Recommended Hardware Setup for CUDA GPU Paralellization

Hi,

We're looking for the best GFLOP/$ setup for matlab. We want to spec out a machine for fast computation.

NVIDIA's GE Force GPUs are significantly cheaper than their Tesla and Quadro GPUs.

- Can we use the GE Force GTX 980 with Matlab, does it have any limitations with double precision computations?

- Can we use multiple GTX 980s in the same machine?

- Which CPU would people recommend for such a setup - i7? X99? Xeon?

Thanks very much for your help,

[Max]

I notice you posted this question in March, so if you've already made an internal decision, please share.

The limitation of GPUs is from the number of PCIe lanes the CPU has, and by how many large PCIe slots the motherboard has. The X99 architecture has many more PCIe lanes. The other thing to consider is the heat output of so many video cards in one box, as adequate cooling can be a challenge as well as the noise output of cooling such a machine.

Compare the 5960x vs the 4790k

http://ark.intel.com/products/82930/Intel-Core-i7-5960X-Processor-Extreme-Edition-20M-Cache-up-to-3_50-GHz

http://ark.intel.com/products/80807/Intel-Core-i7-4790K-Processor-8M-Cache-up-to-4_40-GHz

Any NVidea series GTX 600 or higher card can use Matlab.