I'm sure I remember reading somewhere that nVidia were going to compete against Aegia by making physics software run on SLi rigs (so rather than a dedicated PPU, you had flexible GPU/PPU options). I suppose this sort of supports that, but wouldn't like to have been an early adopter of physics cards if thats the case...
Submitted by babychaos on Tue, 2008-02-05 09:37
Yeah.. both nVidia and ATi looked into running a card as a Physics accelerator. There's even a variant of the popular Havok engine, Havok FX, designed for it... While they can do it, its a bit overkill. The PPU developed by Ageia has 125 million transistors, which sounds like a lot, but a chunk of those will be the memory interface. I would have thought you could integrate it into a GPU without amking the chip mahoosively bigger (for reference, the 8800 GT chip is already 754 million transistors).
I'd hope they'd integrate PPU acceleration in much the same way that transform and lighting were integrated into the original GeForce...
Submitted by byrn on Tue, 2008-02-05 10:30
from what I remember ageia in addition to their hardware physics had a pretty competitive software physics library too that was rivalling havok. It was actually causing them issues since it was in direct competition with their hardware solution. I would guess nvidia will slate that and just go hardware.
Comments
I'm sure I remember reading somewhere that nVidia were going to compete against Aegia by making physics software run on SLi rigs (so rather than a dedicated PPU, you had flexible GPU/PPU options). I suppose this sort of supports that, but wouldn't like to have been an early adopter of physics cards if thats the case...
Yeah.. both nVidia and ATi looked into running a card as a Physics accelerator. There's even a variant of the popular Havok engine, Havok FX, designed for it... While they can do it, its a bit overkill. The PPU developed by Ageia has 125 million transistors, which sounds like a lot, but a chunk of those will be the memory interface. I would have thought you could integrate it into a GPU without amking the chip mahoosively bigger (for reference, the 8800 GT chip is already 754 million transistors).
I'd hope they'd integrate PPU acceleration in much the same way that transform and lighting were integrated into the original GeForce...
from what I remember ageia in addition to their hardware physics had a pretty competitive software physics library too that was rivalling havok. It was actually causing them issues since it was in direct competition with their hardware solution. I would guess nvidia will slate that and just go hardware.