CPU/GPU cross Computing

[Seraphimsan]

I had an interesting idea, though I'm sure some one more qualified must have attempted it before me.


But would it be possible to use both the cpu of a computer as well as the gpu's on all graphics cards on the board for mass computations? I'm sure there are much greater ideas for mass computations, but for a single user, with a slightly above standard home built PC. This might come in handy...

Any one have any thoughts on this?

[z05DSL3A]

Is this the sort of thing you are talking about?
http://www.nvidia.com/object/cuda_what_is.html

[mcleano]

Sounds very cool!! Although I imagine someone more experienced is going to put a massive bulldozer on this one :(

[Seraphimsan]

holy mother of....yea, thats it XD and I hadnt been to nvidia's website in like 6 months... I probably had heard of it somewhere a while ago and forgot about it. Byt yea, straight on the mark

[helios]

My experience with CUDA is that if you better be careful with what you give the GPU to do. One of the few times my computer has crashed was while running the GPU version of Folding@home. Granted, FAH is exceptionally resource intensive, but still.
Also, CUDA works best for "embarrassingly parallel" problems. The GPU has 32 cores, but each of the cores is slower than the CPU.

I just checked with Wikipedia and I was, to my own surprise, completely right:

While GPGPU can achieve a 100-250x speedup vs a single CPU, only embarrassingly parallel applications will see this kind of benefit. A single GPU processing core is not equivalent to a single processing core found in a desktop CPU.

[Denis]

I think OpenCL is the best in this context