Hacker News Reader

> There is an argument that says that if you have the GPU anyhow, you might as well use it even if it is inefficient. I do not like this argument.

Why not? People use CPUs all the time where GPUs are much more efficient, why not the other way round as well? If you have GPU, why to keep it idle?

View on HN

I think he touches on why not during the comments about block chain. It is probably about “wasting” power and energy with inefficiency.

Alright, with proof of work you indeed waste energy, but it's still just tangentially related - if you do it on CPU, you are probably wasting even more. However, I see e.g. my Core-M laptop as vastly underused - if I run Deep Learning there, it's on CPU, which is inefficient, but given it has only Intel HD graphics with OpenCL for which I don't have a working TensorFlow/Keras, I might have it 100x slower than I could. But I still use CPU for simpler Deep Learning models when I need it "on the road", though wish I could use iGPU. OTOH when I have a problem that requires AVX-512, it's much more efficient to compute it on latest Xeon, but if I only have a stack of single-precision GPUs, I might use those even knowing my results will be less precise and they would take longer to compute?

Reply on news.ycombinator.com