How fast are new computer systems with chess!

[Albert Silver]

I don't really understand the test at all. The results are all showing different depths or iterations. What exactly does the test consist of?

[bob]

Those numbers look horrible to me. I get a higher NPS than that on a quad-core Intel. On a dual quad-core I am seeing numbers between 16M to 20M so 7M is not exciting at all...

[bob]

Hi ernest,

Nehalem is still a quad core cpu,so 2 x 4 =8cores + with HT on = 2 x8threads
so the chess engines see it as a 16cpu !

Nehalem 10% slower (it's 7,3418%) but his speed is 36,81% lower then the Skulltrail @4.2Ghz!!

JP.

Turn hyper-threading OFF. It is not a good thing for most chess programs. the parallel search introduces more overhead than the extra threads can account for, so you run _slower_.

[jpqy]

Hi Albert,

This is very easy to understand..i let people with heavy computer systems just run a chess engine from the begin position (most off them don't know anything about chess) If i have to use a test position it will be too complicated...i do this already many years,so i know what problems i get from people who don't understand how to do this test.
I say,let your computer run 5-10min. and give me your screen and/or analyses.
The most computers in my list can make in less then 5min. reach Depth 21...so all nodes/s. are taken at the moment the first depth 21 is calculated.Gives me a very nice average from the nodes/s.

Only now 2 people have not reach depth 21..so i take 20..but it gives me very clear what system is faster or slower!!

This i always do before i buy a new computer systems...many years ago i wait when a computer get 10 times faster..a few years ago it's just 2 times faster...i had a AMD 3400+ always testing...keep a test position who need 30min. to find the solution...till the core 2 cpu's came!! I buy a E6600 the same test was done in 3min. Wow,long time ago that it goes 10 times faster...from dual core to quad..buy a Q6600 g0 @3.6ghz...again 2times faster...Now comes the Nehalem and he wil give me at least 2Miljoen nodes/s. more when he overclock well!!

For me is it just a other way then using pi,prime,ortal,sp2004,cinebench,...
to check your speed and how stable your computer is.I overclock my computer till he is Chess stable!

The person with the first score or highest nodes/s. till now with his Skulltrail system at 4.2Ghz has only the best computer parts you can have till today! (tomorrow is different) He has a really fast build up..look to the second Skulltrail @4.0GHz,is already much slower and still good 1Ghz higher clocked then the Nehalem..a normal Nehalem @2.93Ghz is almost so fast as this Skulltrail...i know what my next cpu will be!

Regards,
Jean-Paul.

[jpqy]

Hi Robert,

Please do the test for me,give me your screen and analyses!!

You use Arena + TogaII142JD multi core!!

And show me 16M nodes/s. !!

That would be shocking news for me

hyper-threading "ON" give on the same computer (Nehalem @2.93Ghz) with the same programs -> 1.3Milj. nodes/s. more!!

So i think you get a nice speed gain with HT "ON"

Regards,
Jean-Paul.

[bob]

Hi Robert,

Please do the test for me,give me your screen and analyses!!

You use Arena + TogaII142JD multi core!!

And show me 16M nodes/s. !!

That would be shocking news for me

hyper-threading "ON" give on the same computer (Nehalem @2.93Ghz) with the same programs -> 1.3Milj. nodes/s. more!!

So i think you get a nice speed gain with HT "ON"

Regards,
Jean-Paul.

I don't run toga or any other program, I run Crafty. On a dual quad-core, 16-20M is what I get, depending on the position. Be happy to run one for you if you want. I might even have something...

time=5:18 mat=0 n=5347594847 fh=91% nps=16.8M
time=28.82 mat=-6 n=489084409 fh=88% nps=17.0M
time=4:25 mat=0 n=4371820456 fh=89% nps=16.4M
time=3:02 mat=-1 n=3281143514 fh=90% nps=18.0M
time=57:52 mat=0 n=72832255513 fh=42% nps=21.0M

Those are from various positions, run on a dual xeon (E5345) 2.33Ghz, 8 cores total. No hyper-threading.

