CUDA applications that support more than one graphics card?

[Monkeypillow]

From what I've used, it seems that most applications will only support one card. Is there any way to leverage the power of more than one card simultaneously with applications like Badaboom?

 

[Silus]

From what I've used, it seems that most applications will only support one card. Is there any way to leverage the power of more than one card simultaneously with applications like Badaboom?

Any CUDA application, if coded accordingly, can take advantage of any number of GPUs on a machine (and I don't mean SLI, I mean just plugged on a PCI-e slot). The FASTRA project is a good example. However, the applications that exist, are not for the ordinary consumer, but rather for investigation purposes, oil & gas prospects, etc. Think NVIDIA's Tesla platforms.

But again, for the ordinary consumer, I don't know any.

 

[Elledan]

One commercial CUDA application I'm developing at the moment does also scale across multiple GPUs. As said, it is something that has to be explicitly added to the application, as the host (CPU) thread(s) have to send the CUDA kernel & data to the proper GPU/other CUDA device.

This particular application is aimed at FPGA/ASIC design simulation and thus unlikely to be used by most people