Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Anandtech's Chess Benchmarks shows that Pentium is better than AMD

Author: Aaron Gordon

Date: 16:01:20 12/31/04

Go up one level in this thread


On December 31, 2004 at 18:33:01, Eran Karu wrote:

>Please click below.
>
>http://www.anandtech.com/linux/showdoc.aspx?i=2213&p=4
>
>In the graph, you can see that Pentium runs more nodes per second than AMD does.
>So, while many people say AMD is better than Pentium for chess, why does the
>graph show the opposite?
>
>Eran Karu

Because Anand, like many other review pages, pick and choose what programs
they'd like to favor for Intel. If you compare Divx encoders, Chess programs,
etc. Anything that is optimized for the P4s and doesn't do well on the Athlons,
they use. I emailed Anandtech multiple times about using Crafty, he did it once
in a Xeon EMT (64bit) vs Opteron test. The P4 got killed, as usual, so he
decided to go back to TSCP.. which so far favors P4 chips more so than Athlons.
The only chess program that I've seen that does it.


Lots of people will say I'm crazy, why would they do that, etc. I've already
seen it before (Tomshardware faking pictures, benchmarks, Anandtech and HardOCP
doing the same). Been like this for quite some time.

Kick backs from Intel are just one of the reasons they do this. Do you really
think some kids that started a benchmarking page are going to give Intel bad
reviews and stop Intel sending them $800 worth of chips, $500 of ram, $200
motherboards every other month?

One of the reasons they don't use DVD2AVI in the encoding tests. Athlon beats
the P4 by quite a bit. Of course Intel doesn't want you to know that.. and the
review doesn't want to stop getting free stuff. So you'll basically never see
the Athlon optimized stuff, ever. Only recently has HardOCP started to use
DVD2avi (and the P4 gets absolutely annihilated). They do the same thing with
MP3 encoders, encryption, you name it. Even OpenSSL (pick and choose tests).

Read this to see one of the MANY things they do:
http://www.vanshardware.com/articles/2001/august/010814_Intel_SysMark/010814_Intel_SysMark.htm

and
http://www.vanshardware.com/reviews/2002/08/020822_AthlonXP2600/SYSmark%202002%20Analysis%20Presentation%20FINAL.pdf

In short, what Intel did was take the 1 filter the P4 did well at, did it over
and over.. making almost the entire score based off that 1 single test. Just
imagine someone taking Crafty, Shredder, Chess Tiger, TSCP, and Fritz. Testing
them all and showing the Athlon is faster overall. Is this a good test? Yep.
Now, lets do it like this.. TSCP, TSCP, TSCP, TSCP, Crafty. Way to go Intel.

Basically don't bother trusting any review pages. Either test it yourself or ask
someone that has both systems that is competent. I've had both a P4 (2.4 @
3.2GHz) and an Athlon XP (1.83 @ 2.6GHz), and have tested various Athlon and P4
systems over the years. I would use my Athlon ANY day over my P4, and I do,
because the P4 is just plain slow. Some things it is MUCH slower (over twice as
slow). Yes, it is configured properly. Infact, the memory latency, timings AND
memory speed is faster on the P4 (I use PC3500 overclocked, 2-2-3-6 timings) vs
3-4-4-11 timings on my PC2700 that is in the Athlon.



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.