Why is NVIDIA against 3DMark03?

[jbirney]

Originally posted by Spiritwalker
Just to refresh your memory, Doom]l[ is an OpenGL game NOT a DX9 game so you cannot make the comparison.

That and the fact that JC has specifically coded the game for the nV30.

There are 5 paths for the Doom3 to render only one of which is NV30 specific. There is one that was coding for the R8500 as well. This path that the 8500 uses the OpenGL version of PS1.4.

[Spiritwalker]

R8500 not 9700

Benchmarks should test spec against spec or vendor specific against vendor specific.

Using the 8500 path the 9700 is faster but uses lower quality textures and rendering (like the nV30 does at points) but ATi hardware is all or nothing, so JC is using the higher quality slightly slower ARB2 path for the 9700. He could code a specific 9700 path (and I dont know why he is not going to), but all I am saying is lets play fair with benchmarking.

Remember the stink we used to hear in the cpu benches when Intel SSE2 code was run against AMD FP? Now most sites state the fact and/or use both versions of the test to highlight the differences. That is all I am asking.

[vitocorleone]

Look, it's tagged as being the "Gamer's Benchmark", NOT the "Gaming Benchmark". There's a very distinct difference. "Gamer's Benchmark" means that people who play games on their computer - hence "gamer's" - use it to benchmark their system, primarily the graphics. It does not imply that the benchmark (no one is disputing that it produces a benchmark score) is meant to represent real games. If it did, then I would agree that the tag was misleading and that it should instead be the "Gaming Benchmark".

[sebazve]

vote 1

[Myrmecophagavir]

Originally posted by Spiritwalker
Benchmarks should test spec against spec or vendor specific against vendor specific.

No, they should test the fastest method the hardware exposes to get the same outcome. If one card only exposes the minimum spec use that, if another card provides extensions to spec which enable it to get the job done quicker then use them. The benchmark tells you which card can get the desired results the fastest, regardless of how it does it.

BTW, PS/VS 2.0+ is still DX9 spec.

[kyleb]

some one aught to send the url for this pole to Jen-Hsun Huang, it would be nice if he would realise he is wasteing time trying to pull strings and if he realy wants to make more of an effort then he should put it into doing good for the future of the nvidia.

[kyleb]

Originally posted by Spiritwalker
Benchmarks should test spec against spec or vendor specific against vendor specific.

ya, and runers should all be alowed to use their own tracks to race on as well!

[tamattack]

Originally posted by Spiritwalker
He could code a specific 9700 path (and I dont know why he is not going to).

Simply because there is no proprietary ATI OpenGL extensions for DX9-level pixel shaders. The optimal codepath for R300 is the same as the OpenGL ARB standard. ATI is all about standard APIs these days.

Whereas JC has to resort to using NV's proprietary extensions just to get competitive performance on NV30. (And there's some debate as to whether the use of NV's proprietary extensions is dropping to FP16 mode or, even worse, integer mode).

[Spiritwalker]

Originally posted by Myrmecophagavir
No, they should test the fastest method the hardware exposes to get the same outcome. If one card only exposes the minimum spec use that, if another card provides extensions to spec which enable it to get the job done quicker then use them. The benchmark tells you which card can get the desired results the fastest, regardless of how it does it.

BTW, PS/VS 2.0+ is still DX9 spec.

I agree with you, but in this case both cards are more than spec so it is up to the drivers to make proper use of the hardware, not the benchmark. The benchmark is written to spec and that is all it claims to be. Some may dissagree because of PS 1.4 but they are DX8.1 spec.

What I am trying to say is test using all available avenues, whether they be vendor specific or spec and compare them. That is the true purpose of a benchmark. But let ppl know which is using what.

[Nutty]

You refer to 3Dm03 as a demo, when it's clearly more than that.

Is it interactive? No, its a rolling demo.

If it were interactive, the character would have to have head collision to stop them jumping up through ceilings. As this doesn't happen in the rolling demo, it's probably omitted.. Along with alot of other items.

[kyleb]

ohh ya well go check out the game nwo, it had some serious collsion problems last i knew and they still called that a game.

[Nutty]

They called it a game, because you could actually play it. Man, am I talking to morons here or what?