Guys, I’ve been breaking my head about this for quite some time now. I work with Computer Vision (with and without Deep Learning techniques). I am trying to build a PC for the same with more focus on Deep Learning techniques(80%). GPU, memory and storage requirements for this PC are covered (2070 super, 16gb 3200 b-die corsair lpg and adata xpg 1tb nvme m2 ssd). I am not able to decide on the CPU. I came across the mkl issue. I understand matrix computations on the CPU are faster on Intel than an AMD processor. I’ve come across blogs which show this difference. I got aware of the benefits of using libopenblas for speeding up stuff and matching the speed for amd and intel. I also came across articles where they mention using debug variables and patched versions of numpy to match the speed.

So, it takes a bit of an effort but, techinically, it is possible to match the matrix related computational speed of an amd processor to that of an intel processor.

• Am I correct in assuming this?
• can we assume things will get better? Like, will Intel ever start supporting AMD processors? Intel can find some other ways to cut mkl support for amd, though. That seems more likely. The only reason I have this doubt is because AMD does have the architecture for supporting mkl.
• should I not worry too much about this?
• I did not, however, find a direct comparison of i7 9700k vs 3700x for matrix computations. I want to know if it is significant.

I have the following summary about the 2 processors. Please feel free to add/correct/rank the key features that I mention below.

AMD:

cheaper lesser power setup cost is less - i need not buy a fan for the processor. also, ryzen's fan looks cooler. more threads - correct me if i am wrong here but this allows for more computations when compared to an i7. 

Intel:

mkl support - easier to setup/no headaches about support loss sli support motherboards are more common compared to x570 chipset motherboards for ryzen 

Thanks in advance!