SLI

Hello, i’ve got some questions about SLI and CUDA.

  1. If i use two GTX 480(connected by SLI) for example, does cuda see a big GPU with 960 cores or 2 GPU with 480 cores? Is it compatible?
  2. Both same cards without SLI, is it possible to run a kernel on the first one and at the same time run a second kernel on the second card ?

Thank you

  1. 2 Separate GPUs.
  2. Doesn’t matter whether you enable SLI or not. You still see 2 GPUs and can launch kernels, allocate memory, etc. on each independently. Read the multi-gpu section in the CUDA 4.0 programming guide for the details.

Thank you i’ll check it!