Anybody working with AMD and CUDA Multi-GPU?

Hi, sorry if the name of the topic isn´t right, plus my english is not top notch.

To bussines. I have been ask to benchmark with Linpack a future HPC server with:

4 AMD Opteron 6200 (12 cores)
2 EVGA Gforce GTX680 (4 GB GDDR5)

so I followed some tutorials, and I manage myself to compile it XD

I used a Linpack_FERMI_v15 and follow all the steps.
But I find myself with 2 questions,

  1. Does the amd chipset support multiple GPU? because when I run the xhpl it does the memswap but it only uses 1 of the GPU

  2. Everywhere I looked, can not find a single person/manual where I can “Confirm” that all I did has been done sort of right :S

I will appreciate any kind of help. Thanks

ps: I´m a neewbie on all these GGPU/CUDA/Parallel Programing.

oh… also is there a Kepler_HPL? where I can find it?

The chipset has nothing to do whether multi-GPU is working or not. I am not familiar with Linpack, but so long as the driver recognizes both GPUs, there is nothing that could prevent multiple GPUs from being used simultanously. It is either that the host application is not even trying to use both GPUs, or some threading issue yet again in the calling application is preventing from both GPUs being used simultanously.

ok, thanks. Now I can disscard that idea.