Regarding this somewhat alarmist post, NVIDIA were kind enough to contact me and provide some clarification.
After Fermi ships, NVIDIA plan to extend OptiX support to at least GT200 GeForce cards, and possibly down to G80 as well. So eventually you will indeed be able to run Optix on pretty much all consumer NVIDIA cards.
Humus votes for this being the coolest costume ever. Pretty great, I agree, and technically lovely, but I vote for this one since I’ve seen foldy heads before and never saw this idea until today.
We’ve written about the NVIDIA Optix ray-tracing API (which used to be called NVIRT) once or twice before. Well, today it is finally available – for free. While it’s very nice of NVIDIA to make this available, there are a few caveats.
We already knew Optix would only work on NVIDIA hardware (duh), but the system requirements reveal another unwelcome fact; it does not even run on GeForce cards, only Tesla and Quadro (which are significantly more expensive than GeForce despite being based on exactly the same chips). They say GeForce will be supported on their new Fermi architecture – I call shenanigans.