show what timings are.

[KTE]

Windows stock theme...? :p:

[VictorWang]

timings everything on auto at bios vicktor, booted up and ran it .....no memset used....just to show what timings are.

:shocked: now plz show us the magic. 0.3sec means 100mhz on 45nm.

[massman]

Is the boost only due to the Maximus or will it happen on other boards as well? :)

[saaya]

trd/pl phase adjust? :D

[camouflage]

:D Victor, memset is bugged and shows 2T no matter what the setting is.

Are You running 1T or 2T ???

:up:

[Rol-Co]

Is the boost only due to the Maximus or will it happen on other boards as well? :)

For now its only maximus + yorkfield.....but wil test with wolfdale too, and i''m waiting on some ddr3 so i can test blitz extreme if it has the same..

edit:
lol,......everybody take a wild guess.....i wil tell later guy's, but in good hands wr 1m will be damaged.

@camouflage/saaya
704 mhz memory 1T ?? wow that would be great isn't it ..... nope 2T , like i said all auto except voltage cpu and memory.

[massman]

Could you post some Sisoft/everest bandwidth and latency screenshots before and after the bios tweak?

Does it has any effect on 3D benchmarks?

Too much questions, I know :p:

[Rol-Co]

Could you post some Sisoft/everest bandwidth and latency screenshots before and after the bios tweak?

Does it has any effect on 3D benchmarks?

Too much questions, I know :p:

ow maybe thats nice , show some difference....what to run?

1m
sisoft bandwith
sisoft cpu ??
any more idea...?

[massman]

lavalys everst cpu and memory benchmark?

example:

http://www.madshrimps.be/articles/DFINF680iLT-T2RReview-anOCdreamcomingtru-thorgal-17068.jpg

[Johnny Bravo]

Rol-co you are loving this arn't you :p:

[Rol-Co]

Rol-co you are loving this arn't you :p:

Yep.....how do you know :D but they keep asking also
how is the tube m8?? did u use it already ??


ok difference

1m
without

http://82.173.172.10/Rol-Co/1mzonder.JPG

1m
with

http://82.173.172.10/Rol-Co/1mmet.JPG

everest without

http://82.173.172.10/Rol-Co/cachememzonder.png

everest with

http://82.173.172.10/Rol-Co/cachememmet.png

:up:

[massman]

With the tweak the bandwidth is lower and latency higher?

[Johnny Bravo]

Yep.....how do you know :D but they keep asking also
how is the tube m8?? did u use it already ??


Yeah I guess you could say it does the job ;)

http://valid.x86-secret.com/cache/273720.png

http://valid.x86-secret.com/show_oc.php?id=273720

[Rol-Co]

@massman

yes ...i guess so :confused: still running some everest benchmarks

@johnny
and a nice job too i see...:clap:

[SF3D]

Ok Rol-Co! You have teased us enough :D

Tell it now.... :yepp:

I'm so bad at tweaking, that I really need this info. I have QX9650 + Maximus combo, so I could test it too.

[Rol-Co]

Ok Rol-Co! You have teased us enough :D

Tell it now.... :yepp:

I'm so bad at tweaking, that I really need this info. I have QX9650 + Maximus combo, so I could test it too.


and youre 1m score at 4g is ??

[SF3D]

The same time is in your first screen--> 11.563 :(

No tweaks, just boot and run.

[camouflage]

For now its only maximus + yorkfield.....but wil test with wolfdale too, and i''m waiting on some ddr3 so i can test blitz extreme if it has the same..

edit:
lol,......everybody take a wild guess.....i wil tell later guy's, but in good hands wr 1m will be damaged.

@camouflage/saaya
704 mhz memory 1T ?? wow that would be great isn't it ..... nope 2T , like i said all auto except voltage cpu and memory.

:D Where is the problem mate? I'm running 934 MHz memory 1T(no tweaks - running even Norton in background - only stability test)::shocked:

http://img525.imageshack.us/img525/7681/42ghz32mte7.jpg

BTW You little bandit - wanna get payed to offer Your secret or what?:confused:

If so tell me, I'll start a collection for You...............:rofl:

[Rol-Co]

M8
you're missing something here, 704 mhz 1T with ddr2 :stick:

and no please, you think bad of me. :slap: ...payed for this?....nope

[SF3D]

But what are you playing here? Why you can't tell us what bios setting you change and get 0.3s off from your pi time?

