3 ms·
Serious question, is there are reason to have a GPU and that many cores? My assumption would that either the GPU or the CPU is the bottleneck, most likely the
by joe_the_user 6y ago
Serious question, is there are reason to have a GPU and that many cores?
My assumption would that either the GPU or the CPU is the bottleneck, most likely the GPU. Why not spend money for more GPU and fewer cores?
- datameta 6y agoFor shallower models CPUs aren't overshadowed by GPUs as much - but beyond a certain # of parameters the CPU loses out as GPUs do vector math highly efficiently. Some experiment results: https://azure.microsoft.com/en-us/blog/gpus-vs-cpus-for-deployment-of-deep-learning-models/ https://azure.microsoft.com/en-us/blog/gpus-vs-cpus-for-depl...
- joe_the_user 6y agoBut that's a fight between either CPUs OR GPU. It doesn't given indication that both could be useful simultaneously.
- chrisbroadfoot 6y ago"The team also tested their approach on a collection of 30 challenges in DeepMind Lab using a more powerful 36-core 4-GPU machine. The resulting AI significantly outperformed the original AI that DeepMind used to tackle the challenge, which was trained on a large computing cluster." Well, they presumably tested the same CPU with 4 GPUs (2080 Ti I think) - maybe they wanted to compare.