Originally posted by: chizow
That's a fact, but it still doesn't translate into pure performance.
Of course shaders are not end-all thing in performance. Like I said, there are other things besides it.
Yes, a GF3/4 will require multiple passes to render the same scene, but if those multiple passes are just as fast as a single pass, there isn't any real-world benefit. Efficiency for efficiency's sake means nothing if it doesn't translate into performance. Its no secret that the GF series cards use fragment programs to accomplish the same work in multiple passes that a R200/R300 would in a single pass, but the end result is negligible when it comes to image quality, and actual performance/"efficiency" is self-evident in REAL benchmarks. The same can NOT be said for 3dmark2k3.
LOL! people whine that 03 is not a "real" benchmark, yet they happily used 01 to benchmark their vidcard, NV included! WHy is 01 a "real" benchmark (after all, why would NV use it?), whereas 03 is not? Because it supports standard-features that NV is too lazy to implement?
FWIW: I have seen 03-benchmarks with R9700Pro using PS2.0, 1.4 and 1.1, and the difference in performance is quite striking.
PS 1.4 is a DX 8.1 compliant feature, so please stop confusing it with DX 9.
I'm well aware that PS1.4 is a DX8.1 feature, thank you very much
Yes, DX 9 falls back to 1.4 natively, but it also falls back to 1.1 if 1.4 isn't supported. I think that's the rub here. All the supporters of 3dmark2k3 keep saying its a true test of DX9 hardware, but in reality 95% of the tests are composed of DX 8.1 features.
Like I said, PS2.0 is used where it's smart thing to do. And that's how games are made. There propably wont be any games that are 100% PS2.0.
That's the issue here, FutureMarketing and ATI obviously collaborated on the project to make it a test of DX 8.1 or better cards (they say as much in their responses to nVidia's remarks). What shifted from a focus on a DX9 benchmark has shifted to a DX 8.1 or better benchmark.
And the problem is.... What? Like I said, all DX9 (PS2.0) cards support DX8.1 (PS1.4), so using PS1.4 is valid thing to do.
The irony is that a Radeon 8500 gets higher |3SMarks than a Ti4600, yet there isn't a single game that an 8500 (even with the more "efficient" PS 1.4 rendering methods) would outperform a Ti4600.
In games that take advantage of PS1.4, they would be differences (assuming 8500 wouldn't be otherwise bottlenecked).
Doom 3 included. In fact Carmack has said numerous times that a GF3 requiring multiple passes still outperforms a R200 that only requires a single pass. I still feel that 3DMark2k3 is either optimized to render using PS 1.4, or it arbitrarily assigns a penalty to non-PS 1.4 parts.
No, it doesn't arbitarily penalize 1.1 cars, other than requiring them to do more passes to achieve same results.
Another common defense for 3DMark2k3. If its just a "vid-card test" or a "3D capabilities test", why do they claim to be "The Gamer's Benchmark".
Now you whine about their slogans
? Why don't you whine NV "the way it's meant to be played" (with DX7!)-slogan?
Why are references to Doom 3 all over their white papers?
D3 uses PS1.4, just like 03 does.
They claim that GT 2 and 3 were specifically designed to "mimic" the rendering engine of Doom 3, yet both fail miserably.
Yes, it mimicks D3 in that it uses PS1.4. But why do you say it "fails"? Because NV loses? oh no, we can't have that, so the benchmark is obviously flawed! Yes, that must be it.
Oh wait, I guess that's what they do anyways LoL!
The most glaring shortcoming is still the fact that CPU/Platform differences have almost no effect on the total score. Good luck explaining to some poor fellow that gets 4500 in 3DMark2k3 with his Celeron 1.1 and 9700pro why he can't run Doom 3 smoothly (as his score would indicate).
How do you know that D3 will be CPU-limited and not vid-card limited?
You said it yourself, there are no games that are limited by either.
Not yet. 3DMark is a forward-looking benchmark.
PS 2.0 is only used in certain instances "where its not overkill". I don't think a game will extensively use either of these features until DX10 at the earliest, and by then this benchmark will be even more useless. Not to mention any card that currently has full DX 9 support will be obsolete or a value part.
Like I said, 3DMark is forward-looking benchmark. And there are already games coming up that take advantage of advanced shaders.
Again, confusing a DX 8.1 feature with DX 9.
No I'm not, I'm well aware that PS1.4 is DX8.1 feature. It's a feature that will see wide use even with DX9, it's officially part of DX-standard (but of course it shouldn't be used in benchmarks 'cause NV is too lazy to implement it
) and it's supported by all vid-cards that support DX9. So I REALLY fail to see the problem here.
Just as all DX 9 parts support 1.4, DX 9 games will support PS 1.1 (with obviously better results than those portrayed by 3DMark2k1).
How do you know? PS1.4 is ALOT better than 1.1 is!
I find it funny when people use the "inferior tech" argument; I'm sure you're one of those people that rushes out and buys a card b/c of some great feature, only to realize there isn't a single game within 6 months to a year that will implement it.
You are sure eh? You are wrong! I currently have GF2 GTS.
And yes, NV pushes inferior tech. Carmack has said that MX doesn't cut it with D3, Ati's entire lineup is DX9 (with exception to 9100). NV has fallen behind, that is a fact.
When it finally shows up in a game, the card is obsolete and the next generation of cards is already upon us.
And they can still use 3DMark03 to benchmark those cards.
Personally, (and most reviews echo the same sentiment) I'd rather see implementations of current features over emphasis on future capabilities.
3DMark has always been future-looking. And ALOT of reviewers use it, so there doesn't seem to be a problem
The best example is the R200, a completely overambitious part that suffered miserably b/c it banked on "future tech".
Really? I think it's a pretty good product. First it competed succesfully against GF3, then it competed against GF4.
As for 3DMark2K3, its still a broken benchmark.
Yeah, because NVIDIA says so
They might as well slap a "Built By ATI" sticker on it and call it "The ATI Video Card Benchmark".
What do you suggest they do? Cripple the benchmark so it fitst better with NV's crippled feature-set
?
Its certainly not "The Gamer's Benchmark", and does nothing to mirror real-world gaming performance, in future or current games.
You have a crystal-ball that tells you all of this? You know more about game-developers and their intentions that Futuremark does?
Head over to beyond3d-forums. There are ALOT of smart people there (many are in one way or another involved in 3D-business). They don't see a problem with 3DMark03, neither did NVIDIA, untill Futuremark decided to support advanced tech that they were too lazy to implement. And, again, that really is NV's problem.
What you are basically suggesting that every benchmark should be designed around NV-hardware, otherwise the benchmark is inaccurate.
Facebook
Twitter