Waiting...

[VictorWang]

@.@ all the exactly same timing / fsb. but 0.3sec faster, that's holy !#@$&#37;@^%

the best 4ghz 1m i've ever had on 45nm qx9650 = 11.468sec.
normally 11.485/11.5/11.516sec
but never under 11.4sec :(

set 1T in bios. 6-5-5-18-3-48-3-3, static read=enable, transaction boost=auto, dynamice read=auto.

[Zeus]

Could it have something to do with a strapchange?

Like booting up at some speed and using setfsb to get the desired frequency?

I'm not familiar with X38 chipset but it sounds plausible to me. :shrug:

[massman]

Victor, can you ever forgive me?:D . I forgot that you have broken the 45K PP as well :clap: :shocked:

http://www.unihw.com/attachments/forumid_16/20071030_5c0bf02110d114208201Wk0Ri6JLDwsT.jpg

[Rol-Co]

@sf3d
Not playing anything m8, i just told from the beginning i will test some wolfdale and blitz ddr3 first, but people keep on asking questions i can answer. only showing some screens that i'm not talking b sh*%t here.


vicktor
you're running ddr3 board ? will test ddr3 maximus too soon....maybe same thing.

[camouflage]

M8
you're missing something here, 704 mhz 1T with ddr2 :stick:

and no please, you think bad of me. :slap: ...payed for this?....nope

:clap: My fault - I see DDR2......, but I wanted to know if Victor on DDR3 is running 1T with the 6-5-5-18 timings............:confused:

:welcome:

[dinos22]

nice work man

this is crazy good

200 strap?
Phase adjust?
static read adjust?

actually can someone post bios screens for the mobo

this would be fun guessing what it is :D:D:D:D::D:D hahahah

what "unique" features does that mobo have over others since Rol reckons it's something to do with that :)
i thought the bios was the same as other machines


what is 32M like >>> this one's very important too????

and FFS tighten that PL in memset :D

[T_M]

SF3D - please let the man do his work, then im sure he will be willing to share this monster tweak once he is comfortable

[Rol-Co]

nice work man

this is crazy good

200 strap?
Phase adjust?
static read adjust?

actually can someone post bios screens for the mobo

this would be fun guessing what it is :D:D:D:D::D:D hahahah

what "unique" features does that mobo have over others since Rol reckons it's something to do with that :)
i thought the bios was the same as other machines


what is 32M like >>> this one's very important too????

and FFS tighten that PL in memset :D

Did some 32m for you dinos22, boot up at 4g and tighten up timings, and large system cache, but i have no idea if its good....not realy a 32m guy. :)

http://82.173.172.10/Rol-Co/32myork.JPG

You tell me ....:shrug:

[VictorWang]

Victor, can you ever forgive me?:D . I forgot that you have broken the 45K PP as well :clap: :shocked:


yup, i know that, but i can't break 11.4sec 1m with 4G :(





victor
you're running ddr3 board ? will test ddr3 maximus too soon....maybe same thing.

sure, ddr3 mobo, plz show us ur results on ddr3 :D :up:

[dinos22]

Did some 32m for you dinos22, boot up at 4g and tighten up timings, and large system cache, but i have no idea if its good....not realy a 32m guy. :)

[IMG]http://82.173.172.10/Rol-Co/32myork.JPG

You tell me ....:shrug:

VERY slow

[Rol-Co]

ok only for 1m i think.....:D

[proxis]

32M "with" and "without" tweak, please (same OS optimisations) :)

[dinos22]

32M "with" and "without" tweak, please (same OS optimisations) :)

yeah that's actually a good question

[Rol-Co]

32M "with" and "without" tweak, please (same OS optimisations) :)

like i thought because of the other test, its not "helping" in 32m

http://82.173.172.10/Rol-Co/32myork4g.JPG

http://82.173.172.10/Rol-Co/32myork4g2.JPG

Pure 1M power...

[T_M]

"Rol-Co's 1M-Powa Tweak"

[Rol-Co]

"Rol-Co's 1M-Powa Tweak"

:rofl: ....:up:

[massman]

That would explain why the bandwidth is lower and latency is higher ... :)

[hipro5]

For 1M and ONLY:
"WITH" = Bios "NOT fully" supporting York and
"WITHOUT" = Bios fully supporting York.......;)

[Rol-Co]

For 1M and ONLY:
"WITH" = Bios "NOT fully" supporting York and
"WITHOUT" = Bios fully supporting York.......;)

Ah george , now youre spoiling the fun....:D

why don't you have a sub 45k score on that 7.7 @ blitz.

