Guru,

NVIDIA GeForce FX is behind ATI RADEON 9500-9800 in every single game or benchmark when pixel shaders are used. Moreover, NVIDIA GeForce FX somehow manages to be slower in games like Splinter Cell.

Our general expectation based on performance in Highly-Anticipated DirectX 9 Game, Tomb Raider: Angel of Darkness as well as some other games and numerous synthetic benchmarks, such as 3DMark03, is that NVIDIA GeForce FX series will be behind comparable models from ATI RADEON 9500-9800 series. We maintain this expectation at this time.

Well I don't disagree with FX being slower than Radeons in most cases, that's pretty much known facts. What I disliked was the indirect statement that FX cards would be total useless for next gen DX9 games which is not the case.

It might not run as fast in pure benchmark tests as the radeons but DX9 games will of course be playable with good fps.

It's the statements below that I think is unprofessional. How about OGL games, heard of DOOM 3?

- Those who are not interested in the next generation games;
- NVIDIA products fans :)

What we see now is too negative for NVIDIA GeForce FX. We cannot base our expectations on pure belief.

We have never tested Doom III game. But we are very excite about it.

That is a typo. Thanks.

Cleeve,

We have never tested Doom III game. But from what John Carmack said in his plan or interview it was clear that the RADEON 9500-9800 is highly likely to be faster than the GeForce FX in ARB mode (which is an apples to apples comparison). We are not sure what NV3x mode for Doom III is, so, let us leave it till later date.

I agree completely ... this is not a professional review and the results should be taken with a serious dose of salts. It's obvious that the guy who wrote this is yet another Ati fanATIc.

Look, if the ATI slacks off then the results are surely misleading and need looking at ... yeah yeah ... It's fairer to say that ATI are better at some things and nVidia are better at other things.

I notice that there are no comments like this when the 5700 scores far higher then the 5950 ... there was even one benchmark when the 5600 owned the 5950 ... ffs , is this a serious comparison ??? I think not.

To absolutely rule out the FX range of cards for all DX9 applications is wrong, stupid and blind.

It is a fact that nVidia have the best driver developers in the world. For pure terms of stability you simply cannot compare ATI to nVidia, thats like comparing a 386 to an AthlonFX 51 ... nVidia have proven that they are improving their drivers in DX9, just compare the 52's and the 53's ... and even a site as reputable as tomshardware indicate that therf is something 'dubious' going on in ATIs cataclysm driver.

How can you mention Doom 3 results when you HAVENT EVEN BLOODY played the game ??? I mean how skewed is that .... do not believe everything you read on a website.

How legitimate is your comparison of your so called Secret DX9 game ????

Ok firstly you make no claims to legitimate benchmarks, so how do we the discerning public know that this isn't the version you downloaded from the newsgroups when it was hacked from valve ... and why have valve seemed to release such an exclusive benchmark to your website, when they haven't released it to the major sites such as toms or anand ... or are you breaching NDA ?

Oh and gee whilst I'm at it ... lets look at the real benchmark issues with HL2. ATI has 24b precision ... nVidia has 32b precision ... Higher is better.
When the drivers are ready, and lets face it they will be (I know you dont agree but dont tar nVidia's programmers with the same brush that you tar ATI's) , this alone means that HL2 will look better on nVidia than ATI ... I really dont care if a 9800 can do an average of 9fps better if it looks worse ... at 60fps who can tell the difference ?

When the HL2 code path for the NV is complete I wouldn't be suprised if the high end nvidias matched or beat performance with the 9800's as they certainly already beat visual quality....
Just a thought here , dont forget that more and more people are moving to TFT screens , and in the majority of cases the Max resolution you can acheive is 1280x1024 ... ergo wtf do a lot of ppl care how the performace is at 1600x1200 ... you have to level the playing field to the level that people will use it ...

I know its hard to do any serious benchmarking , I understand that at the detail you guys went to , it can't be easy ... I for one appreciate greatly how much time this must have taken ... but the answers and your comments show for sure that you are biased towards ATI and that you haven't given 'The Fastest Graphics Cards of 2003' a fair assesment.

That reminds me, my main Graphics card is a Radeon 9800XT.

in addition -

Summing up the results obtained in X2 – The Treat we can see that GeForce FX is much better fit for this game. However, the gap between the GeForce and RADEON families is not dramatic enough to draw a definite conclusion once and for all.

.... If you read your article the WHOLE WAY through ... even though you have cunningly laid out the graphs, the difference between the 9800XT and the 5950 is also not dramatic enought to draw a definate conclusion.
I will defy anyone without a frame graph on screen to tell the difference between 9 fps whem you are above 40fps.


There are only a couple of benchmarks where the ATI completely destroys the nVidia ... and obviously we can mention Serious Sam as a counter to that.

Guys, please be more constructive.

Troj,
The only constructive thing I figure out from your comment is about precision. Permit me to explain.

MS and ARB allow full precision to be 24-bits to 32-bits. There are no difference in image quality between them. NVIDIA wants everyone to use 16-bit precision for Pixel Shaders to get higher speed on its hardware. 16-bit - they call it part precision - is not supported by MS or ARB. All mixed modes for NVIDIA I heard about either used DX8.1, degraded precision to 16-bit, or degraded something else.

