Nehalem benchmarks

Discussion of anything and everything relating to chess playing software and machines.

Moderator: Ras

WuShock
Posts: 182
Joined: Thu Jul 19, 2007 3:13 am

Nehalem benchmarks

Post by WuShock »

Image
Vinvin
Posts: 5320
Joined: Thu Mar 09, 2006 9:40 am
Full name: Vincent Lejeune

Re: Nehalem benchmarks

Post by Vinvin »

With 1 core, the core i7 965 (3.2 Ghz) is 30% faster than the QX9770 (3.2 Ghz too), great achievement by Intel !
ernest
Posts: 2057
Joined: Wed Mar 08, 2006 8:30 pm

Re: Nehalem benchmarks

Post by ernest »

Vinvin wrote:great achievement by Intel !
If this was obtained with HT (hyperthreading) the real value for chess is questionable...
User avatar
Ovyron
Posts: 4562
Joined: Tue Jul 03, 2007 4:30 am

Re: Nehalem benchmarks

Post by Ovyron »

ernest wrote:If this was obtained with HT (hyperthreading) the real value for chess is questionable...
0?
User avatar
hgm
Posts: 28480
Joined: Fri Mar 10, 2006 10:06 am
Location: Amsterdam
Full name: H G Muller

Re: Nehalem benchmarks

Post by hgm »

How many cores has this thing? It seems the benchmark gives the combined performance of all cores, as I don't see any duals amongst equally clocked quads.
Vinvin
Posts: 5320
Joined: Thu Mar 09, 2006 9:40 am
Full name: Vincent Lejeune

Re: Nehalem benchmarks

Post by Vinvin »

ernest wrote:
Vinvin wrote:great achievement by Intel !
If this was obtained with HT (hyperthreading) the real value for chess is questionable...
Hmm , may be i made a mistake ... I though it was Fritz 11 but it's more probably Deep fritz 11 !!

Sorry for my mistake !
Erik Roggenburg

Re: Nehalem benchmarks

Post by Erik Roggenburg »

hgm wrote:How many cores has this thing? It seems the benchmark gives the combined performance of all cores, as I don't see any duals amongst equally clocked quads.
the i7 family discussed each have 4 cores. only the top extreme version has
an unlocked multiplier.

as you might guess, these things are pricey. the lowest model of the three is expected to be out at somewhere around $300 USD.

for most people that do not poop gold, upgrading to a i7 920 from a 9550 doesn't make much sense. you'll need a new mobo - and those things (for now) are as expensive (if not more expensive) than the i7 920. if, however, you are struggling along with something slow, and it has been a couple of years since your last upgrade, it might be worth it. it is a lot of money to spend for a 12% increase (i7 920 v. 9550, which is the closest in price, IIRC).
Karmazen & Oliver
Posts: 374
Joined: Sat Mar 10, 2007 12:34 am

Re: Nehalem benchmarks

Post by Karmazen & Oliver »

Ovyron wrote:
ernest wrote:If this was obtained with HT (hyperthreading) the real value for chess is questionable...
0?
NO. HT is not similar real cores, but is better that nothing... except in games... in other software is good. it´s a patch, but intel = patch. = sale.

it is better for cache... if you have a lot of aplications runs in cpu.
User avatar
M ANSARI
Posts: 3734
Joined: Thu Mar 16, 2006 7:10 pm

Re: Nehalem benchmarks

Post by M ANSARI »

Nehalem will only be interesting to upgrade once it goes to 8 cores or is available in Xeon form with a dual socket Skulltrail type motherboard. Although by quickly looking at the chart it looks like the Corei7 cleans the clocks of a Q6600 ... you have to realize that most people that are running Q6600's have them clocked at 3.2 Ghz to 3.8 Ghz. I am not sure how the Corei7's are clocking in their non multiplier unlocked forms ... but with only 4 cores on the Corei7 the Q6600 are definetely still good.
User avatar
hgm
Posts: 28480
Joined: Fri Mar 10, 2006 10:06 am
Location: Amsterdam
Full name: H G Muller

Re: Nehalem benchmarks

Post by hgm »

The best machine I have is 'just' an E6600, and I have been postponing buying a new one already for quite some time in the hope I could have an octal core. I as severly tempted last week to buy an HP box of very nifty design, based on the Q9300 (a Core 2 Quad from the Penryn family). It tured out the Q9300 has only 6MB L2-cache, though (3MB per core pair), while in princple Penryns should have 6MB per core pair. And anything below 4MB would make it unsuitable to run the EGTB generator I am developing.

I guess I can still wait; I just started writing an engine that is suitable for implementation of SMP (on my exiting engines that would be quite hopeless), abnd before I am at the stage where it even runs on a single core will take some time.