thefixer opened this issue on Jun 16, 2009 ยท 161 posts
svdl posted Fri, 19 June 2009 at 5:37 PM
moogal: I concur. In fact, rendering on the GPU instead of on the CPU would be a HUGE plus. Even with a fairly modest graphics card you could get near realtime render times, say 5-10 frames/sec. The render QUALITY won't be as high as a still render, though.
The whole idea of using a high quality gaming render engine to render out animation is brilliant. No, you don't need 60 FPS realtime rendering, you need the highest quality the game render engine can deliver, and devil may care if frame rate drops to 5 FPS!
Recording an animation with the Crysis engine at highest settings, and then playing back at the correct framerate should result in a stunning animation.
There's one big caveat that I know of. Rendering on the GPU is certainly possible, but how does one capture the output? nVidia has CUDA and Gelato, which only works on the 8xxx and higher series of graphics cards. I don't know whether ATI has a comparable library to program/access the GPU. Anyway, if they have, it is most certainly totally incompatible with nVidia's solution.
While ATI and nVidia dominate the graphics market, there's also Intel for embedded graphics. Yet another totally different GPU, and I'm pretty certain that the Intel embedded graphics chips can't be used for anything else than just plain graphics.
Matrox is another player, and how about the professional high end cards? I remember a card called Wildcat, not from ATI or nVidia or Matrox, that was quite popular at the time.
The advantage of OpenGL is that it's just OpenGL. Write a graphics program against OpenGL and you don't have to worry about the hardware. You can't do that when using CUDA or similar libraries, those are very hardware specific.
Once a general hardware abstraction standard for GPU processing is established (hopefully as an OpenGL extension, that would be the most "acceptable" standard IMO), GPU based calculations, including rendering will flourish. After all, 128 parallel streaming processors in a GPU will be QUITE a lot faster than 4 or 8 cores in a CPU!
The pen is mightier than the sword. But if you literally want to have some impact, use a typewriter