Sandy Bridge - Bulldozer - Computer Chess Test!

[Kensek]

That's the key - turn off everything. I also set the priority to Realtime also.

I can't chase you any more. My 965 tops out at 4.41 GHz.

I might have to go to Fry's and pick up a $230 i7 920.

[rge]

yeah, turning everything off made big difference. I am near my max as well, I can run it no prob at 211x21, 1.55v cpuz, but 212x21 shuts down ~ 1 min mark using 1.6cpuz. maybe later in year I will run at ?1.62V 212x21, if I knew for sure I would get 9, might try later on, though may not make it.

[Kensek]

Oops almost had a 9.3. Went back to the G. Skill Ram DDR3-2021 at 8-8-7-20 1N

Had to settle for a 9.2 score. Still ran at 35 x 126. 4.41 is all my 965 can do. Temps now at 100% load is 58C.

[speckled10]

Has anyone tried this with Windows 7 Beta to see if its as fast as WinXP using same hardware ?

[JPQY]

Kensek has past the 9.2Milj.nodes/sec. -> 9.223.516 makes his position 1 stronger!

But rge comes close to the 9 Milj.nodes -> 8.951.825 and take the second place!

2009 just started and we have already 2 higher scores...

And i'am also curious what the chess program will do under Windows 7! ...soon we will know!

List updated at page 1.

JP.

[rge]

That is interesting, this program will use all the bandwidth you throw at it. I even got 23x196 or 4.5 to run to 20th run before crashing, but was only at 9.13 there, 21 run would probably be just over 9. No way will I get 9.2, unless can get memory to work at 1960 or 10x. Maybe next weekend will have time to play with secondary timings, but I dont think my mem can do over 1870 even with 1.98v, just 1600 rated.

[Kensek]

That is interesting, this program will use all the bandwidth you throw at it. I even got 23x196 or 4.5 to run to 20th run before crashing, but was only at 9.13 there, 21 run would probably be just over 9. No way will I get 9.2, unless can get memory to work at 1960 or 10x. Maybe next weekend will have time to play with secondary timings, but I dont think my mem can do over 1870 even with 1.98v, just 1600 rated.

I tried with my 965. Running 22 x 200 it scores a lot lower than 35 x 126. Even Tried 157 x 28 (4.396)
All are roughly 4.4 GHz. But the Higher Multi helps me immensely.

[rge]

I tried with my 965. Running 22 x 200 it scores a lot lower than 35 x 126. Even Tried 157 x 28 (4.396)
All are roughly 4.4 GHz. But the Higher Multi helps me immensely.

Yep, it must be multi or the associated higher qpi (you may know, I cant test higher qpi). It is not bandwidth or uncore, though uncore helped more than bandwidth.

I got 9.05 no problem, by just increasing to 212x21, this run in pic was at 8x mem 7,7,7,21, but even 6x mem at 8,8,8,24 came in at 9.03. I ran my ram single channel, dual channel, triple channel all with same score ie 9.0 to 9.05. With triple channel I ran mem at 6x multi with 7,7,7 timings and 8,8,8 timings, 8x at both timings, and even got my 1600 ram up to 10x..2120 with 1.7qpi and over 2v DRAM to run to 20th run. All those runs given same MHZ, bandwidth was almost meaningless, all within statistical error. Most all were between 9 and 9.05 on 21 run, and 9.07 to 9.13 20th run.

[Kensek]

@ Rge - at least you can run above 4.4 GHz easily.

My i7 965 maxes out 4.42 GHz at 1.585v after vDroop. But CPU Temp is only 58C. It may have to do with the fact that Asus R2E has a TDP power limit on the CPU.

I'd would've considered the Blood Rage, but I don't like the PCIe slot layout the board has.

I just ran a 9.238.112 but I'm not submitting it. I'll submit something if it is over 9.250

[rge]

58C...is that load? My 24/7 temps on max load are 70C at 4.2 and 1.376v load. I realized after posting the run in pic, I actually hit 100C, see thermal LOG in red under core 1 on real temp...LOL. That one probably throttled at one point, maybe right at end, but did not affect score, have other runs that did not throttle or hit 100C with same score. I need to open a window and let some cold air in when running at that volts....27C ambient is not cutting it.

[Kensek]

58C...is that load? My 24/7 temps on max load are 70C at 4.2 and 1.376v load. I realized after posting the run in pic, I actually hit 100C, see thermal LOG in red under core 1 on real temp...LOL. That one probably throttled at one point, maybe right at end, but did not affect score, have other runs that did not throttle or hit 100C with same score. I need to open a window and let some cold air in when running at that volts....27C ambient is not cutting it.

58C @ 100% load. That's because I moved my WC setup into my sparsely heated bedroom above my garage. It is 48F in there.

My 965 needs 1.48v to run at 4.2.

My Old i7 920 needed 1.44v to run Prime stable at 21*200.

You got a great chip

[rge]

I wanted to see what mine would do when not fighting 98-100C temp. opened two windows and closed door and vents and let room cool down from 26C to about 12C or 54F. That way my load temps were running about 84C max instead of 100C. max i could get was 215x21 or 4.5 ghz, also ran on extreme perf, 7,7,7 21. I need to try some different bioses, but did not have much time to play with it.

But I did hit 9.237.7. What is interesting I hit 20 on a down spike, and 21 on upspike, so 21 ended up higher than 20. I think if I ran it 20x, might get more just by luck of variability. My runs varied from 9.15 to 9.237, had 1 run that had much higher 20 score 9.28....then 21 took a nose dive at last seconds to 9.21....thought I was going to hit 9.25.

