I received my 5950X yesterday—primarily for compiling stuff at work (I work from home, like so many others during Covid-19), but also to evaluate Zen 3 for a possible Threadripper upgrade for the next iteration of analysis.sesse.net.
I'm not getting the boost clocks I had hoped for, so I'd like to ask other Zen 3 users if they have any experiences:
Are you getting the promised single-threaded boost clock, on any core, on any load? My cores seem to vary between 4.5 and 4.7 GHz, with the cores on CCX 1 being much worse than CCX 0. I've seen 4.85 occasionally, but never the promised 4.9 (and most reviewers' samples seem to hit over 5.0). I'm running a bit over 50 degrees, so this specific test is unlikely to be thermally bound.
When I run single-threaded workloads (like Stockfish in single-threaded mode, or just a plain infinite loop), all other cores on that CCX boost to a bit over 3 GHz for no apparent reason. I've bound the task to a single core, so it shouldn't be migrated around. Is anyone else seeing the same? (I'm running Linux, and I don't have an easy way to test with Windows.)
What all-core boosts do you get when running Stockfish (with NNUE) on all 32 threads? (I get 4.1 GHz on CCX 0 and 3.75 GHz on CCX 1, dropping to 4.0 GHz and 3.7 GHz as things slowly warm up to 87 C. I haven't tried really long runs.)
What nps numbers do you get? I get what I perceive as really good single-thread numbers (approx. 2300 knodes/sec) and medium-good all-core numbers (approx. 38500 knodes/sec when thermals have stabilized-ish). Start position, default hash, Stockfish from latest git (d706ae62d), GCC 10.2.1, PGO build, BMI2.
Any data points are interesting!
…at least, it compiles Stockfish in basically zero time.
That's interesting. I'm getting ~62 Mnps on that benchmark, which is markedly worse than the ~72 others are getting, but I'm unsure if the other 5950X are overclocked? The slowest one is listed as “5950X @ 3.95 GHz”, and I'm unsure if that means the base clock has been raised from 3.4 GHz to 3.95 GHz, or if that is just the frequency it's boosting to when all cores are active.
I received my 5950X yesterday—primarily for compiling stuff at work (I work from home, like so many others during Covid-19), but also to evaluate Zen 3 for a possible Threadripper upgrade for the next iteration of analysis.sesse.net.
I'm not getting the boost clocks I had hoped for, so I'd like to ask other Zen 3 users if they have any experiences:
Are you getting the promised single-threaded boost clock, on any core, on any load? My cores seem to vary between 4.5 and 4.7 GHz, with the cores on CCX 1 being much worse than CCX 0. I've seen 4.85 occasionally, but never the promised 4.9 (and most reviewers' samples seem to hit over 5.0). I'm running a bit over 50 degrees, so this specific test is unlikely to be thermally bound.
When I run single-threaded workloads (like Stockfish in single-threaded mode, or just a plain infinite loop), all other cores on that CCX boost to a bit over 3 GHz for no apparent reason. I've bound the task to a single core, so it shouldn't be migrated around. Is anyone else seeing the same? (I'm running Linux, and I don't have an easy way to test with Windows.)
What all-core boosts do you get when running Stockfish (with NNUE) on all 32 threads? (I get 4.1 GHz on CCX 0 and 3.75 GHz on CCX 1, dropping to 4.0 GHz and 3.7 GHz as things slowly warm up to 87 C. I haven't tried really long runs.)
What nps numbers do you get? I get what I perceive as really good single-thread numbers (approx. 2300 knodes/sec) and medium-good all-core numbers (approx. 38500 knodes/sec when thermals have stabilized-ish). Start position, default hash, Stockfish from latest git (d706ae62d), GCC 10.2.1, PGO build, BMI2.
Any data points are interesting!
…at least, it compiles Stockfish in basically zero time.
If you use Ryzen Master to get OC-ed CPU what you experienced these are natural phenomenons.
If you want fixed CPU clock you can set it in BIOS only.
I can set on my Ryzen 9 3050x 4.40 GHz on all 16 cores, but it was more stable (and lesser hot using NOCTUA air cooler) setting on 4.0 GHz. I am not interested in OC-ed only one core.
"Zero time" is a poetic magnifying.
Sesse wrote: ↑Sat Dec 12, 2020 6:10 pm
That's interesting. I'm getting ~62 Mnps on that benchmark, which is markedly worse than the ~72 others are getting, but I'm unsure if the other 5950X are overclocked? The slowest one is listed as “5950X @ 3.95 GHz”, and I'm unsure if that means the base clock has been raised from 3.4 GHz to 3.95 GHz, or if that is just the frequency it's boosting to when all cores are active.
The list is rather disturbed and incorrectly parametrized.
I frequently do single-threaded work, and it's fine if that's 25% faster.