Computer Chess Club Archives


Search

Terms

Messages

Subject: Helps if people knew how to benchmark..

Author: Aaron Gordon

Date: 23:08:55 12/31/02

Go up one level in this thread


On December 31, 2002 at 21:11:10, John Rice wrote:

>On December 31, 2002 at 17:22:42, Rick Terry wrote:
>
>>A friend of mine who works at USA Computers seems convinced that The Pentium 4
>>512 FSB outperforms the Athlon in Every Bench Mark. I didn't know enough to
>>argue with him, repeating only what I heard here about AMD Processors being
>>Superior to the Pentium in Running Chess Programs, But Why?, Why would the
>>Pentium Perform better then the Athlon in all other Benchmarks but Chess?
>
>That's an easy one, FPU's (floating-point unit). AMD CPU'S are much better at
>dealing with integer mathematics. Compare the Sandra FPU benchmark results
>between a P4 and the Athlon XP and you will see what I mean.
>JR

You're another unfortunate victim of Sisoft's BS. If you notice you'll see
Sisoft compares the Athlons RAW (non-SIMD) FPU vs the P4's SSE2 SIMD fpu. This
is like taking two ferraris, each having turbo and nitrous and turning the turbo
off and not using nitrous on one of the cars. Completely unfair. Try comparing
the raw FPU (listed in the sisoft test) vs the RAW Athlon FPU.

If you compare the chips equally (both non-SIMD, since sandra doesn't even have
an SIMD test for the Athlon) you'll see the P4 gets smoked beyond belief.

AthlonXP @ 2.4GHz: 3670 Mflops
Pentium4 @ 2.8GHz: 1571 Mflops

AthlonXP is 133% faster even running 400MHz less than the P4.

Even with the Athlon as far ahead as it is do I trust Sisofts benchmarks? Not
hardly. Their benchmarks are the epitomy of festering retardation. Version after
version do they bother to include an SIMD test for the Athlon or AthlonXP? Nope.
Makes me wonder what other stuff they could have 'left out' of benchmarks behind
the scenes (in the code itself, as far as Athlon optimizations go).


As far as other benchmarks.. most review sites haven't a clue how to test
properly (as in setting CAS-2, enabling 4-way interleave, etc in the bios on
Athlons). Others know how to but hinder the results on purpose (Tomshardware,
HardOCP and most recently Anandtech, have proof in all 3 cases). Now, most
people by looking at all the benchmark reviews would think Quake3 runs better on
the P4. Again, not true. Quake3 is extremely memory bandwidth dependant. As
I mentioned previously most review pages run horrible ram timings in the bios.
This causes a massive reduction in memory performance and, of course, frames per
second in Quake3.

I compiled a list of some Quake3 results that are purely cpu/fsb/mem related.
The videocard makes next to no difference. Most review sites do stupid things
like run 1600x1200x32 in cpu reviews. Why is this stupid? It makes the videocard
a massive factor in the framerates, running higher resolutions increases the
dependency on the fillrate of the videocard.

Here are the results and instructions on how to run the benchmark yourself.
http://speedycpu.dyndns.org/bench/q3viafps/

As you can see a P4-2133MHz/133(533 RDRAM) gets 42.7fps. This is actually faster
than a P4-2.4GHz/100fsb(400 RDRAM) in Quake3. With this in mind look where the
AthlonXP 1900+ (1.6GHz, stock speed) is with the Epox 8KHA+. 43.8fps.

This trend holds true to most applications, games, etc. available today. If the
Athlon is setup properly it's going to be VERY fast. How might you set one up?
Well, first you need good ram and a good chipset. For SDR systems the KT133a is
best (not KT133). For DDR systems definitely go for a KT266a (not KT266), KT333,
KT400 or Nforce2 (best at the moment). With those chipsets in mind look at this
bios optimization guide. It'll tell you most everything you need to know.

http://www.rojakpot.com/bog.aspx




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.