7800GTX fails Shader Model 3.0 test!

[Dashken]

WITH EVERY new product generation, the competition employs a whole marchitecture machine to churn out its powerpointless work for AIBs, disties and selected "members" of the press.

What was interesting an ATI presentation that recently got leaked onto the wibbly web was a claim that Nvidia "lies" about Shader Model 3.0 and that its hardware isn't SM3.0 compliant. The presentation got extra credibility from another well-known web-site. The only problem is that the drivers in question are over 16 months old, and GeForce 6800 Ultra isn't Graphzilla's top dog.

A bold claim indeed. So, we have taken our 7800GTX and ol' trusty 6800 Vulgaris for a spin. We have downloaded drivers from the slide - ForceWare 66.81 - old as a bible, but ATI likes to use these in their slides - current WHQL-certified babies ForceWare 78.01 and Rel80 beta-hoping-to-be-WHQL driver, 81.85. We have used Windows XP Professional in both 32- and 64-bit tastes. For 64-bit, 78.01 and 81.84 were used.

Secondly, we downloaded Display Compatibility Test Kit (short: DCT) in its latest incarnation, 5.3. DCT is a Microsoft in-house tool used for testing integrated and discrete graphics for qualification for WHQL and "Designed for Windows" programmes. The Link for downloaded is at the end of the article, so if you want to check your hardware, be Microsoft's guests.

We have used the ol'trusty theINQ system:

AMD Athlon 64 X2 4800+
Zalman CNPS7000-Cu
1GB Corsair (XMS-3200XL PRO)
GigaByte GA-K8N-SLI-Pro
7800GTX + 6800 Vulgaris, both reference nVIDIA boards
Western Digital Raptor 74GB
MSI DVD burner
Hiper 580W moddable PSU

To avoid any mix-ups and allegations of single-runs, each driver was tested on freshly installed Windows and this was with flat-out, format c: style installations, and then DCT 5.3 was started. The test was configured for full length, and each one took a gruelling 36 hours, and this is one of the reasons why yours truly has been distant for a week.

With every tested driver revision, Nvidia hardware fails to comply with Pixel Shader 3.0 test, and what's odd - Pixel Shader 1.2 and 1.3 are joining the failing club. To add to the confusion - 6800/7800 fully complies with the ATI endorsed 1.4 from the Radeon 8500 days. Each and every tested driver revision acted the same in Pixel Shader 3.0 test. So, it took us 150 hours of testing to confirm the claim behind the powerpointless slide.

But, Captain Canuck isn't a cheerful compliant boy either. For a company that is touting "Shader Model 3.0 Done Right!" spinner talk, it's odd to see a Vertex Shader 3.0 feature (Vertex Texturing) gone from their X1K series.

The best part is, ATI's Xbox 360 GPU dubbed Xenos fully supports the missing feature. Vertex Texturing is even used in Xenos presentations as one of the strengths of that unified architecture. But X1K will have none of it.

Yes, Vertex Texturing is a DirectX 9.0c feature for "Done Right!", so don't flood my inbox with claims it isn't. Link is provided in our complimentary L'INQ section. ATI did offer an alternative, called R2VB (Render To Verex Buffer), and that feature is supported by both Xenos and X1K GPUs.

So, Nvidia does not 100% comply with Pixel Shader test, ATi with Vertex one. Confused? Ask game developers, and please - offer them at least one cup of coffee extra. And an aspirin to boot.

I am currently running a full set of DCT tests on an ATI X1800XL board, to see how the Captain Canuck offering fares. There is no point in running X800GT or GTO I guess. Sadly, we don't have any ATi X1K mainstream hardware... for now.

Source: http://theinquirer.net/?article=27084

[s p r a n k i k o]

Sheesh!!! Another 3Dmark03 situation?

Why can't they all be all just good boys and girls?

It's almost Christmas anyway!!

[kayFX]

give me 7800 GTX and i'll take it.. no shader model 3.0? no problem.. still take it..