[hipro5]

Ah george , now youre spoiling the fun....:D

why don't you have a sub 45k score on that 7.7 @ blitz.

Because I haven't "fully" tested it yet....These days I figured out what's going on and what settings shall I put....Though I'll re-try it.... :D

[DeDaL]

Is it :D


http://www.topmods.net/dedal/img_7252.jpg

[hipro5]

Yes......That's it..... :D

[DeDaL]

It works only on 1M. (3D not compared yet)

I found this "tweak" week ago :D

[Rol-Co]

George
Can you test if the blitz got the same thing??, have one overhere but no ddr3 yet.
want to know... so i can get a nice 7.7 too ;)

[SF3D]

What a tweak! :rolleyes: Damn, I flashed my bios on monday.

Thanks George! :up:

[hipro5]

What a tweak! :rolleyes: Damn, I flashed my bios on monday.

Thanks George! :up:

You MUST say thanks to Rol-Co that teased you for it and not me.... ;) :)

EDIT: It's ONLY for Super-Pi 1M and ONLY.....NO OTHER benches...... ;)


George
Can you test if the blitz got the same thing??, have one overhere but no ddr3 yet.
want to know... so i can get a nice 7.7 too ;)

Yes I'll give it a go propably this weekend...... :)

[Sampsa]

Yes I also noticed this "tweak" 2 months ago with Maximus Formula

0401 bios is ~0,3s faster in Spi1M than newer ones (tested against 0504 which was the newest at that time)

http://support.asus.com/download/download.aspx?SLanguage=en-us&model=Maximus&#37;20Formula

[T_M]

Thanks for sharing version number.
Suprising how many people knew about this!

[KTE]

This is a tweak or a system bug? :confused:

[camouflage]

:D Haha, funny things happen here. BTW, 1M does not really say much about RAM twaeking cause today's CPUs have a large L2 cache, especially the QX..;)

That's why I'm a 32M guy............:ROTF:

:up:

[Rol-Co]

But it sure can make a difference between 8.2xx and 7.9xx s-pi ;)

and it sure kept you all buisy..... :rofl:

[Extera]

well, everybody knows now...

Now we need to push harder rol-co... :rofl:

[Kain XS]

Which difference in BIOS does that difference @ 1m?
Somebody knows?

[Rol-Co]

0401 at maximus formula se

[eva2000]

interesting sounds like non-supporting yorkie bios might think the QX9650 is 1066MHz bus cpu and default to 266FSB strap latencies versus 1333FSB bus cpus ???

[Rol-Co]

also works with qx9770 :D

http://82.173.172.10/Rol-Co/fastpc.JPG

[Kain XS]

interesting sounds like non-supporting yorkie bios might think the QX9650 is 1066MHz bus cpu and default to 266FSB strap latencies versus 1333FSB bus cpus ???

So, changing starp would be the same than doing this=?

[Zeus]

interesting sounds like non-supporting yorkie bios might think the QX9650 is 1066MHz bus cpu and default to 266FSB strap latencies versus 1333FSB bus cpus ???

That is most likely what happens.

Perhaps one can simulate this by booting at 266 FSB and use setfsb to desired frequency.

Has anybody tried yet?

[hipro5]

That is most likely what happens.

Perhaps one can simulate this by booting at 266 FSB and use setfsb to desired frequency.

Has anybody tried yet?

Nope....It's not this......I ALWAYS play on 266 strap.... ;)

[Rol-Co]

That is most likely what happens.

Perhaps one can simulate this by booting at 266 FSB and use setfsb to desired frequency.

Has anybody tried yet?

not working....you can try

tried several settings here with a suported bios but never came close...
in everest you see some diference in subtimings, but those timings you can't find it at the bios.
i think some hidden board timings/settings.

[Rol-Co]

with this, first sub 10 score at 4515.....:shocked:

[eva2000]

maybe 266FSB Strap for 1066Mhz bus cpus is different from 266FSB Strap for 1333mhz bus cpus.

