5 ms·
This is the perfect chip for continuous, low-memory number crunching. For everything else... not so much. I mean, this chip consumes 74W when idle, of which a
by deng 8y ago
This is the perfect chip for continuous, low-memory number crunching. For everything else... not so much. I mean,
this chip consumes 74W when idle, of which almost 90% are spend on the interconnect. That's insane. Most important bit in the review for me:
"After core counts, the next battle will be on the interconnect. Low power, scalable, and high performance: process node scaling will mean nothing if the interconnect becomes 90% of the total chip power."
- close04 8y agoMost people used TR for day to day desktop use. This really pushes it into the workstation segment. While efficiency is important, results are even more so.
- Symmetry 8y agoThat's what it consumes when one core is active. Idle should be much lower, though you're right that to the extent you're doing anything on most desktops it's going to be just a few threads.
- Coding_Cat 8y agoWhy do you say low-memory? I got the opposite feeling. I've been drooling over EPYC and TR2 ever since they released the specs for my memory bandwidth limited projects.
- snuxoll 8y agoThe NUMA configuration on TR2 results in two 8-core CCX's without direct memory access, so you still "only" have four memory channels and half the cores have to hop over the infinity fabric to access any of them.
- celrod 8y agoIf that number crunching is highly vectorizable, avx-512 of skylake-x will likely still be faster. Eg, an Intel 7900x is over twice as fast at multiplying big matrices (multithreaded) as an AMD 1950x. Of course, a similarly priced GPU would stomp both whenever CPU-GPU latency isn't much of an issue. Meaning "highly vectorizable number crunching that you can't practically run on a GPU for some reason" is probably fairly niche. In LINPACK the even the 1950x beat the 7980XE on phoronix: https://www.phoronix.com/scan.php?page=article&item=amd-linux-2990wx&num=4 https://www.phoronix.com/scan.php?page=article&item=amd-linu... and I'd have thought LINPACK would really reward AVX... Maybe it isn't optimized to take advantage of it? I love wasting time playing with number crunchinng microbenchmarks, and when you do that avx-512 seems so unassailably impressive, it's surprising to see this not bourne out in larger benchmarks. (Also worth pointing out you need to tell gcc -mprefer-vector-width=512 on top of march=skylake-avx512, otherwise it'll prefer 256 bit vectors). Anyway, those benchmarks definitely support compiling and number crunching. Pretty ideal for stuff like monte carlo.