Please forgive the novice question, but I am curious if any of Mathematica's functionality might see improved performance using the new AVX-512 extensions available using Skylake-X (recently announced by Intel). In other words, does any part of Mathematica's functionality use AVX extensions to enhance performance, and therefore might perform even better with AVX-512?
As a non-computer scientist, I've often wondered if functions with the Listable attribute achieve their amazing performance by using AVX extensions, or some other computer science "wizardry."
As I contemplate purchasing new hardware in the next 12 months, I'd really like to know if purchasing Intel is worth the price premium over the new AMD Ryzen chips. Mathematica's performance is one of the major deciding factors for my decision, and I do rely heavily on code the takes advantage of low-level listability.
Any comments / speculation / warnings are appreciated! Thank you.
Hi Todd. I am interested in whether or not you found any information related to this post. I am wondering this same thing, now 3 years later, as I seek a processor most naturally aligned with Mathematica. -Mark
Hi Mark. (Sorry for the slow reply.)
In a word, "no." There really isn't much available speaking to the AVX issue, nor on whether the AMD slowdown "bug" due to MKL received any attention from Wolfram. I've asked multiple questions, and largely received no answers or unhelpful answers.
Here are a few threads that may mean something to you based on your personal background:
In the end, I decided to purchase a new AMD system recently and went with the 3960x Threadripper from AMD.
I've been happy with it's performance using Mathematica so far!
Good luck in your decision. Todd
It uses the MKL library, so to some extent it does.
These threads may be useful:
(same as above)