Navigate: Advice Beginners BIOS Guide CPUs Links Mainboards Memory Network Storage Video/Sound Cards Contact Forum SiteMap Sponsors WebNews Home	.	.
Prices: Mainboards ABIT ASUS Chaintech Shuttle Soyo Tyan CPU Intel P4 2.4C-800 P4 2.6C-800 P4 2.8C-800 P4 3.0-800 P4 3.2-800 AMD AthlonXP XP 1700+ XP 2000+ XP 2400+ XP 2500+ XP 2700+ XP 3000+ XP 3200+ Athlon64 Athlon64 3200+ Athlon64 FX-51 Opteron Opteron 240 Opteron 242 Opteron 244 Opteron 246 Memory Corsair Crucial Kingston Mushkin OCZ Search Prices:

Some Benchmark Considerations

Before going into the main part of the review, that is system configuration and performance, there are a few things that should be said about benchmarks used. The first point concerns the use of BAPCo's Sysmark2001. Sysmark2001 is a benchmark that runs several applications simultaneously forcing multithreading if it is supported by the CPU. Because the Windows(r) Media Encoder 7.0 encoder in its original form does not properly recognize the specific instruction set of the AMD Athlon XP, AMD in collaboration with Microsoft released a patch to fix this deficiency. Unfortunately, this leads to what might be claimed a violation of the licensing agreement with BAPCo because the benchmark is not running in its originally intended environment. For any reviewer, this causes the problem of getting caught between a rock and a hard place, either way you play it, one party will claim that the use of the benchmark was flawed.

So far so good, personally, a more important issue is whether the benchmark really contributes valuable information. Multithreading is certainly interesting but to be honest, in any PC, there is simply no situation thinkable where a user can input enough data to run several applications at a pace that will result in true multithreading. By extension, the same argument holds for any office application. In light of these issues, I consider office application benchmarks as somewhat irrelevant for the power user and the average person shouldn't be concerned too much anyway because no matter of which CPU can be purchased nowadays, they all have more power than anyone can claim to use in an office environment.

Another issue about benchmarks is that in order to evaluate CPU power, it is necessary to exclude bottlenecks such as I/O latencies caused by hard disk transfer. This by itself disqualifies most current ZDBop benchmarks like the different Winstone suites that are influenced more by the speed of the hard drive used than by real CPU performance. We are interested in performance here and not bottlenecks, it is easy enough to run and add the benchmarks but I feel that they might distract from more important issues. I wish to emphasize that this does not mean that we try to discredit other reviews using other benchmarks, rather, we are trying to use a somewhat different approach and provide complementary information to what has been shown elsewhere.

This, of course makes it somewhat different to come up with meaningful benchmarks, with meaningful specifying applications where a user can actually benefit from higher performance. Whoever expects reiterations of SiSoft Sandra here will be heavily disappointed. This doesn't mean that SiSoft Sandra has no merit, it certainly does but for an evaluation of a CPU and the scrutinizing of the arbitrary QuantiSpeed Architecture algorithms it is not suitable for the simple reason that I have never seen anyone benefit from higher SiSoft numbers Because of my own conception of most standard benchmarks being inadequate, recently, we have started to move to some alternative benchmarks based strictly on real life applications. For 3D games we stick with Quake3 for the time being since it is the one benchmark that favors the Pentium4. There are other benchmarks like Return To Wolfenstein that will be included shortly to give a more rounded picture.

• Quake3 Arena Demo 001
  • At low resolutions, the graphics card becomes a negligible factor. Quake3 Arena makes efficient use of prefetch algorithms and is one of the most favorable benchmarks for the Pentium4 and Rambus memory technology.
• Ulead Media StudioPro6.0
  • The leading video editing software. A 50 MB AVI file is converted into a compressed MPEG format.
• Caligari TrueSpace 4.2 PublicTest D3D and OpenGL
  • A widely used CAD and rendering program heavily optimized for SSE instructions and, thus a home game for the P4. TrueSpace is also heavily influenced by memory bandwidth and latencies. The benchmark was developed by Caligari for Intel with the debut of the SSE instructions.
• Caligari TrueSpace5.1 Raytracing
  • For this review, we have a new addition in the form of Caligari 5.1. The benchmark renders a scene kindly provided by Adam Trachtenberg using raytracing and as such the performance is mostly limited by the raw CPU power, more precisely by the FPU power of the processor whereas the system platform does not appear to matter too much. Raytracing is a beautiful tool and is widely used and what has prevented its entry into the average household are the extremely long rendering times depending on the CPU used. Another advantage for measuring raw CPU performance by means of ray tracing is that it is independent of the graphics adapter used. This means that the rendering times were identical, irrespective of whether a Matrox G200, a TnT2 or a GeForce were used. A demoversion of TrueSpace5.1 can be downloaded free from the Caligari Website and I'll be glad to email the bench file to anyone interested.

Screenshot of the raytracing procedure after about 1/2 of the scene is rendered. The system used in this run was the Tyan Tiger MP which is visible in form of the dual render lines (each CPU renders its own part of the scene independent, very neat)

There are other benchmarks we used for internal control and sanity checks like

• StreamD
  • StreamD is the most commonly used benchmark program for an evaluation of the pure memory bandwidth without additional tasks such as cache transfer or AGP DMA.
• SiSoft Sandra Memory Benchmark (8.21 beta)
  • A composite memory benchmark that targets the entire system data bandwidth including cache transfers and AGP DMA / DiME).
• Expendable Timedemo
  • Execution of a gaming script with rapid alternating reads and writes that is highly dependent on memory and bus latencies as well as FPU performance.
• Incoming
  • With framerates peaking at over 500 fps, Incoming Gameindex is certainly no longer a relevant benchmark for playability, on the other hand as we'll show below, it does very reliably show differences between driver versions to allow selection of the most suitable video driver for a given platform.

    next page:    => Test Platforms: Gigabyte GA-7DX, VIA KT266 Reference Board =>

    All advice and educational articles on LostCircuits are free, but if you feel you can, please make a small donation to us!
    Thank you!

    ---

    General disclaimer: This page only reflects the author's personal opinion and assumes no responsibility whatsoever regarding any of the contents or any damages that may occur explicitly or implicitly from reading the contents of this site. All names and trademarks mentioned in this review are the exclusive property of the respective parent companies.
    All contents of this site are protected by international copyright laws. Reproduction of the contents even in parts is not allowed except after written permission by the author and referral to this site.
    Copyright 2002 - 2008 LostCircuits