Not strictly a CUDA question - but this is related to the Tesla, and I couldn’t find a better place to post this question.
I have this OpenGL/Cg/GLUT code that does some signal processing on the GPU. (Most of this code was largely based on Dominik Godekke’s tutorials) When I first wrote this, I was using a Quadro FX 1400, and everything worked just fine. Now we’ve acquired an NVIDIA Tesla, and I want to run this code on the Tesla.
My setup now has a GeForce FX 5500 PCI, and a Tesla C870. (I had to get rid of the Quadro FX 1400 because my stupid PC has only one PCIe x16 slot.)
I dug around a little and figured that an easy way to do this would be to use the GPU Affinity extension and set the affinity to the Tesla and then use the rest of the code unchanged. (I am not really an OpenGL expert so if there is a better way of doing this, I am very interested in hearing about it).
Unfortunately, I tried using the glewinfo tool and it appears that when I plug the FX 5500 and the Tesla into the PC together, the NV GPU Affinity extension is not supported. (At least, I can say for sure that it is supported on my dev PC which has a Quadro FX 1400, and it is not supported on the test machine that has the 5500+C870). Both PCs are running the 169.09 drivers.
My question is: why is this so? Does the GPU affinity function have to be supported by the GPU as well as the driver? (This seems like an all software feature to me …) If so, does the FX 5500 support the feature. What should I do to get this setup to work?
All help will be much appreciated.
Thanks in advance,
Pramod