Octodad on PS4 uses hardware PhysX?

[Dribble]

Of course it works on PS4 and xbone, and yes it will use whatever resources (gpu's or cpu's) that is most efficient. It doesn't *require* AMD to add it, or cuda - nvidia will do a physx port for each of the consoles using whatever languages are available on those consoles. It's their library, they get it to work on all sorts of hardware, the fact it doesn't work with PC's except in cuda is an artificial limitation nvidia choose to make, not because they couldn't (quite easily) do it.

For example in the PS3 it used the cell SPE's which are basically compute cores like gpu's now have.

That said as the consoles have an excessive number of cpu cores I would have thought it uses 1 or more of them, not the gpu which is almost certainly going to be required to run flat out drawing pretty pictures.

 

[bystander36]

This is an artificial limitation though. There is no real reason it could not be run with Direct Compute or OpenCL. There is nothing special about the CUDA API that is required for PhysX to work.

It just happens to be developed for CUDA, and to some extent Direct Compute currently.

Yes, it is developed on CUDA and a few CPU's, therefor it requires CUDA or the specific CPU's it was designed for.

I'm not sure where you get the "artificial" part. They would have to completely rewrite it to work on other hardware.

 

[Stuka87]

Yes, it is developed on CUDA and a few CPU's, therefor it requires CUDA or the specific CPU's it was designed for.

I'm not sure where you get the "artificial" part. They would have to completely rewrite it to work on other hardware.

No, it would not require a full rewrite. Do you understand how an API works? Yes, there would be some change, a change they have already done for Direct Compute. But certainly not a full rewrite. The fact that it can run on a CPU without any issues shows that there is nothing CUDA dependent.

 

[bystander36]

No, it would not require a full rewrite. Do you understand how an API works? Yes, there would be some change, a change they have already done for Direct Compute. But certainly not a full rewrite. The fact that it can run on a CPU without any issues shows that there is nothing CUDA dependent.

It is CUDA dependent because they wrote it for CUDA.

When people talk about artificial requirements, they are talking about just restricting use with a simple line of code. This requires a rewrite of at least major portions of the code. While it is possible to do, that isn't an artificial limitation.