Nehalem vs Istanbul

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

Rémi Coulom
Posts: 438
Joined: Mon Apr 24, 2006 8:06 pm

Nehalem vs Istanbul

Post by Rémi Coulom »

I have some money to spend (15-20k Euros), and I wonder which would provide the best performance for chess-like programs on a 4x6-core system, for the same price. (ie, integer computation, with contention on memory accesses and bad scaling, so that hyperthreading does not help).

It is difficult for me to figure out by looking at benchmarks, so I thought I'd just ask here.

Advice about where to buy would be appreciated, too. DELL ? HP ? Other ?

Thanks,

Rémi
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: Nehalem vs Istanbul

Post by bob »

Rémi Coulom wrote:I have some money to spend (15-20k Euros), and I wonder which would provide the best performance for chess-like programs on a 4x6-core system, for the same price. (ie, integer computation, with contention on memory accesses and bad scaling, so that hyperthreading does not help).

It is difficult for me to figure out by looking at benchmarks, so I thought I'd just ask here.

Advice about where to buy would be appreciated, too. DELL ? HP ? Other ?

Thanks,

Rémi
I would not ask. I would run the program you want to use and measure the speed. These processors are getting pretty tricky, and individual programs do things differently, and may well hinder the processor.
Rémi Coulom
Posts: 438
Joined: Mon Apr 24, 2006 8:06 pm

Re: Nehalem vs Istanbul

Post by Rémi Coulom »

bob wrote:I would not ask. I would run the program you want to use and measure the speed. These processors are getting pretty tricky, and individual programs do things differently, and may well hinder the processor.
Yes, of course, that would be the best way to decide. But it may not be so easy to have access to such hardware. I'll ask them, but I don't expect hardware retailers would have such machines available for testing by potential customers.

Thanks for your advice.

Rémi
Vinvin
Posts: 5228
Joined: Thu Mar 09, 2006 9:40 am
Full name: Vincent Lejeune

Re: Nehalem vs Istanbul

Post by Vinvin »

Suj, qui est de temps en temps connecté sur playchess saurait très probablement répondre à ta question, il teste souvent de grosses configurations AMD et Intel.

Si tu y tiens, je peux essayer de te mettre en contact avec lui, écris moi à vincentlejeune@hotmail.com (même adresse sur facebook).

à+
Vincent
Gian-Carlo Pascutto
Posts: 1243
Joined: Sat Dec 13, 2008 7:00 pm

Re: Nehalem vs Istanbul

Post by Gian-Carlo Pascutto »

I think you can work on the assumption that scaling will be very similar, with maybe a slight advantage to Nehalem. Integer performance will be 15-20% faster clock for clock on the Nehalem.

So the question is more or less what specific configurations you've found for your budget.

(PS. That money will buy you 50 quad core cluster nodes :), or 200 cpus)
Rémi Coulom
Posts: 438
Joined: Mon Apr 24, 2006 8:06 pm

Re: Nehalem vs Istanbul

Post by Rémi Coulom »

Gian-Carlo Pascutto wrote:I think you can work on the assumption that scaling will be very similar, with maybe a slight advantage to Nehalem. Integer performance will be 15-20% faster clock for clock on the Nehalem.

So the question is more or less what specific configurations you've found for your budget.

(PS. That money will buy you 50 quad core cluster nodes :), or 200 cpus)
Yes, I know. My boss told me we have to spend that budget before the end of the year, otherwise it would be lost for our research group. I already have access to very big clusters. So I told him if he buys me a trip to Tokyo and a super PC, then I'll try to defend the UEC Cup, and he agreed :-)

Rémi
Rémi Coulom
Posts: 438
Joined: Mon Apr 24, 2006 8:06 pm

Re: Nehalem vs Istanbul

Post by Rémi Coulom »

Vinvin wrote:Suj, qui est de temps en temps connecté sur playchess saurait très probablement répondre à ta question, il teste souvent de grosses configurations AMD et Intel.

Si tu y tiens, je peux essayer de te mettre en contact avec lui, écris moi à vincentlejeune@hotmail.com (même adresse sur facebook).

à+
Vincent
Merci Vincent. Je vais d'abord essayer de contacter des experts localement. Ça ne m'étonnerait pas qu'il y ait des machines de ce genre à l'INRIA. Au cas où, je te recontacterai.

Rémi
tano-urayoan
Posts: 638
Joined: Thu Aug 30, 2007 8:23 pm
Location: San Juan, Puerto Rico

Re: Nehalem vs Istanbul

Post by tano-urayoan »

Hello Monsieuer Colom:

Today the site anandtech published a review between 2 and 4 server sockets between intel, amd. http://it.anandtech.com/IT/showdoc.aspx?i=3653
Rémi Coulom
Posts: 438
Joined: Mon Apr 24, 2006 8:06 pm

Re: Nehalem vs Istanbul

Post by Rémi Coulom »

tano-urayoan wrote:Hello Monsieuer Colom:

Today the site anandtech published a review between 2 and 4 server sockets between intel, amd. http://it.anandtech.com/IT/showdoc.aspx?i=3653
Thanks.

The problem with these reviews is that they are focused on virtualization of database applications. Chess and Go are very different, so it is difficult to figure things out. That's the reason why I posted my question here.

I'll probably settle on that machine:
http://h10010.www1.hp.com/wwpc/us/en/sm ... 81880.html

Rémi
Vinvin
Posts: 5228
Joined: Thu Mar 09, 2006 9:40 am
Full name: Vincent Lejeune

Re: Nehalem vs Istanbul

Post by Vinvin »

Great choice ! I'd go for AMD too ...

No chance to get a 8*6 cores machine with lower clock rate ?
For chess no improvement but as go is better in "parallelization" (because of monte-carlo algo) may be 48*2.4 Ghz is 50%+ faster than 24*2.8 Ghz ...

BTW, I'll follow the go tournament closely !

Zen go is very strong, I wish you success to catch it !

Here's a very interesting link for those who are curious about computer go : http://www.computer-go.info/