Anandtech review of S3 DeltaChrome

[vb]

http://www.anandtech.com/video/showdoc.html?i=1989

Edit: Sorry, http://www.anandtech.com/printarticle.html?i=1989

We have chosen not to do a full image quality analysis at this point becuase we feel it would be a waste of time.

Edit2:

they support what ATI and NVIDIA are calling 4x rotated grid antialiasing, and upto 16x anisotropic filtering.
Currently, S3 calls thier antialiasing algorithm 2xAA, but under 2x mode, the pixel is sampled twice both horizontally and vertically (which is technically correct, but doesn't fit the established standard). It is likely that the nominclature will end up changing at some point. The maximum resolution at which they support AA is 1024x768, and, eventhough the functionality isn't exposed, it would be possible to do 16 sample AA (what S3 would call 4xAA) at lower resolutions. Obviously this is not going to be fast, and it is unlikely that this functionality will be exposed in the shipping version of the product.

Is it possible to use rotated grid AA when doing SuperSampling? Or does S3 support Multisampling?

Rather strange...

 

[jb]

Is it possible to use rotated grid AA when doing SuperSampling?

Say hello to my Voodoo5 which showed us back then how good AA looks when its using a RG using SS methods

 

[tEd]

nm

 

[I.S.T.]

Here's another review: http://www.techreport.com/reviews/2004q1/deltachrome-s8nitro/index.x?pg=1

 

[Fodder]

If they got every game running as well as Unreal Tourney 2003-4, they'd be set.

 

[tazdevl]

Kind of funny the 5950 Ultra is absent from the benches.

 

[StealthHawk]

We have chosen not to do a full image quality analysis at this point becuase we feel it would be a waste of time.

Edit2:

Well, it is a preview, so I can excuse not looking at IQ.

they support what ATI and NVIDIA are calling 4x rotated grid antialiasing, and upto 16x anisotropic filtering.
Currently, S3 calls thier antialiasing algorithm 2xAA, but under 2x mode, the pixel is sampled twice both horizontally and vertically (which is technically correct, but doesn't fit the established standard). It is likely that the nominclature will end up changing at some point. The maximum resolution at which they support AA is 1024x768, and, eventhough the functionality isn't exposed, it would be possible to do 16 sample AA (what S3 would call 4xAA) at lower resolutions. Obviously this is not going to be fast, and it is unlikely that this functionality will be exposed in the shipping version of the product.

Is it possible to use rotated grid AA when doing SuperSampling? Or does S3 support Multisampling?

Rather strange...

Someone needs to tell Anand that NVIDIA doesn't support 4x rotated grid AA.


More previews! Is this card ever going to see the light of day?

Does anyone know if Volari is shipping in the US yet?

edit: Victory for Futuremark! From the TR preview:

The S3 driver revision we used was version 6.14.10.1632-15.08.09.b. We used ATI's CATALYST 4.2 drivers on the Radeon card and Forceware 56.56 on the GeForce FX 5700 Ultra. One exception: at the request of FutureMark, we used NVIDIA's 52.16 drivers for all 3DMark benchmarking and image quality tests.

edit2: oops. TR compared ATI, NVIDIA, and S3 in the Pixel Shader 2.0 test of 3dmark03.

 

[ChrisRay]

Well This card might be interesting assuming they get the driver issues worked out.

I want a new toy without the ATI or Nvidia logo

 

[Tagrineth]

edit: Victory for Futuremark! From the TR preview:

The S3 driver revision we used was version 6.14.10.1632-15.08.09.b. We used ATI's CATALYST 4.2 drivers on the Radeon card and Forceware 56.56 on the GeForce FX 5700 Ultra. One exception: at the request of FutureMark, we used NVIDIA's 52.16 drivers for all 3DMark benchmarking and image quality tests.

edit2: oops. TR compared ATI, NVIDIA, and S3 in the Pixel Shader 2.0 test of 3dmark03.

No victory at all.

They only used Forceware 52.16 for 3DMark03... not for any games.

That just skews the results even farther...

 

[3dilettante]

edit: Victory for Futuremark! From the TR preview:

The S3 driver revision we used was version 6.14.10.1632-15.08.09.b. We used ATI's CATALYST 4.2 drivers on the Radeon card and Forceware 56.56 on the GeForce FX 5700 Ultra. One exception: at the request of FutureMark, we used NVIDIA's 52.16 drivers for all 3DMark benchmarking and image quality tests.

edit2: oops. TR compared ATI, NVIDIA, and S3 in the Pixel Shader 2.0 test of 3dmark03.

