In a recent interview, Tony Tamasi, senior vice president of Content and Technology at Nvidia, said that he doesn't have a particular preference to what physics library game developers use, as long as it runs on Nvidia's GPUs.
"We've been working on physics for a very, very long time - back in 2005, we showed off our first GPU-based physics demos using the Havok library,
" explained Tamasi. "Havok could never really decide if they wanted to do the GPU or not and I'm a little sceptical of whether we'll ever see GPU-based Havok physics.
"In the end, we decided to acquire Ageia because they'd already got a really nice SDK designed for parallel processing - we just did a C for CUDA implementation and we've been doing that ever since. We've also been helping the Bullet Physics guys with their GPU implementation - it's based on our OpenCL samples and they've been using Nvidia hardware to do their development.
"Anyone who wants to do physics on the GPU is all goodness for us, so if the Havok guys get around to doing some OpenCL stuff on the GPU, that's good for Nvidia. All we really want is to see lots of GPU based implementations - we did the PhysX thing because we wanted to get the ball rolling.
We then went onto discuss the future of game physics and whether we'll ever see real gameplay physics in PC games given that there are very few PC exclusive titles these days. Tamasi said that we're already starting to see gameplay physics being introduced into games such as Batman: Arkham Asylum
- there's just more of it in the PC version.
"Partly, this is more of a game design choice than a technological barrier [for consoles],
" he explained. "It is clear if game developers want to really do a good job of taking physics to the next level, they actually do have to really think about it. If they want to have walls that can be destroyed, their AI needs to solve for all new visibility and path finding capabilities.
"Frankly, we're still in the early days of physics, much like 3D in the early days where you had Z-buffer and bilinear filtering - things are obviously quite different nowadays. I'd say we're on the cusp of physics taking off, and it's not just physics - it's simulations in general.
It's pleasing to hear Nvidia say that it doesn't matter what physics library game developers use, as long as it runs on the GPU, because GPU-agnostic game physics is what gamers have wanted ever since the major players first started talking about improving gameplay physics. Whether developers will adopt something other than PhysX though is another question entirely - a lot of that is going to depend on the tools available for alternative libraries to PhysX. Moreover, game developers are unlikely to get the same level of support from Nvidia's DevTech team if they opt for a GPU-agnostic library, such as Bullet Physics.
We spoke to a number of Nvidia's DevTech engineers during our visit to San Jose and one thing they all said was that the investment in improving the tools for implementing PhysX isn't likely to slow down. Ease of implementation is hugely important to developers (and publishers) and so it could be a while before we see widespread adoption of GPU-agnostic physics libraries.
Discuss in the forums