[JPQY]

Today a new leader: rge -> 9.237.692 Nodes/sec.

@rge:"might get more just by luck of variability" -> I call it not luck,it's normal,because the chess engine has to do millions and millions calculatings,and you have so many
posibilities that he at some depth find a other move or combinations from moves that you get a different score! And the drop can also to make with HT(Hash Tables) are full like on your screen 99% (in Analyse window,right) explained on page 4,message 2.

And also today i can publish the first results from the Phenom II X4 940!
73) TogaII142JD-4cpu 21/45 03:44 125.657.505 4.228.992 +0.18 Phenom II X4 940 @3.612Ghz DDR2: 5-5-5-15 xoqolatl New
83) TogaII142JD-4cpu 21/48 04:13 86.724.835 2.826.454 +0.17 Phenom II X4 940 @3.0Ghz DDR2: 5-5-5-15 xoqolatl New

List updated!

JP.

[bowman]

Okay, now I'm really wondering what's going on with that four-way Opteron server down there. Something is seriously off.

[rge]

One last run, forgot to try my 23 multi (turbo enabled). Again with windows open this evening, around 10C or 50F in the room. 23x198 or 4.55 Ghz, mem still 8x 7,7,7 21. Load temps about same mid to high 80's, score was 9.246.923.

[Kensek]

One last run, forgot to try my 23 multi (turbo enabled). Again with windows open this evening, around 10C or 50F in the room. 23x198 or 4.55 Ghz, mem still 8x 7,7,7 21. Load temps about same mid to high 80's, score was 9.246.923.

I just tried again - and I could only muster up a 9.155 with my usual 35x126 settings. Tried for an hour.

The room started at 52F and in an hour was up to 60F.

The Gig Extreme is a great board for 920s and 940s. My 920 kicked but when i owned the Gig and ran with F4c. It ran HT and Turbo at 21x20 Prime stable at 1.46 BIOS volts and 1.406 VDroop.

But the Gig sucked with the 965 in it.

Oh Well Great Job. The Cold really helps in OCing.

[rge]

I just tried again - and I could only muster up a 9.155 with my usual 35x126 settings. Tried for an hour.

The room started at 52F and in an hour was up to 60F.

The Gig Extreme is a great board for 920s and 940s. My 920 kicked but when i owned the Gig and ran with F4c. It ran HT and Turbo at 21x20 Prime stable at 1.46 BIOS volts and 1.406 VDroop.

But the Gig sucked with the 965 in it.

Oh Well Great Job. The Cold really helps in OCing.

Thanks, yep cold is great.... wish I could hang the computer out the window, have to cool the room down then heat it back up when wife gets back. Would like to see what, if any, voltage decrease I would get with the lower temps at my 24/7 settings.

[jcool]

Well contrary to what some people said, i7 does respond to cold. I tested my new 3838A 920 on air and was rather disappointed that it needed 1,440V to run 4Ghz prime with HT on. Now on phase, I can run almost 4,2Ghz with the same Vcore, which makes the chip seem halfway decent... but i haven't tested a lot yet so we'll have to see.

[JPQY]

Thanks for your new nr1 high score rge -> 9.246.923

Kensek & rge move there system to a cooler room or even open window to have lower temps,they stay in a cold place to get higher results!
So now we get even with High Multi x Lower Bus Speed (35x126) or Lower Multi x High Bus Speed (23x198) scores above 9Milj.nodes/sec.

JP.

[Kensek]

[rge]

Run 3 of about 10 runs. Still at 23x198, mem timings same. 23x198 is very stable at 15C.

When gets down to near 0C one night, should be able to make it to 21 run consistently at 23x200 or 4600, had one at 9.5 at 20th run at 4600, then froze at 1:45 still on 9.4...sucked. But temps too high for that now, and rising with afternoon.

[Kensek]

RGE - My run was done this morning when my room temp was 50F. Ran at 35 x 127. As I said I use those new low volt low Cas Ram. DDR3-2037 8-8-7-20 1N timings

[JPQY]

New nr1... yes Kensek again -> 9.386.667

It's now after new year and i get almost every few days a new record...between rge and Kensek

JP.

[rge]

I installed windows 7, 64 bit yesterday, build 7000, and like it, so I am done with xp. But will be interesting to see how windows 7 does, when I can get some cold temps to test.

yep, if you can run 2000+ at low timings, I definitely cant do that, at 2000 I was at 9, 9, 9, 27 and had to loosen up some subtimings as well, probably killed benefit of extra bandwidth on mine.

I will compare windows 7 to XP at previous runs just for interest sake, then try for 23x200...but need some cold weather. spi 1m on wind 7 (beta) is only about .05 seconds slower than XP, and that is with xp stripped down, havent seen what services I need on 7 yet...so should be near same.

[Kensek]

I installed windows 7, 64 bit yesterday, build 7000, and like it, so I am done with xp. But will be interesting to see how windows 7 does, when I can get some cold temps to test.

yep, if you can run 2000+ at low timings, I definitely cant do that, at 2000 I was at 9, 9, 9, 27 and had to loosen up some subtimings as well, probably killed benefit of extra bandwidth on mine.

I will compare windows 7 to XP at previous runs just for interest sake, then try for 23x200...but need some cold weather. spi 1m on wind 7 (beta) is only about .05 seconds slower than XP, and that is with xp stripped down, havent seen what services I need on 7 yet...so should be near same.

I tried Win7 last week - 32 bit and it performed worst than XP by quite a margin.

All my benches on Win7 come out poorer than XP, but better than Vista.