Both nVIDIA and Diamond told us that the TNT2 Ultra is capable of yet even greater clock speeds than the ones that achieved the incredibly high scores with our board. A bit more experimentation will be done no doubt before and indeed after TNT2 Ultra's ships so it could be a case of 'the best is yet to come'. Either way, we think you'll agree with us, when we say that the Viper V770 is out-and-out champion when it comes to visual quality and yet maintaining frame rates that are close to that of the Voodoo3 and in some cases higher (in D3D). Although the Voodoo3 outperforms the TNT2 by a couple of frames per second at high resolutions (1600x1200 and 1024x678) in Quake2 Massive1, the likelihood of this small gap being reversed is a distinct possibility. We could give you a rather boring maths lesson etc.. But let's just say that the Viper V770's Ultra TNT2 graphics clock still has a bit of mileage left in its tank before it fills out at 183MHz. If indeed the Viper V770 was to go up to (or above) 183MHz and nudge clear of the Voodoo3, then the end result would not be a welcome one for 3Dfx.

The greater attainable clock speeds are not only possible due to memory vendors now being able to run reliably as high as 200MHz (for SDRAM) but also because of the shrink in the die down to .25micron, which nVIDIA successfully achieved. No mean achievement…

Although Glen Fry (from Steely Dan?) sang that wretched 'The Heat is On' song for Beverly Hills Cop in the 80s, the heat is certainly OFF in the Ultra TNT2s case. Set at the high clock speeds mentioned above, the Diamond Viper V550 (which did harbor a fan) ran unbelievably cool. After 12 hours of solid Turok 2 T2-marking (we left it running all night), it ran flawlessly and wasn't anywhere near as hot as you might expect (Voodoo3 gets much hotter). NVIDIA certainly deserves a lot of praise for this as our fingers (manual heat probes!) were far from being scalded when pulling the board in and out after usage. We'll update you with some results from our Electronic Heat probe if the desire is there. Again this whole reliability issue at high clock speeds comes down to nVIDIA's successful die shrink from .35micron for the TNT down to .25micron for TNT2. Sharing the same FAB as 3Dfx has its advantages as TSMC has of course given 3Dfx a successful die-shrink from .35micron down to .25micron as well.