A GPU cluster would work better but if you're only testing things out using CUDA and want 200GB networking and somewhat low power all in one this would be the device for you