[jpqy]

Hi Robert,

I have PM you...i would like to send you the files with explenation and would be glad to put your result in my list with your system!

Kind regards,
Jean-Paul.

[Albert Silver]

I ran the test. Rig is E8400 @ 3.6GHz (can OC much more, but this is the day-to-day clock as it doesn't require voltage changes or cause temp increases) and DDR2 PC8500:

FEN: rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1

TogaII142JD-2cpu:

18/38 00:26 35.837.695 2.793.600 +0.21 e2e4 e7e5 Ng1f3 Nb8c6 Bf1b5 a7a6 Bb5xc6 d7xc6 Nb1c3 Bc8g4 h2h3 Bg4xf3 Qd1xf3 Qd8f6 Qf3g4 Ra8d8 00 Bf8c5
19/43 00:51 71.577.071 2.801.921 +0.24 e2e4 e7e5 Ng1f3 Nb8c6 Bf1b5 a7a6 Bb5xc6 d7xc6 00 Bc8g4 h2h3 Bg4xf3 Qd1xf3 Qd8f6 Qf3e3 Bf8d6 c2c4 Qf6e6 d2d3
20/48 01:49 152.016.606 2.830.533 +0.20 e2e4 e7e5 Ng1f3 Ng8f6 Nf3xe5 d7d6 Ne5f3 Nf6xe4 d2d3 Ne4f6 d3d4 Nb8c6 Nb1c3 d6d5 Bf1d3 Bf8d6 00 00 Bc1g5 Bc8e6 Rf1e1
20/48 02:23 202.910.528 2.839.278 +0.23 e2e4 e7e5 Ng1f3 Ng8f6 Nf3xe5 d7d6 Ne5f3 Nf6xe4 Bf1d3 d6d5 Nb1c3 f7f5 Nf3e5 Qd8d6 Bd3xe4 d5xe4 d2d4 Nb8d7 Ne5c4 Qd6g6
21/49 06:16 536.903.119 2.854.673 +0.31 e2e4 e7e6 d2d4 d7d5 e4xd5 e6xd5 Bf1b5+ Bc8d7 Qd1e2+ Bf8e7 Nb1c3 Ng8f6 Ng1f3 00 00 Be7b4 Nf3e5 a7a6 Bb5d3 Nb8c6 Ne5xc6 Bd7xc6 Bc1f4 Nf6e4 Nc3xe4 d5xe4

Albert

P.S. I e-mailed the screenshot in case this is a requirement.

[M ANSARI]

Let me know what test you need. I have changed my Skulltrail setup from Vapor to air. For quite some time I had it running at 4.8 Ghz but have decided to put it in a box where I can use it without the noise. It is now at 4 Ghz on air but I should have it on water by next week which is probably 4.2 Ghz. I also have a Skulltrail running air with 2 x E5430's at 3.2 Ghz doing some engine matches which should finish in a two days. I am surprised at the results you have with Skulltrail using E5430's at 3Ghz as compared with Skulltrail at 3Ghz using different QXxx. They should be almost identical but results show a big drop for the E5430's. On my system a QX9775 at default 3.2Ghz gives same result as E5430 at 3.2 Ghz. The limiting factor for the E5430's is the memory ... FBDIMMS are notoriously hard to overclock and cause a crash which is difficult to recover from on the motherboard ... which is why I avoid anything above 400 Mhz for 400 Mhz ram. If faster FBDIMMS are available I would think the E5430's would be good for at least 3.6 Ghz as they run at 3.2 Ghz for weeks on end 24/7 on the lowest Skulltrail voltage setting.

[jpqy]

Hi Albert,

Thanks for your result...if i have your memory timings then your result is complete!

Maybe it's better i put my list also on this forum...

JP