[KTE]

It can't be faster consistently unless it involves lowering some hidden latencies somewhere. That's all that can really happen but good to know. Cheers. :)

[hipro5]

I think that it has to do with the "instractions" of the CPU given and nothing more.....Maybe it doesn't work it's SSE3 or something like that....Who knows....?.... :)

[VictorWang]

ah...yes, I remembered, my 5689mhz 7.969sec is done on Blitz Extreme 0804 bios.

but after I update to 0904...I never reached that speed ag. :(

i'm too lazy to flash back the bios :D

[BenchZowner]

i'm too lazy to flash back the bios :D

Buy a new motherboard :D :rofl:

Ok, no more jokes :D

This thing is weird.
What's making the SuperPi 1M faster ( with older BIOS ) and 32M slower at the same time ?

Instructions can't be, because as far I remember SuperPi uses only the SSE instructions ( the 1.5 XS version ) [ not SSE2/SSE3/SSSE3/SSE4 ]...nothing new here.

What could it be ?
Somebody checked if the same thing happens with a C2Q KF ( KentsField ) ?
Just to check if it's just the newer BIOS degrading the 1M performance anyway.

[Rol-Co]

Only difference i found was:old/new

Read delay phase adjust : +1/normal
dim 1 clock fine delay 7t/9t
dim 2 clock fine delay 2t/5t
dim 3 clock fine delay 6t/9t
dim 4 clock fine delay 2t/5t

but personaly i don't think that is making the difference to the 0.3 sec

[before]

@Rol-Co or someone else, can you please take screenshots of MCHBAR sequences for me? If yes, just drop me a pm, and I'll explain how to do.

Dunno but may be I can identify what's going on.

[sobih25]

Have you seen 500 fsb yorkie 32mpi stable scores?

I've played with this cpu and get some Vic's comparable scores, (higher fsb, lower ddr3, strap 200). I'll try better ddr3 soon.

500x9 32m pi
http://img87.imageshack.us/img87/2038/1059532mi1.th.jpg (http://img87.imageshack.us/my.php?image=1059532mi1.jpg) http://img246.imageshack.us/img246/4592/1100453as4.th.jpg (http://img246.imageshack.us/my.php?image=1100453as4.jpg)

500x8 32m pi
http://img91.imageshack.us/img91/9733/0959797gd5.th.jpg (http://img91.imageshack.us/my.php?image=0959797gd5.jpg) http://img246.imageshack.us/img246/5246/1001031df4.th.jpg (http://img246.imageshack.us/my.php?image=1001031df4.jpg)

1m pi don't likes me or I don't know some tweaks, get just near 11,5 at 500x. don't try 804 bios atm.

[SeLecT]

Nice testing!

Its good to see the FSB 500 :D

[sobih25]

small update brought by kielbasa squad. this time the same yorkie, the same mb, better ddr3 (strap 266, 1:2 divider and 7-6-5-18 timmings) gives a 5s better times than few days ago :D.

http://img166.imageshack.us/img166/580/0954625hu2.th.jpg (http://img166.imageshack.us/my.php?image=0954625hu2.jpg)

http://img107.imageshack.us/img107/5674/1054562wi3.th.jpg (http://img107.imageshack.us/my.php?image=1054562wi3.jpg)

it's not really maxed out yet;).

[pmp]

Nice testing!
Its good to see the FSB 500 :D
But it's MB dependant.
With X48 driven P5E3 Premium it seems, it's not so obvious like with BE. (i meant FSB500).

[massman]

Has someone figured out why that particular bios is so much faster?

Anywho, thought I posted these here: graphs from 499 45nm chips (DDR2/3 split up counts 115+233=348 results):

http://i154.photobucket.com/albums/s261/Massman-/XS/4100time.png

http://i154.photobucket.com/albums/s261/Massman-/XS/DDRsplit.png

http://i154.photobucket.com/albums/s261/Massman-/XS/PPDDR.png

Now, I have made a few drawings on some of the graphs as well, as it seems that something rather weird is happening, at least it looks weird to me :).

http://i154.photobucket.com/albums/s261/Massman-/XS/4100timeedit.png

This is the graph which shows the Mhz vs 1M information. Normally, all these points should be representated by 1 y= comparison, however it seems that there are two comparisons instead of one.

http://i154.photobucket.com/albums/s261/Massman-/XS/PP.png

Performance product. Here we seem to spot two groups of performance products, why?. If you look at the DDR2/3 split up graph, you see it has nothing to do with the ram. It seems like there's an empty space between 45500 and 46000 :confused:

[before]

Any of you noticed a difference in memory performance that is related to a boot FSB? Not like a cut-off as it was for P965 chipset, but sometimes a particular FSB, I mean a single FSB value among others gives far more bandwidth than expected...

