Intel V8

Biffa · 11 Jun 2007 at 16:59

We have been working behind the scenes with some of the Folding @ Home crew and I managed to get a hold of an internal beta benchmarking application which enables us to run a script on each machine and leave it for 48 hours to ascertain CPU performance. The Folding@Home project uses SSE on intel processors, 3dnow! on AMD processors and Altivec on PowerPCs.

From http://www.driverheaven.net/reviews/intelV8/folding.php

FreeStream · 11 Jun 2007 at 17:27

Running normal Workunits I see.

my C2D @ 3.4 gives me 3.6-4k Points every 48 hours running SMP.

Quads average about 5/6k points every 48 hours.

So 8 cores and other optimisations should see 10k :eek:

oceaness · 11 Jun 2007 at 18:28

Concorde Rules said:
Running normal Workunits I see.

my C2D @ 3.4 gives me 3.6-4k Points every 48 hours running SMP.

Quads average about 5/6k points every 48 hours.

So 8 cores and other optimisations should see 10k

I was thinking that that looks a bit low for 8 cores

A.N.Other · 11 Jun 2007 at 22:01

Holy Moley!

If that were running SMP clients, well ... no more need to be said.

rich99million · 11 Jun 2007 at 23:34

That looks rather nice it has to be said :eek:

Mattus · 11 Jun 2007 at 23:40

Pretty much as expected really - a system that basically has two QX6800s is twice as fast as a QX6800

shadowscotland · 12 Jun 2007 at 09:33

It's the X2 5600 wooping the E6700 that gets me, overclock anyone

oceaness · 12 Jun 2007 at 11:52

shadowscotland said:
It's the X2 5600 wooping the E6700 that gets me, overclock anyone

The 5000 is only 15 points below the E6700 :eek:

Mattus · 12 Jun 2007 at 13:08

If this is the benchmark I think it is, it runs an Amber WU, a Tinker WU and a Gromacs WU and averages the scores. Ambers and Tinkers run very well on AMD.. they're also now very rare, so the bench isn't really very accurate. C2D is where it's at.

shadowscotland · 12 Jun 2007 at 14:19

Mattus said:
Ambers and Tinkers run very well on AMD..

Had this problem with benchmarks that were going to be used with the stalled comp. data many months ago (were working to a 20/80 ratio at the time - more like 5/95 now)