Sandy Bridge - Bulldozer - Computer Chess Test!

gainestown still under NDA.

Well, mine are that is... :wasntme:

Core I7 920 @ 3.8ghz
Gigabyte UD5
Corsair 3Gb DDR3-1600Mhz 9-9-9-24
Windows Vista 64bit




http://img143.imageshack.us/img143/4564/cheeskn6.jpg




:up:

Hi guys,

I tested a little bit more :)

http://img222.imageshack.us/img222/2469/chesshe2.th.jpghttp://img222.imageshack.us/images/thpix.gif

Monstru & Kensek...Thanks for your results:up:

List updated!

will be joining once given further instruction.. jpqy, you got a pm with my information :)

Sagatanas & Jcool...Thanks for your results:up:

List updated!

Hi Jean Paul,

I did it. This program is also good for instability test.

http://ssyybg.blu.livefilestore.com/...6YGTTYQ/t1.JPG

Thanks,

Metroid & Flopper...Thanks for your results:up:

Metroid shows a perfect screen!

List updated.

Guys, This is awesome!
Let's put some explaination behind the amazing speed of nehalem.
I did a little quick vtune on it, just to be sure ...
So, you get this amazing speed because the code, as you should expect, is VERY branchy, it does branch like no tomorrow. Hyperthreading get some space to be efficent here. On the top of this, the L1 instruction and L1 data are very happy ... almost perfect hit, the prefetchers are doing wonder too to keep the all stack happy. The L3 to memory is very successfull too
As you would expect from a chess game, the memory pattern access is "almost predictable", but the multiple hierachy of caches help filtering the unpredictable... each of them, because the height of each cache line, remove a part of the unpredictable sequence ...

Hyperthreading and Bandwidth/Shorter latency help a lot here.

Francois

Hio DrWho?,
I was absolutely stunned when I ran this bench. I did no tweaks at all. Just ran it box stock with default BIOS and even default 1333 mem speed. I wasn't even running my RAM at it's advertised spec. This platform is a real beast. I would not want to "play the computer" against a Bloomfield...I'd get owned! :ROTF:

This thing is so far out ahead in technology that we don't even have the proper benchmarks for it. This is the only one I've seen that takes advanatge of it's architechture, and instructions.

We will be seeing games and programs that will take advantage of it more and more, and I hope the Dev's and Intel both work closely with each other to fully utilize the power this thing has. This is a memorable platform. It really is the best I have ever seen.

i had another run today
i7 965 @ 4.142ghz
rampage 2 extreme
gskill 6gb 1600 9-9-9-24
http://alexiel.stor.li/x58/chess2.jpg

Yeah well, my guess is that the Bloomfield's IMC is mostly responsable for this gain in speed here.
This chess test might not be representative though, as any other bench/app out there that isn't completely memory-dependant shows my Dual Harpertown at 4Ghz to be 40-50% faster than my 4Ghz i7 (i.e. Cinebench, BOINC, 3DMark CPU etc).

Sargatanas, nice cpu you got there :D Hope mine will clock similarly well!

ty jcool.
should check the lowest vcore...should be around ~1,37v
but since i had no time for that i just took 1,4v +llc ^^

My test.:up:

Processor: Intel Core i7 920@4104Mhz "216x19" - 1.23v;

Motherboard: Gigabyte GA-EX58-UD5 - Bios F4K;

Ram: Trial Channel by Corsair 6GB "3x2GB" - DDR3 1600 Cas 8-8-8-24 - 1.65v "TR3X6G1600C8D"@1728Mhz - Cas. 7-8-7-22 - 1T - 1.62v;

Heatsink: Air by Asus Triton 81;

System: Windows Wista 64 bit - Service Pack 1.

http://img254.imageshack.us/img254/2...3795nx0.th.jpg

http://img504.imageshack.us/img504/5...2707xy1.th.jpg

Very good Program!:clap:

Hi Jean,

It looks like I kept my promise,

http://ssyybg.blu.livefilestore.com/...142JD-8CPU.JPG

T_Flight...Nilos & Metroid....Thanks for your results:up:

Metroid...yes you get the 8Milj.nodes/sec.:clap:

dj883u2...thanks for your result...close all other programs in the background and you will get much higher nodes/sec.!!
Your system can do much more;)

JP.

Hi Jean,

A better score right here. I have been working hard to get this result, always improving :)

This result below is the best I could get by the fact even the Realtemp shows how exactly the temperatures of the 8 logical cores were when the test was being processed.

Load time 100% Toga 8 cores:
Minimum: 67c
Maximum: 74c
Average Maximum and Minimum: 564/8 = 70.5c

This program is magnificent. I will use it altogether with many others for a research that I will be conducting in few days about core management.

http://ssyybg.blu.livefilestore.com/...142JD-8CPU.jpg

Metroid [08.12.2008] [21.46] [8.066] [I7 920 4.0Ghz 1.27V 21x191 P6T Deluxe 1531Mhz 8.8.8.24 Win XP] Arena+TogaII142JD-8CPU.txt

Thanks,

Did another. This time I added the extra CPU-z with the Memory Tab.

This is at 3780Mhz. This will be the first OC'ing Run for me in 5 years! :)

Forgot to Add: This is on Intel Box Stock Cooling. Temps were about 41-43C idle and 68-73C Full Load 8 Cores.

Thanks T_Flight...and that for first OC in 5 years:clap:

Guys...if possible,don't forget the analyses data!!
It's much easier to copy your results in the list..else i have to do it manually from your screens:yepp:

JP.

PS. List updated.

Yes, I was looking for that. I couldn't find it. How do we get that text data?

9. Go to menu above the chessboard and click on 'Position' and select 'Copy Analyses to Clipboard'
10.Now you can easy paste these analyses in a e-mail or text file and send it to me (also with the screen, if possible!)

Or how Metroid did it with Attached Files.

JP.

Got it. Will do next time. It's gonna be a bit before I get all my water stuff. I do have some of it here, but have to wait until after Christmas, and maybe the first of the year before I can get the rest. After that, I feel like this CPU may have 4.2 in it on water for benching. I'll have to see. It's very stable at fairly low volts. I can actually bench at 1.2v instead of the 1.23v I had in that last screen. I was just testing to see if it was any more stablke and might improve, but it doesn't look like the extra volts are adding anything to those clock levels. I'm gonan take it back down to 1.2 for now. You may have also noticed that my mem is running at a 2:12 ratio which is pretty dang high for the clock level. I may be able to increase mem clocks and back off the ratio. Still playing with stuff.

This benchmark is very useful for showing multithreaded performance. I think this could very be THE standard for direct measure of actual full load multithreaded performance. You can see it now in the scores. There is a direct correlation to RAW CPU Power and nodes/s.

http://i291.photobucket.com/albums/l...ghar/arena.jpg
here r my results .. e6750@ 3.6ghz with 2gb rams so set the BOTTOM LINE :D no one dare to cross me :P

It's liking it. I can't wait to get my watercooling and bump it up to 4GHz or 4.2 possibly. This is fun! :yepp:

At last could manage to work it out,even though not so awesome i'm just happy can make it work for now. and will do some benchie again soon.