I have posted it in another thread, but nobody seems to be interested because I've got no reply; BTW it might be something important to look at.

Ok; I experienced that while investigating MSI P35D3 functionning, but maybe it exists for other brands with P35 chipset.


Found something interesting on this board... :rolleyes:

Started with a boot @460FSBx11 to perfrom Everest benchmarks and then rebooted @450FSBx11; and so on; down to 380FSBx11. Last boot was done @399FSBx11, and system was overclocked with setFSB under windows up to 440FSBx11. (Available) Timings were the same, divider remained unchanged; I've just played with PL but results are given individually.

It looks to work only with FSBSEL jumpers set to 333. :)

http://img339.imageshack.us/img339/7584/msibootmrsl2.png

http://img184.imageshack.us/img184/6683/msibootmwbt3.png

http://img339.imageshack.us/img339/5696/msibootmcay0.png

http://img339.imageshack.us/img339/2103/msibootmliz6.png

A boot at 399FSB shown to be far more efficient than anyothers. Memory Read, Copy and Latency appeared significantly improved (even with PL8).

[KTE]

I hope its not my eyesight massman, but are your graphs really showing sub 44750 with DDR2? :confused:

Is this only Kev by any chance as well?

[massman]

Sub 45k:


Name Score Cpu Memory PP cpu

OPB 7,735 5739,4 551,9 44394
OPB 7,89 5657,7 638,5 44639
OPB 8,25 5412,4 563,8 44652
Rolco 11,234 4000,2 640 44938

OPB 7,468 5913,8 844,8 44164
OPB 7,578 5811 1020,3 44036
OPB 7,578 5811,6 715,3 44040 (maybe DDR2, not sure)
Victor 9,469 4750,4 950,1 44982

Coola 7,813 5746,4 44897


Edit: I do not have all 1M's of course, so there might be even more sub45k's out there. I have used HWbot's and Ripping.org's data along with some scores of OCX/XS/Cooaler.

[Pt1t]

Nice tweak Rol-Co, i m going to try it on maximus Extreme :D
... if first bios doesnt reconize correctly the yorkfield.

[massman]

Any of you noticed a difference in memory performance that is related to a boot FSB? Not like a cut-off as it was for P965 chipset, but sometimes a particular FSB, I mean a single FSB value among others gives far more bandwidth than expected...

I have posted it in another thread, but nobody seems to be interested because I've got no reply; BTW it might be something important to look at.

Ok; I experienced that while investigating MSI P35D3 functionning, but maybe it exists for other brands with P35 chipset.

To be honnest, I'm not surprised to see this :). When testing my DFI P35 (well, I'm still testing:p) I noticed that 500x7 is 32M stable when I boot at 500FSB. When I boot at 399FSB, 490x7 wasn't even 1M stable!

I reckon the same problem would be visible at 265 and 332 FSB.

[dinos22]

hey before i thought my Gigabyte was weird as i started to notice some strange behaviour like that heh :up:



absolute best FSB i can get out of SS cooling and 9xmulti with E8500 and SS

need that half a multi :(

45K efficiency too :D on DDR2

http://img132.imageshack.us/img132/3343/8766s1msuperpi5137mhzhh5.png

[massman]

Heavy stuff, Dino. Do you still have headroom? You're still not sub45k :D (3Mhz less or 0,06s less)

[massman]

Damn, my previous drawing was pretty crappy, here's what I really meant by two lines:

[dinos22]

Heavy stuff, Dino. Do you still have headroom? You're still not sub45k :D (3Mhz less or 0,06s less)

of course
i want to break 45K past 5.5GHz ;) :D

if i could only find a motherboard that can handle these RAM sticks it would be soooooooooooooooooooooooooooooooooo much fun

http://img89.imageshack.us/img89/9683/980mhz7772022v32msuperpsw4.png

this particular gigabyte is shaving off 10-15MHz FSB on another board :(
and slot 3 and 4 clock a LOT worse so no dual channel serious DDR3 benching until i find something else
my asus P5E3 DLX is dead :(

[massman]

to break 45k, you don't need DDR3 ;)

[dinos22]

to break 45k, you don't need DDR3 ;)

i would like to smash it :D
DDR2 is possible to
there are no tweaks there at all LOL in my runs

[massman]

Seriously? Raw system power at 45k PP? :eek:

[dinos22]

Seriously? Raw system power at 45k PP? :eek:

yeah man :D

but 32M times suck lol so gotta work that out heh

[massman]

Should do more research, but I reckon it's all about the 570FSB, no?

