A month ago I ran benchmarks on CPUs vs. GPUs on Google Compute Engine, and found that GPUs were now about as cost-effective as using preemptible instances with a lot of CPUs: http://minimaxir.com/2017/11/benchmark-gpus/
That was before preemptible GPUs: with the halved-cost, the cost-effectiveness of GPU instances now doubles, so they're a very good option for hobbyist deep learning. (I did test the preemptible-GPU instances recently; they work as you expect)
That was before preemptible GPUs: with the halved-cost, the cost-effectiveness of GPU instances now doubles, so they're a very good option for hobbyist deep learning. (I did test the preemptible-GPU instances recently; they work as you expect)