Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : 3DFX -- Ignore unavailable to you. Want to Upgrade?

To: Marc who wrote (11102)	3/13/1999 3:04:00 PM
From: Tae Spam Kim	Respond to of 16960

March 13, 1999, early AM ------------------------ Okay, surprise plan update before I start going into grind ahead mode for my GDC speech. We ran some benchmarks tonight testing out some of our newer optimizations for Quake3, and decided to compare the top four accelerators that we consider relevant: Voodoo2 (for obvious reasons), ATI Rage128, and the NVidia RivaTNT and RivaTNT2. We tested 640x480x16bpp very thoroughly, since this was the lowest common denominator among the accelerators, however we also made a few tests at 1600x1200x16bpp to test fill rate and 640x480x32bpp to test likely operating conditions. Tests were run on a best case machine, a P3/500. 640x480x16bpp (16-bit textures): RivaTNT2 - 50.7 Rage128 - 48.1 RivaTNT - 48.0 Voodoo2 - 44.7 (Ed: HOLY SHIT THAT'S GOOD!) 640x480x32bpp: Rage128 - 47.0 (16-bit textures) RivaTNT2 - 43.7 (32-bit textures) RivaTNT - untested Voodoo2 - N/A 1600x1200x16bpp (16-bit textures): RivaTNT2 - 18.1 Rage128 - 15.1 RivaTNT - 11.7 Voodoo2 - N/A Comments: I'm really really really REALLY amazed and giddy at how fast the Voodoo2 performed. John and I were absolutely dumbfounded at the performance we saw, which was just phenomenal given that it's fairly old by accelerator standards. Major kudos to 3Dfx's OpenGL team for improving their driver as much as they have recently. I wish we had managed to test the Voodoo3 we have at 1600x1200, since it should have performed phenomenally well because of its strong fill rate. Unfortunately I "upgraded" the BIOS on my V3 and it's not feeling very well right now. But if extrapolation proves correct, the V3 may actually prove to be the fastest board of the lot even at 640x480, since it's 38% faster than the V2 on a PII/300 and thus could be that much faster on the PIII/500. Once my board is healthy again I'll put up the numbers. You actually don't notice much of a visual difference between 1600x1200 and 640x480 believe it or not. I think 640x480 is sort of the sweet spot still, but I could see gamers bumping up to 1024x768 when this resolution essentially becomes "free". I plan on doing a very comprehensive set of benchmarks (hopefully with Permedia3 and Savage4) after Q3TEST has been released. Finally, props go out to the OpenGL guys at ATI and NVidia (especially NVidia, which has shown a strong commitment to OpenGL for a very long time and who are one of the pivotal figures in making OpenGL a viable option for developers and players alike). NVidia has always had a very fast (and just as important -- robust) OpenGL driver, and their hardware has always been top notch. The Rage128 is giving TNT and TNT2 a run for their money, but we won't have anything definitive until actual production parts and drivers are available. Note: the demo that this was tested on was an actual recorded demo of a very very violent game of deathmatch with about 7 players in a restricted space, so this should be actually be more aggressive in terms of performance requirements than a real deathmatch would be in actual game play. What this means is that Q3 should have higher average frame rates on modern systems than Q2 did a year ago, which is a good thing. In addition, Q3 is far more scalable than Q2 was, and should run on Rage Pro and Permedia2 fairly well (it will be ugly, but it should still be fast and playable). Wow, I'm in a much better mood now.