[dinos22]

Should do more research, but I reckon it's all about the 570FSB, no?

nah
i can do that at lower FSB too

it's this board and bios

[massman]

Same advantage as the 0401 bios of the Maximus Formula SE? Do you have a higher PP when using another bios of the same board? If so, what's the difference with the other biosses?

So many questions :p:

[dinos22]

GB has crap bioses aside from first releases IMO for extreme overclocking
for normal use and stuff they should be similar donno

[massman]

So F2 outperforms F5 (with updated 45nm support)?

LOL, it seems like bios updates slow down the 45nm cpu's :D

[dinos22]

So F2 outperforms F5 (with updated 45nm support)?

LOL, it seems like bios updates slow down the 45nm cpu's :D

problem is that you can't clock FSB on anything other than F2

so if you plan to be stuck around 450MHz FSB on a dual core that can do another 100+ good luck ROFL

clock for clock it was a touch slower too
1M uses CPU cache should it should be some kind of indicator i guess and first bios is just fine

i asked gigabyte for half multies to be added to F2 DDR3 bios and F3 DDR2 bios

if that happens all is good and i will not change bioses after i get the half multies heh

[dinos22]

to break 45k, you don't need DDR3 ;)

true that

http://img185.imageshack.us/img185/1848/44963kefficiencyzf3.png

not optimised yet

[massman]

Still no tweaks?

[dinos22]

still no tweaks

ran it again

http://img182.imageshack.us/img182/3452/44909kefficiencytt7.png

[massman]

As far as I know, you now are the most efficient bencher behind OPB and Coolaler ... :D

[Zeus]

As far as I know, you now are the most efficient bencher behind OPB and Coolaler ... :D


He's just another cheater. :rofl:

[massman]

Haha, Zeus :)

Just wondering at what PP he would be with tweaks and even better rams :)

[dinos22]

:rofl: :rofl: :rofl: :rofl: :rofl: :rofl: :rofl:

well i have been hanging out at OCX :shocked: :D


anyways here is the problem

no tweaks here either 'cept for LSC of course

http://img523.imageshack.us/img523/7461/36ghz32msuperpi616mhz45uv8.png


it could be the case that my OS is just really dirty

did i mention i keep reusing the same install on all Intel chipset motherboards cause i'm lazy to install an OS from scratch :eek: :D

[dinos22]

Haha, Zeus :)

Just wondering at what PP he would be with tweaks and even better rams :)

my 1M tweaks are not that great
probably just low maxmem value

i reckon it would probably go down to 44.85K at most

[massman]

Low maxmem? Why is the real memory a number with 10 digits then?

[dinos22]

Low maxmem? Why is the real memory a number with 10 digits then?

i haven't applied any tweaks yet
i am saying when i bench 1M i just lower the maxmem value and start benching

[kiwi]

hey Dinos, you don't have a separate HDD for benching? :eek: Why no perf. level 6/7 @ 513?

[dinos22]

hey Dinos, you don't have a separate HDD for benching? :eek: Why no perf. level 6/7 @ 513?

this board can't do better than 9 LOL
maybe 8 may be ok but yeah it's not that great PL wise

the thing is my 32M times were very nice with E6850 so not sure what's up

[massman]

Maybe bios related?

[dinos22]

Maybe bios related?

this is the same as F3 bios
GB techs just added half multies

F3 bios is what i usedwith conroe OCs

but i guess it could be bios related for Wolfdales donno

[massman]

This thing really is bios related, can't go past that. There are only three (DDR2) motherboard/bios combinations I have seen hit 45k or even sub 45k and those are:

Gigabyte X38-DQ6 - Bios: F2
Asus Maximus Formula SE - Bios: 0401
Asus P5K-E - Bios: 0503

Just checked cpu compatibility of those bioses: none of these bioses support 45nm chips as support has been added in later revisions.

