New dynamic branching demo

[Scali]

Effectively, Nvidia has elected to take a speed hit on current/near future games (compared to ATI), while gambling that they can get that loss back using SM3.0 over the next year or so.

Funny that you mention that, since ATi did exactly the same with R9700.
Instead of having optimized integer pipelines, like the FX, they chose to have FP pipes only, and emulate integer operations on it, at the cost of performance.
For ATi it worked out, because SM1.x games were 'fast enough', while SM2.0 ran great.
And we all know what became of the FX.
Ironically the tables are turned now. I think NV's choice is the right one (as far as I've seen, NV's SM2.0 performance is not that much less than ATi's, and SM3.0 is already giving them gains in FarCry in some cases). And I think that ATi will not wait as long with a good SM3.0 card as NV did with a good SM2.0 card (I consider the 6800 their first good SM2.0 card).

 

[g__day]

Funny but I see both NVidia's and ATi's approaches as correct and culturally valid for these companies. The current generation of video cards is barely the ones that will usher in SM 3.0, it will be the next two generations that deliver the goods, kinda like the Geforce 1 and T&L.

NVidia has its SM 3.0 platform and now has to dial up the speed for its next two generations. ATi has raw speed in SM 2.0 advantage and I guess a decent platform to achieve SM 3.0 with real speed in its next generation of GPU. So neither lose out - its like AMD saying we made it to 1 GHz first - an important milestone but not the end of the race.

Well NVidia got to SM 3.0 first, but that's just the start of the race really. ATi has always said no new features until they are fast enough - NVidia trials a generation or two before the real deal that packs performance is delivered.

I can live with either approach - knowing that is what these companies actually do!