Hacker News new | ask | show | jobs
by runholm 3184 days ago
Performance per $ isn't great, but the hardware is good. If you want to crunch data like a champion you need to spend the same $ on a desktop with a big fat GPU suited for GPGPU tasks and with proper cooling. The rest of the system doesn't really need to be insane, but you have to be able to keep up with the GPU.

I doubt the iMac can keep running with a high load over time without thermal throttling.

1 comments

The other crux is that it is a Radeon GPU and not a Nvidia GPU, so you cannot (effectively) use any Deep Learning library that targets CUDA/cuDNN.