[jmax_oc]

This thing really is bios related, can't go past that. There are only three motherboard/bios combinations I have seen do close to 45k or even sub 45k and those are:

Gigabyte X38-DQ6 - Bios: F2
Asus Maximus Formula SE - Bios: 0401
Asus P5K-E - Bios: 0503

Just checked cpu compatibility of those bioses: none of these bioses support 45nm chips as support has been added in later revisions.
You can add P5E3 WS with 0124 Bios. Hipro was this your BIOS during spi benches ? Me yes (7,891s @ 5733Mhz)

[massman]

Thanks, Jmax, Hipro was using the Blitz extreme and bios 0904.

There seems to be a difference between DDR2 and DDR3 motherboards, though.

[DMH]

Massman they don't need an update because they already support them

[massman]

Not officially :).

Those three DDR2 boards give awesome 1M performance with biosses before the micro-code update of the 45nm cpu's. With other biosses it seems to be less performant (or I haven't payed attention to it, which could very well be)

[dinos22]

gigabyte does not have a single decent bios aside from first release for both DDR2 and DDR3 boards :( (as in poor FSB performance and very buggy RAM dividers)

i can do similar efficiency with DDR3 GB board too so you can add that to the list lol

and i used two bioses this F2 0115 and F3 on DDR2 board


btw massman F2 0155 bios is basically F3 bios with yorkfield/wolfdale support massman ;)
same efficiency

[dinos22]

with F6j bios 32M SuperPi is quicker with QX9650 :S

http://img175.imageshack.us/img175/7581/signature1f58f596nl2.png

[Rol-Co]

Dinos22

Its bios related, not supporting bios is fast @1m and slow @32m so newer bios give you better performance on 32m .

@massman
Blitz extreme ddr3 has the same issue....45k on DDR3 :)

http://rol-co.nl/hwi/3600-45k.jpg

[before]

Its bios related, not supporting bios is fast @1m and slow @32m so newer bios give you better performance on 32m .


It is. @IxTk we are currently investigating this matter. So far, I can only say that there are changes among both MSR and MCHBAR registers. Not sure if we'll find exactly what's happening.

BTW We are looking for dumps done with other motherboards than the P5E3 WS Pro which is the one we actually use. Making comparison with several different boards can help saving so much time.

If you're using Gigabyte X38-DQ6 - Bios: F2; Asus Maximus Formula SE - Bios: 0401; Asus P5K-E - Bios: 0503; or any other one with which the phenomenon occurs... Could you please save CPU-Z registers dumps (.txt) and then send them to me? What we need are 2 dumps, (1) with pre-45nm bios and (2) with post-45nm bios; both done under the same conditions (I mean the same bios settings).

If you wanna help but still have some questions, please feel free the ask me in pm... :up:

[dinos22]

i'll help out next time i bench

i will also use X38T-DQ6 (DDR3 board) early bios to see it does the same

[kiwi]

Same here with Blitz Extreme :) I was also going to investigate mchbar registers, could be on of the reasons

[before]

Thanks guys :toast: Any help will be greatly appreciated! :up:

http://www.ixtremtek.com/images/AsusP5E3WSPRO/QX9650PI1M@7,984.png

Working on efficiency here. :)

http://www.xtremesystems.org/forums/showthread.php?p=2746325#post2746325

[Rol-Co]

You need to shave of 0.145 before :D

dumps coming up,have new cpu and try te get you dumps of maximus formula/blitz extreme/striker II formula

[before]

Yeah; thx for your help!

An hardware tweak is needed :) Actually, it looks quite improbable to get sub-45k @ nearly 5.7GHz.

[before]

Nobody wants to help? :(

What you guys can get in term of lowest pp with a post-45nm post? Any sub-46k out there?

[T_M]

so far my lowest was around 46.2

[before]

Mine was 45900pp; dunno if that's because I've started to play with the right MCHBAR offset, or if that's thx to a well tweaked OS.

[Misko]

Mine was 45900pp; dunno if that's because I've started to play with the right MCHBAR offset, or if that's thx to a well tweaked OS.

45900 with which multi and clock?

[before]

400.4FSBx10 1:2 7-6-5-15 IIRC

[Misko]

400.4FSBx10 1:2 7-6-5-15 IIRC

hmmm hard to tell, you are on the edge, I can go 460xx, maybe even 46 flat with DFI, but not under it, and can make 449xx with 45k BIOS with Asus on 600fsb 1:1 ddr2. What does runs say if u just playing with mchbar? (leave some win setting and tweaks and just play with mchbar, any different scores?)