So, you do not seem to be right about precision.

I apologise, that was not meant to be totally constructive :), it was meant to prove that (according to you) in one benchmark where the results are TOO similar, then its not a conclusion, but you pass this off as being inconclusive, when you fail to point at that on average there is no noticeable difference between the highend products. As I mentioned, how can you tell the difference betweend 45 and 50 fps when you are playing and CANNOT see the FPS counter... If you tell someone somethings faster they will believe you ... if you dont tell them and ask them to judge themselves they cannot tell thedifference , this is psychology 101.

Sorry , I digress further to mention the precision , this is all quoted from Beyond3D.


With DX9 and its compliant hardware, we now have floating point ("FP") calculation choices available to developers. Note the word "choices". DX9 specifies that for hardware to be DX9 "compliant", the minimum specification for floating point is 24-bit.

...

The entirety of ATI's R300 series pipeline is 32-bit per component, bar the pixel shader processors, which are at 24-bits per component of precision, or 96-bit total.

The NVIDIA's NV30 series internal pixel pipeline provides for three options -- 12-bit integer ("FX"), 16-bit FP per component or 32-bit FP per component.

... Ergo the difference, the ATI Pixel shader deals with colors in 128bits but only has 24bit individual shaders, which limits precision the deeper you get into dx9 programming ... The nVidia has 128bit color, but its shaders can render at 32bit precision and I understand that this is the default, so for every pixel operation the nvidia is doing a byte more work ... slower .. sure , better ? when its used, sure.

You are wrong about partial precision. It is in the DX9 specs/API and is 100% supported by both ATI and NV cards. ATI only has 24bit precision and it will use that when doing partial while NV will use 16bit since it can do 16bit or 32bit(full precision).

Tyler2K

Can you point out more cheating incidents with a major graphics company happened recently in addition to NVIDIA's ultimate cheating episodes?

In case you read our article carefully, you should probably noticed that the obvious problem with the F1 game is that it locks the performance at two points basically.

I don't quite understand your next two claims. Sorry.

Ok , forgetting the Quake - Quak incident with ATI, lets look at Aquamark ... It is shown here at toms hardware

http://www6.tomshardware.com/graphic/20031023/nvidia-nv38- v36-19.html#questionable_optimizations_in_atis_drivers

Understandably ... Toms refuse to say for certain that ATI are optimizing in a method unbecomming a major graphics retailer ... however a simple look at the screen grabs show an Obvious difference ... In fact if you read the entire review on toms from here http://www6.tomshardware.com/graphic/20031023/index.html

and another comparison here ... http://www.guru3d.com/article/Videocards/88/

You might get a possibly fairer assesment.

Again kudos to alexy stepin for a very hard job mostly well done.

ATI and Quake issue happened only once and more than 2 years ago. Nobody is interested in it.

We do not comment on findings of other web-sites, especially in cases when a company explains its positions. In your cases ATI did that.

Ohh . but you comment on the findings of Doom 3 ............ Without having played it ...

We retell you John Carmack's comments, Troj.

Amd thas it so much more valid then reading another websites comments ?

If you read the entire draft then you are only paraphrasing Carmack ... Granted he said the R300 path is faster using defaults, but if you optimize for nVidia, then (again paraphrasing) its a damn site faster again ... I believe he used the word 'noticeable'.

Nevermind the fact that at the time that Jon Carmack said that the NV was slower, the Forceware drivers had NOT been invented at all, and the 5950 simply didn't exist.

Troj,

You can "optimize" Doom III to run fine on the GeForce2 MX by making an ultimate degrade to image quality. What I am pointing out is that when compared apples to apples and there are no specially tweaked modes for particular hardware, but loads of pixel shaders, the GeForce FX is slower than the RADEON.

GeForce FX 5950 Ultra is a faster flavour of the GeForce FX 5900 Ultra - nothing new. ForceWare do their work fine, but not that fine to recommend the card as a solution for future titles.

That is my point exactly though, it was years ago, just as this one with eventually fall out of major news. Yet the point is ATI fixed it and nVidia have fixed most of their problems, my main concern is why the people who won't let it drop. It is in the past, don't hold it against a company who was there in the past and most certainly has a future.

The company continues to play against the rules - it continues to replace shaders in 3DMark03 and may behave dishonestly in other benchmarks.

Have you ever tried to READ something??? I mean you "quantity of BS" is quite staggering. Where the hell you’ve been for the last year and a half? Please be so kind to read some articles about R200, and then some reviews of R300. Maybe then you will bring a comment worth anything except nothing!

IIRC

Doom3 is a game using OpenGL API , BUT only requires DirectX 8.1

except if Id Software add support for PS ver 2.0 hence DirectX 9 class requiring
DirectX 9 or OpenGL 2.0 Hardware...

My advice is this : Wait for Id Soft to finish the game , then decide...



ps : If you are gonna decide based on one game only.....

Peace!