Hi,
In the programming guide pdf file, it is said that CUDA devices by an application running on a multi-GPU system is only guaranteed to work if these GPUs are of the same type.
Can some one please tell me what’s the problem if an application (multiple threaded) running on different types of GPUs even though they are all 1.1 compatible?
Can it run such application in some situation? Thanks!
Regards
-Steven
As far as I know, there should be nothing preventing an app from running in that situation. The only important consideration to keep in mind is multiple GPUs of different types complicates load balancing.
We’ve got a number of machines with multiple different CUDA-capable GPUs in them, and they all work fine. If you want to load balance between GPU models with differing numbers of SMs, things get a little more interesting, but nothing terribly difficult to deal with.
Cheers,
John