Do machine learning and deep learning programs take advantage of multi-core multi-thread CPU?

Still trying to decide which CPU to buy. Some mentioned that machine learning and deep learning programs such as Tensorflow and Python are single threaded. Having a good GPU such as 1080Ti is more important than having a high-end CPU. Others mentioned that it is good to get multi-core, multi-thread CPU. So, what is the recommended number of cores and threads to build a system to do machine learning, deep learning and data science research?

 

TF is single threaded? Since when? If you train or do inferencing on CPU it's going across cores.

1080 Ti should be the first priority. Get 8 cores minimum.

 

Thanks. What is the recommended number of cores and threads? Somebody also suggested getting a CPU with cores that are twice the numbe of GPUs would be sufficient!? Can't decide between Intel and AMD.

 

If you're doing 4-6 GPU EPYC or dual Intel. 1-3 GPU single Intel is OK desktop W or Threadripper.

Intel MKL is good and you'll see the software performance boost. AMD doesn't have the software investment like Intel does.

If you can find a cheap Gold or Plat on eBay or the forums that'd be good. There isn't much data on ES CPUs from this gen so make sure if you do go that route you find a good one. I wouldn't spend 2k on a cpu

 

I am going to get a 1080Ti first and see how it goes. May upgrade to three more 1080Ti or get four better GPU later. In this case, how many cores in the CPU is recommended?

 

I am new to deep learning. I am doing a project on 3D reconstruction. I want to use both VAE and GAN. Can someone explain to me how GAN work. The input and output processing. Thank you. How can I feed in 2D image and get 3D model. Thank you