*If* I were to gain temporary access to a 3Dlabs Wildcat....

[Dave Baumann]

... what tests would you like to see?

I know that the Super-Scene AA on these boards has interested a few people in the past, not least me, so I've asked 3Dlabs if it would be possible to borrow one for a little while. As it stands it may not happen, but if it did what would you guys like to see tested?

Bear in mind that I have no Workstation software to hand, nor am I likely to be able to gain access to it. However, if anyone wants to supply little testing utlilities then that would be useful.

So, what do you think?

 

[Reverend]

Definitely not 3D games. I'm pretty sure it won't perform well comparatively and I doubt 3DLabs would appreciate it all that much if this is the case (i.e. you use games and it doesn't do well).

BTW Wavey, for WS benching software, try Indy3D. Back in 2000 (or was it 1999?), I was contacted by someone at www.sense8.com (some guy called Erik Garrison... dunno if he's still there...) after he was told of my Quadro review. He sent me a CD containing Indy3D. Indy3D gives you fillrate, fixedrate, polygon rate & MCAD benchies plus etcs.... not sure it this has changed in the latest Indy3D iteration though. Hell, I still have the result running this on the Quadro somewhere...

 

[Dave Baumann]

Although I probably will take benchmarks I wouldn't be looking at this from an out-and-out performance angle, but more about what image quality enhancements can be gained from an adaptive-multisampling solution such as this, which I why I will look at some games (assuming they work).

Given that ATi are moving in this direction with supersampling (or at least, were supposed to!), nVIDIA are gradually enhancing their multi-sampling solution, and 3Dlabs are entering themselves later in the year, then its quite likely that we may see AA solutions moving more towards this type of functionality which is why I would look at some games.

Anyway, thanks for the heads up on Indy3D - I'll look into that.

 

[pascal]

Great Dave 8)

Any image quality (Some JPGs and BMPs) test with a real game (if it works) that we can reproduce at home and have a better feeling of the image quality.

Performance is not the case.

 

[Livecoma]

Perhaps some type of article explaining different feature considerations for professional accelerators over gaming accelerators? I'd be interested in learning more about were a professional featureset diverges from a consumer one. Guess it depends on what crowd your targetting the article too.

The AA capabilities also sound interesting. I wouldn't think benchmarks and numbers would be nessesary for a nice comparison...

 

[Reverend]

Well then that's cool although I can almost guarantee you that you'll find it hard to find differences between its technique and that of out-and-out MS/SS.

Which Wildcat model would that possibly be?

 

[Dave Baumann]

Naturally I'd shoot for the best! But I don't know yet, whatever 3Dlabs let me have access to - assuming they allow any.

However, AFAIK, all the Wildcats are capable of 16 sample AA so I would expect to see some differences - especially with some carefully prepared applications (any coders out there... ).

 

[mat]

Some (uncompressed) screens with this 16xFSAA would be great.

BTW: the german magazine c't has a Wildcat III vs FireGL 8800 vs. Quadro4 vs GF3 Ti500.
i havent read all of it, but i saw that the P4 performance is much better then the Athlon MP with these cards.

 

[Dave Baumann]

I will be using a P4 2.0GHz Northwood at least.

 

[Reverend]

I wasn't talking about differences in number of samples... I was talking about equal number of samples.

Anyway, off to bed! Hehe.

 

[3dcgi]

If you decide you want professional software to test you can try the Softimage demo, which is free. The Softimage demo doesn't use a custom file format so if anyone knows of any good scenes you could try them.

Maya Personal Learning Edition wouldn't work because the graphics performance is slower than the real version so results might not be accurate. Also the custom file format would make it hard to find good scenes to test with.

You might not need these programs if the main thing you want to test is antialiasing.

 

[Sharkfood]

I think PowerVR's dev demo "Tubes" would be good for AA scrutinizing- at least for edge AA. Should also be fairly forgiving on driver support and not requiring rocket science there.

If it's got a decent OGL ICD, even the original Tribes with some 3rd party maps would be good for AA. This game picked lots of wretchful "graph paper" floor patterns and some of the 3rd party maps have geometry alot like "Tubes"- with jungle-gym style structures and joe-generic OGL compatibility for any Win OS.

 

[Ailuros]

OT dumb question: wouldn't theoretically adaptive 16xMS equal roughly in output with 8xSSAA?

I recall the Z3 algorithm whitepaper claiming something along that line, or did I get that wrong?

 

[Gunhead]

Actually, even for what it isn't worth, I'd like to see how it does 3DMark2001's polygon tests, just out of general curiosity. If it doesn't shine there, you can always add a link to 3DLabs' own quality discussion (with the yellow-spotted submarine) to explain things and appease them. And of course you could then run something sufficiently heavy (Indy3D?) comparing the Wildcat to a gamer card, for politeness' sake.

(I know it's not intended for anything like 3DMark2001. But I'd like to *see* that so the differences become very easy to grasp.)

 

[Dave Baumann]

Its been confirmed - I'm going down to their UK office on Monday to have a board installed...

 

[jb]

yea,

lets see some AA screen shots of games if you have the time. I am not worried about FPS numbers...just wanna see what real FSAA looks like

 

[pascal]

...just wanna see what real FSAA looks like

Yeah, I wana see it too

 

[Bambers]

Hmm the old det2 and 3 drivers used to support 4x4 (16x) SSAA. It was only orded grid though and the memory requirements were so great that my 32MB gf256 SDR could only run it at 512x348x16 max . Still Motocross madness 2 ran smooth with it on in that res

 

[Tagrineth]

Actually, even for what it isn't worth, I'd like to see how it does 3DMark2001's polygon tests, just out of general curiosity. If it doesn't shine there, you can always add a link to 3DLabs' own quality discussion (with the yellow-spotted submarine) to explain things and appease them. And of course you could then run something sufficiently heavy (Indy3D?) comparing the Wildcat to a gamer card, for politeness' sake.

(I know it's not intended for anything like 3DMark2001. But I'd like to *see* that so the differences become very easy to grasp.)

The Wildcat III line only supports DX7.

Wildcat II and lower don't support DirectX at all.

They're professional OpenGL cards only.

 

[Dave Baumann]

They said they had some new drivers recently done, so they may now support other versions of DX. I specified that I was currently running XP, which only ships with DX8 so if they have XP drivers (which I assume they do) then I'd guess they must have DX8 support.