[kiwi]

Before, that's 4004Mhz? :)

When I was doing 32M low clock @ 8x500 833 6-5-5 I also did quick 1M (time 11.500) and it was flat 46K or slightly less because cpuz read 3 values - 3999.9, 4000.0 and 4000.1.

[before]

Before, that's 4004Mhz? :)

When I was doing 32M low clock @ 8x500 833 6-5-5 I also did quick 1M (time 11.500) and it was flat 46K or slightly less because cpuz read 3 values - 3999.9, 4000.0 and 4000.1.

Yeah; but 500FSB is very helpful. Anyway; I can't test it more because my cooler master PSU have just died killing my VGA + P5E3 WS Pro... :mad:

[eva2000]

Yeah; but 500FSB is very helpful. Anyway; I can't test it more because my cooler master PSU have just died killing my VGA + P5E3 WS Pro... :mad:
ouch sorry to hear man :(

[MaSell]

My best was 8.860s with 9x570 1:1 4-4-4, so 4545x :) . And archived at not 1M boosted mobo :)

[dinos22]

Yeah; but 500FSB is very helpful. Anyway; I can't test it more because my cooler master PSU have just died killing my VGA + P5E3 WS Pro... :mad:

:shocked: :shakes: :(

[kiwi]

My best was 8.860s with 9x570 1:1 4-4-4, so 4545x :) . And archived at not 1M boosted mobo :)

What mb and bios?

[N3RO]

It is. @IxTk we are currently investigating this matter. So far, I can only say that there are changes among both MSR and MCHBAR registers. Not sure if we'll find exactly what's happening.

BTW We are looking for dumps done with other motherboards than the P5E3 WS Pro which is the one we actually use. Making comparison with several different boards can help saving so much time.

If you're using Gigabyte X38-DQ6 - Bios: F2; Asus Maximus Formula SE - Bios: 0401; Asus P5K-E - Bios: 0503; or any other one with which the phenomenon occurs... Could you please save CPU-Z registers dumps (.txt) and then send them to me? What we need are 2 dumps, (1) with pre-45nm bios and (2) with post-45nm bios; both done under the same conditions (I mean the same bios settings).

If you wanna help but still have some questions, please feel free the ask me in pm... :up:
45112 with P5K-E (0307), have some interess?

BTW, do you know if Blitz 965 has the same issue?

Thanks,

[Misko]

I can't test it more because my cooler master PSU have just died killing my VGA + P5E3 WS Pro... :mad:

:( sorry to hear that

[before]

Thanks guys for those kind words. :)


45112 with P5K-E (0307), have some interess?

BTW, do you know if Blitz 965 has the same issue?

Thanks,

Could you please make some dumps with your P5K-E?

Dunno about the Blitz 965.

[hipro5]

Yeah; but 500FSB is very helpful. Anyway; I can't test it more because my cooler master PSU have just died killing my VGA + P5E3 WS Pro... :mad:

Do you mean the 1250Watts Cooler Master PSU?.....If yes, when I show/investigate it, I didn't even bother to test it....so I was lucky I guess..... :D

[before]

Do you mean the 1250Watts Cooler Master PSU?.....If yes, when I show/investigate it, I didn't even bother to test it....so I was lucky I guess..... :D

Yep; this one... ;) Maybe mine was faulty.

[N3RO]

Could you please make some dumps with your P5K-E?

Some preference about CPU freq.? Default is enought?

Thanks,

[MaSell]

What mb and bios?

P5K pro with some "middle" date bios, don't remember now

[before]

Some preference about CPU freq.? Default is enought?

Thanks,

Lets say 399FSBx10; anyway it doesn't mind at all. Just be sure to boot with same parameters. ;)

[Misko]

New run by Team Serbia :up:

http://img260.imageshack.us/img260/9614/e85004ghz11469syl4.png

[Pt1t]

P5K premium Bios 0202 , no optimisation (i did 11.600 with same os and bios > 0202)

http://www.pt1t.net/p5kpremium/4gp_p5k_premium.PNG

[ME4ME]

http://img163.imagevenue.com/loc912/th_26900_4000mhz_-_11.265_122_912lo.JPG (http://img163.imagevenue.com/img.php?loc=loc912&image=26900_4000mhz_-_11.265_122_912lo.JPG)

on P5K-E wifi/ap with 0602 bios:)