PhysX Acceleration Question - Team Fortress 2.

With the PhysX being implemented into the Nvidia GPUs, which would eventually cover the whole GeForce 8 and 9 Family GPUs, I wonder if the new PhysX engine will help with games that does not have PhysX built in; more specifically will PhysX help with the performance of games such as Team Fortress 2, where Physics plays a major role in the game and that CPU is most likely the limiting factor?

 

no.

physx isn't even going to help performance with games that do implement it. unless a task that requires use of both the gpu and cpu is HEAVILY cpu limited, physx is going to cause a performance HIT. the benefit is that calculating physics on the gpu will allow for more complex physics systems to play out in real time.

 

The Source Engine (thus, TF2) utilizes the Havok physics engine, which is owned by Intel and will be supported on ATI GPUs.

That said, most games will probably not be retroactively worked into Physx or Havok if they weren't built initially for that physics engine.

What needs to be built here (much like with CUDA and GPGPU) is a unified physics engine or else each GPU will only support a fragment of the physics engines available. That, or both engines need to be licensed onto both brands of GPU. Or maybe one will die off like HD-DVD.

 

Thank you both for the detail replies. Too bad the idea of "GPU Physics" is not as matured as I thought it is. As StormEffect said, I wish there is a unified GPU physics standard, so all the physics in any games can be offloaded to the GPU: if not fully at least partially. However, I believe what is most likely to happen is that the game makers will need to spent extra effort support both Physics Engines, and I just don't see ATI and Nvidia operating to create a unified physic engine. At least not this early in the game.....
Reps+ for Both.

 

But consider the fact that technically AMD and Intel are working together to get Havok working on ATIs GPUs.

Also there is a far off bright side, DirectX 11 will require a standardized GPU API, I know this will apply to things like video transcoding, and I also believe it may apply to physics engine processing. In other words, regardless of the engine, using DirectX 11 either GPU could help accelerate physics.

Then again, it would be convenient if we didn't have to wait a few years for this. Nvidia and ATI could just cross-license or share information. Obviously this would the best solution, but of course this is the hyper-competitive corporate semi-conductor industry we are talking about. Who wants to get along and progress technology when there is money to be made!?!?