3dMark 2003 Performance Issues

Omega65Omega65 Philadelphia, Pa
edited August 2003 in Science & Tech
Aces Hardware : 3dMark 2003 Performance Issues

"While doing some research on an old 350 MHz Pentium II system, Johan came across some interesting results while benchmarking 3DMark03. When equipped with a high-end video card, like a Radeon 9700 Pro, the 350 MHz Pentium II was able to more than keep pace with a 1.4 GHz Pentium III Celeron and even a 2.8 GHz Pentium 4 in several of the 3DMark03 game tests.

Perhaps this should not come as a suprise, since 3DMark03 is, after all, primarily a video card benchmark. But how much of one? Given the sheer magnitude of difference between the 350 MHz Pentium II and more modern hardware, we decided to investigate further to determine whether or not these 3DMark03 results would be mirrored in real-world gaming situations. "

Final Thoughts

3DMark03 is characterized as follows on Futuremark's site:

By combining full DirectX®9.0a support with completely new tests and graphics, 3DMark03 Pro continues the legacy of being the industry standard benchmark. The high quality game tests, image quality tests, sound tests and others give you an extremely accurate overview of your system’s current gaming performance.

With the exception of GT1 - Wings of Fury, the 3DMark03 game test performance is almost completely opposite to the game benchmarks presented here with regards to CPU usage. In fact, in all of our testing, we were not able to find a single case where the 350 MHz Pentium II performed on par with the 1.4 GHz Celeron, or even within the same 20-25% range seen in 3DMark03's Battle of Proxycon and Troll's Lair tests. And it certainly never came remotely close to outperforming the 2.8 GHz Pentium 4, despite doing so in three out of four 3DMark game tests.

GT1 - Wings of Fury
350 MHz PII - R9700 - 24.4

1.4 GHz Cel. - R9700 - 80

2.8 GHz P4 - R9600 - 125.9

1.4 GHz Cel. - R8500 - 58.4

GT2 - Battle of Proxycon
350 MHz PII - R9700 - 25.9

1.4 GHz Cel. - R9700 - 30.9

2.8 GHz P4 - R9600 - 20.9

1.4 GHz Cel. - R8500 - 06.2

GT3 - Troll's Lair
350 MHz PII - R9700 - 27.7

1.4 GHz Cel. - R9700 - 28.4

2.8 GHz P4 - R9600 - 19.4

1.4 GHz Cel. - R8500 - 06.5

GT4 - Mother Nature
350 MHz PII - R9700 - 30.3

1.4 GHz Cel. - R9700 - 30.7

2.8 GHz P4 - R9600 - 21.6

1.4 GHz Cel. - R8500 - NA

===================================
The Final analysis - 3DMark 2003 is a Videocard Benchmark only! :thumbsdow :thumbsdow :thumbsdow

Comments

  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited July 2003
    This only confirms what we all have suspected for a while. Good article. Futuremark will take a huge blow now. Even worse than before.
  • Geeky1Geeky1 University of the Pacific (Stockton, CA, USA)
    edited July 2003
    Oookay... maybe I'm missing something, but I fail to see how this is an issue. If anything, it seems to me it's a good thing. Why? Because not everyone has identical systems, and if 3DMark 03 were a more system-oriented benchmark, then the scores of various cards in different test systems wouldn't be compareable. However, since it's obviously a video card oriented benchmark, it should be much easier to compare cards that were not reviewed in the same test system.
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited July 2003
    Geeky,

    Do you have any idea what the program cost on a corporate level? THOUSANDS of dollars believe it or not. If Futuremark claims this program/utility can be used as a benchmark program for the SYSTEM, not the videocard only, companys that have bought (a lot of them has) it, will be furious.

    Not only that, the amount of money Futuremark earns on advertising on their site is unbelievable and the amount of peeps that visits that site very often is massive.
  • Geeky1Geeky1 University of the Pacific (Stockton, CA, USA)
    edited July 2003
    In terms of corporate licensing, no, I don't know how much it is, but now that you mention it, considering the price of other professional level software packages (e.g. Pro Engineer), thousands of dollars sounds about right.

    Admittedly, I haven't read the article yet, but has Futuremark ever advertised 3DMark as a valid benchmarking tool for measuring overall system performance?
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited July 2003

    3DMark03 is characterized as follows on Futuremark's site:

    By combining full DirectX®9.0a support with completely new tests and graphics, 3DMark03 Pro continues the legacy of being the industry standard benchmark. The high quality game tests, image quality tests, sound tests and others give you an extremely accurate overview of your system’s current gaming performance.

    [/B]

    Geeky1 said
    In terms of corporate licensing, no, I don't know how much it is, but now that you mention it, considering the price of other professional level software packages (e.g. Pro Engineer), thousands of dollars sounds about right.

    Admittedly, I haven't read the article yet, but has Futuremark ever advertised 3DMark as a valid benchmarking tool for measuring overall system performance?

    That´s what they do.
  • Geeky1Geeky1 University of the Pacific (Stockton, CA, USA)
    edited July 2003
    Ah... somehow I missed that in Omega's post... :rolleyes:

    In that case, then yes, I have to agree with you- the software is totally incapable of doing what they claim it does...
  • EnverexEnverex Worcester, UK Icrontian
    edited July 2003
    Actually, it does exactly what it says it does. When it finishes, it gives TWO results, the one in big letters is the Graphics score, the SECOND result is the rest of the system, the score which seems to be missing from your Post.

    Main Test Results
    3DMark Score 4501 3DMarks
    CPU Score 401.0 CPUMarks

    Detailed Test Results

    Game Tests
    GT1 - Wings of Fury 139.2 fps
    GT2 - Battle of Proxycon 31.1 fps
    GT3 - Troll's Lair 27.6 fps
    GT4 - Mother Nature 26.8 fps

    CPU Tests
    CPU Test 1 42.7 fps
    CPU Test 2 7.5 fps

    Note the CPU score, which would obviously test the FSB and Processor Sub Systems.

    NS
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited July 2003
    Nightshade, isn´t those only available on the Pro version?

    If so, how much did the proggy cost? I have never looked for the price actually. Only seen the corporate prices.
  • Omega65Omega65 Philadelphia, Pa
    edited July 2003
    NightShade737 said
    Actually, it does exactly what it says it does. When it finishes, it gives TWO results, the one in big letters is the Graphics score, the SECOND result is the rest of the system, the score which seems to be missing from your Post.

    Main Test Results
    3DMark Score 4501 3DMarks
    CPU Score 401.0 CPUMarks

    NS

    Who ever posts their 3DMark CPU score??

    And why use 3DMark to tell you what you already know?? ie a R9800 > R9700 >R9500 > R9600 etc.

    What's funny is that even on a P350 a 9700 scores higher than a R9600
  • WuGgaRoOWuGgaRoO Not in the shower Icrontian
    edited July 2003
    how about '02 and '01 versions...what do they test??
  • SpinnerSpinner Birmingham, UK
    edited August 2003
    WuGgaRoO said
    how about '02 and '01 versions...what do they test??

    Well put it this way, I've tested two very different systems in terms of CPU speed and BUS speed etc etc, one was a XP2000+ and the other was a Thunderbird 800, if I recall. Both systems however had a GF4Ti4400, and they were both identical cards, made by Creative. There was over a 3500 point drop in 3DMARK2001SE score for the slower system. So at least in my experience, previous versions of 3DMARK do seem to bench the system more generally for gaming than 3DMARK2003, at least if that article is as true as it is convincing.

    Cheers
Sign In or Register to comment.