Multi-GPU Understanding Concurrent kernel execution on multi-GPU

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi,

I am new to multi-GPU systems. I shall be grateful if you guys can verify my understanding:

1- Multi-GPUs can be used to run “same” or “different” kernels SIMULTANEOUSLY on different GPUs.

2- Computation speed will be thus doubled on a 2 GPU system as compared to a single GPU system.

3- For getting the advantage of two GPUs we need to create two host threads to control two GPUs.
These two host threads will launch two kernels meant for two GPUs.

4- We will have to disable the SLI mode if we want to utilize the two GPUs for Computations.

5- The SLI can only benefit gaming applications.

Thanks in advance

P.S: Sorry for multiple post, I inadvertently posted it as a reply in another forum.

2

Everything except 2 is pretty much correct.

Because two GPUs have to share the PCI-e bus and host CPU interconnect and memory bandwidth, it is rare to get an effective doubling of computation throughput with two GPUs because of host resource contention.

3

Thanks for your reply. You cleared my doubts :-)

4

4 is old news. Drivers for the past year or so are able to run multi-GPU CUDA when SLI is enabled. The game community needed this so they could run physX on a chosen GPU at the same time as a game in SLI mode.

Now that I think of it, the whole physX/CUDA buzz has died down, hasn’t it. I don’t recall seeing any articles on cool new games with GPU-acellerated physX support since all that hype over a year ago. Odd how quickly a whole industry can loose interest in something so quickly.