No victory at all.

They only used Forceware 52.16 for 3DMark03... not for any games.

That just skews the results even farther...

If they used the approved drivers for 3dmark, then it is a victory for futuremark, all that they're going to get anyway. It's unreasonable to stick to the old revision for all the games just because3dmark's EULA requires it for that one program. There are valid fixes in the drivers amongst the optimizations.

 

[Hanners]

More previews! Is this card ever going to see the light of day?

Well, according to The Tech Report's preview, it looks like cards should ship in Europe in the next few weeks.

 

[vb]

No victory at all.

They only used Forceware 52.16 for 3DMark03... not for any games.

That just skews the results even farther...

3DMark03 340 patch only removes cheats in 52.16 regarding 3DMark03, not any other application.

 

[vnet]

If they got every game running as well as Unreal Tourney 2003-4, they'd be set.

Optimization for that particular game engine.

 

[StealthHawk]

No victory at all.

They only used Forceware 52.16 for 3DMark03... not for any games.

That just skews the results even farther...

Not sure what you mean by that. Surely you don't expect reviewers to use 52.16 on their whole test suite and not use the latest official driver, which may or may not have perfectly valid optimizations for games?

 

[oddfellow]

Well, according to The Tech Report's preview, it looks like cards should ship in Europe in the next few weeks.

I actually might pick one up out of curiosity. Depending on price of course

 

[Tagrineth]

Not sure what you mean by that. Surely you don't expect reviewers to use 52.16 on their whole test suite and not use the latest official driver, which may or may not have perfectly valid optimizations for games?

Which is why there really isn't any victory at all.

The optimisations are still everywhere.

What's the use in having a cheat-free 3DMark03 when everything else is hacked to death? And a cut up 3DMark is useless in and of itself.

 

[StealthHawk]

Which is why there really isn't any victory at all.

The optimisations are still everywhere.

Well, I did quantify that it was a "victory for Futuremark" and not a "victory for truth, justice, and 'optimization'-free benchmarking." You seem to be agreeing with me and haven't contradicted that.

Considering all the threads that popped up here in the last several weeks and whatnot about reviewers using 53.03 in 3dmark03 it is a victory for Futuremark that review sites are playing by their rules.

What's the use in having a cheat-free 3DMark03 when everything else is hacked to death? And a cut up 3DMark is useless in and of itself.

No arguments there. Better than nothing though, right? And who is to say that everything else is indeed "hacked to death" anyway. Furthermore, what recourse is there? Just give up benchmarking entirely because of the possibility that the results are tainted? If review sites use 3dmark03 at all, then they should use approved drivers.

 

[Quitch]

edit: Victory for Futuremark! From the TR preview:

The S3 driver revision we used was version 6.14.10.1632-15.08.09.b. We used ATI's CATALYST 4.2 drivers on the Radeon card and Forceware 56.56 on the GeForce FX 5700 Ultra. One exception: at the request of FutureMark, we used NVIDIA's 52.16 drivers for all 3DMark benchmarking and image quality tests.

edit2: oops. TR compared ATI, NVIDIA, and S3 in the Pixel Shader 2.0 test of 3dmark03.

No victory at all.

They only used Forceware 52.16 for 3DMark03... not for any games.

That just skews the results even farther...

FutureMark have no one to blame but themselves for that. They manage to contradict their own approved driver rules, and think a note in the readme will do fine.

I can't believe they haven't corrected this issue yet. We were told that time restraints meant they hadn't fixed this as they wanted to nail the game issues first. Since then... nothing.

 

[StealthHawk]

FutureMark have no one to blame but themselves for that. They manage to contradict their own approved driver rules, and think a note in the readme will do fine.

I can't believe they haven't corrected this issue yet. We were told that time restraints meant they hadn't fixed this as they wanted to nail the game issues first. Since then... nothing.

Didn't worm say that there most likely will not be another patch in one of these threads?

 

[Quitch]

Precisely, which would suggest time issues are not the only reason the PS 2.0 test wasn't fixed. If no fix is ever coming out, then it's hardly time that's the factor.