CUDA and SLI

Hello,

I am interested in using several graphics boards at once in the context of neurosurgery simulation.

Is it possible to combine CUDA-based GPU programming with SLI-based inter-board communication, and have two GPUs interact with each other in high throughput communications.

Best wishes,

Michel

No. You have to program for multiple GPUs explicitly. The SLI interconnect is not nearly fast enough to support running a single CUDA kernel on multiple GPUs.