Futuremark releases 3DMark03 patch - alleges NVIDIA cheats

[Madstrom]

Sorry if this is a repost, but here's FutureMark's take on the nVidia/ATI difference. This came with my email and had the Q&A at the bottom.



Q: Do scores change with the patch?
A: Yes. Although the actual workload in the benchmark is not changed at all,
we have noticed substantial performance score changes in NVIDIA hardware and insignificant performance change e.g. with ATI hardware. Thus, only build 330 scores should be used from now onwards.

[dante_uk]

Really, what's the point?

You look at the benchmarks for the following:
Q3 Engine ( at least a dozen licenced games and more coming )
Unreal Engine ( Lots of licenced games and more coming )
DooM 3 ( Doom3 , Quake4, unknown title by HumanHead Studies)
3dMark2003 ( none, nothing, petty tech demo )

Compare top Nvidia card with top ATI card.
Currently:
Nvidia wins in Quake3, Unreal, Doom3
ATI wins in 3dMark2003

Okay now which card do I buy ?

How can those results be fair and true?
None of those game engines are DirectX9, but then only one game test in 3dMArk2003 is actually using any DirectX9.

If the top ATI is better than the top Nvidia card why don't ANY other benchmarks show that?

What's Cheating ?
If there's not code to make these game engines run faster in the drivers from both companies then why the hell not - We all play games using these engines!!
I for one want the best performance and best images I can get.

3dMark2003 claims to be about apples-to-apples comparisons. What's the point?
I Write a DirectX function to spin a cube. It works, it shows me a FPS score. One card runs it faster than another.
Is the faster card going to play quake 3 better and faster then the other card? Would anyone be willing to bet money on which card plays unreal 2 better and faster based on how my spinning cube runs?

Then you have the fact that 3dMark is only DirectX.
Most of the games I play use the quake 3 engine ( opengl ) plus I play IL2 using the OpenGL renderer because it's faster, I play UT2003 using the OpenGL renderer because it's faster.
What should I use to see who has the best OpenGL support on their cards?
Answer: easy, I run Q3 and UT2003 !!

Nvidia is only guilty of wasting time bothering to get a better score with 3dMark2003.
Anyone stupid enough to put their faith in the performance of this program deserves what they get.

[digitalwanderer]

Originally posted by dante_uk
Compare top Nvidia card with top ATI card.
Currently:
Nvidia wins in Quake3, Unreal, Doom3
ATI wins in 3dMark2003

Yeah, nVidia wins if you leave AA & AF off...if you turn any of those on the ATi card toasts 'em. I don't know about you, but if I'm paying HALF-A-GRAND for a a video card I'm kind of expecting to use AA & AF.

If the top ATI is better than the top Nvidia card why don't ANY other benchmarks show that?

Again, turn on AA & AF and see the difference both in image quality & speed...nVidia just can't compare right now.

What's Cheating ?

nVidia, I thought we had covered that one already.

3dMark2003 claims to be about apples-to-apples comparisons. What's the point?

A fair and honest head-to-head comparison of a cards features and performance, some of us like that.

[legion88]

Originally posted by dante_uk
Really, what's the point?

You look at the benchmarks for the following:
Q3 Engine ( at least a dozen licenced games and more coming )
Unreal Engine ( Lots of licenced games and more coming )
DooM 3 ( Doom3 , Quake4, unknown title by HumanHead Studies)
3dMark2003 ( none, nothing, petty tech demo )

Compare top Nvidia card with top ATI card.
Currently:
Nvidia wins in Quake3, Unreal, Doom3
ATI wins in 3dMark2003

Okay now which card do I buy ?

How can those results be fair and true?
None of those game engines are DirectX9, but then only one game test in 3dMArk2003 is actually using any DirectX9.

If the top ATI is better than the top Nvidia card why don't ANY other benchmarks show that?

What's Cheating ?
If there's not code to make these game engines run faster in the drivers from both companies then why the hell not - We all play games using these engines!!
I for one want the best performance and best images I can get.

3dMark2003 claims to be about apples-to-apples comparisons. What's the point?
I Write a DirectX function to spin a cube. It works, it shows me a FPS score. One card runs it faster than another.
Is the faster card going to play quake 3 better and faster then the other card? Would anyone be willing to bet money on which card plays unreal 2 better and faster based on how my spinning cube runs?

Then you have the fact that 3dMark is only DirectX.
Most of the games I play use the quake 3 engine ( opengl ) plus I play IL2 using the OpenGL renderer because it's faster, I play UT2003 using the OpenGL renderer because it's faster.
What should I use to see who has the best OpenGL support on their cards?
Answer: easy, I run Q3 and UT2003 !!

Nvidia is only guilty of wasting time bothering to get a better score with 3dMark2003.
Anyone stupid enough to put their faith in the performance of this program deserves what they get.

What a weak commentary.

I'll keep it simple so try to keep up.

Performing well in Quake 3 benchmarks does not translate to performing well in DirectX games. Duh! So why does it matter that 3DMark is "only DirectX"? Duh! Quake 3 is only OpenGL. Duh, again. Using logic goes a long way. Try it.


As I pointed out already, "opti-cheating" to improve static benchmarks (like timedemos in Quake 3 or "synthetic" benchmarks like 3DMark2003) does not translate to improved performance in real-time usage. Duh! So all "opti-cheating" does is mislead the public into believing that their real-time frame rates would go up alot when it, in reality, it won't.