Nvidia Physx SDK 3.3 on ATI 7850 card

Started by
3 comments, last by rave3d 8 years, 8 months ago

I am trying to compile the samples in Nvidia Physx 3.3 on my system.

The system is currently using an Saphire ATI 7850 graphics card.

On compiling in Visual studio 2008, the samples solution gives the error-

"The program can't start because nvcuda.dll is missing from your computer. Try reinstalling the program to fix this problem."

Please suggest on the way forward.

Can Physx 3.3 run with ATI graphics cards.

Advertisement

greetings,

PhysX does not work on ATI, it only works on Nvidia and CPU(bad performance)

To compile the code you need an Nvidia GPU.

Hope this helps.

HyperV

"The program can't start because nvcuda.dll is missing from your computer. Try reinstalling the program to fix this problem."

How have you initialized PhysX? You should be able to avoid using the cuda modules.

To compile the code you need an Nvidia GPU.

This isn't true, I compile my game on an AMD PC (it just fails to initialize Cuda at runtime on this PC).

Yeah I don't know where people get this idea that PhysX does not work on AMD. I guess I'm a special case as I do not have a Nvidia GPU. So as not to confuse the op, PhysX will utilize the CPU for simulation on platforms that do not have Nvidia GPUs installed or does not request GPU acceleration for the PhysX simulation. I've seen that error before, and it most likely attributed to the fact that the SDK samples are trying to initialize the CudaContextManager ( which required code ), since this is linked into the sample executable, you will inherit this implicit dependency. One workaround, would be find all references to where the CudaContextManager is being created in the samples and comment that out and try rebuilding. I've never really investigated the issue any further as I use the samples as reference.

To be exact the PhysX GPU acceleration works not an ATI/AMD or Intel CARDS, but you should still be able to use the non GPU parts on systems where you have no Nvidia GPU

Just as cgrant say make sure the CudaContextManager isn't initialized in your code, und check which libraries you are linking against.

This topic is closed to new replies.

Advertisement