madelectron Posted March 19 Share Posted March 19 I have two unraid servers both running epyc 2nd gen cpus, one is running a 7532 and the other one is running a 7282. This problem exists on both servers. Testing was done on the server with the 7532 32c/64t 2.4ghz base, 3.3ghz turbo. While running unraid, with running cpu govenor at on-demand, frequency scaling works but never goes above base clock of 2.4ghz. If I set it to performance, all cores/threads top out at 2.4ghz and statically stay there. Never any turbo boost. Using the same server, only changing boot drive from unraid usb to a HiveOS usb, and try to cpu mine crypto with the server. Mining on 1 thread, that thread runs at 3.3ghz, at 16 threads it's running at 3.2ghz, at 32 threads it's down to 3ghz on the mining threads, and finally with all 64 threads mining the cpu frequency is at 2.8ghz. So the server is properly configured to handle boost clocks from the cpu, but for some reason it is not happening in UnRaid. Is there any update that can enable this for EPYC cpus? Maybe a newer cpu frequency scaling driver? Quote Link to comment
JorgeB Posted March 20 Share Posted March 20 It's working for me, what do you get from: cat /sys/devices/system/cpu/cpufreq/boost Also, where are you monitoring the CPU clock? Quote Link to comment
madelectron Posted March 20 Author Share Posted March 20 I'm monitoring the cpu frequency per thread in Netdata. The returned result from your command is 1. As you can see with the attached image, every thread that has a burst load it being hard capped at 2.4Ghz, when it should be boosting up to 2.8-3.3Ghz, depending on the overall CPU load. Quote Link to comment
JorgeB Posted March 20 Share Posted March 20 33 minutes ago, madelectron said: The returned result from your command is 1. That suggests it is enabled, I don't know if Netdata shows the correct values, do you get the same clock speeds from: watch -n 1 grep MHz /proc/cpuinfo Quote Link to comment
madelectron Posted March 20 Author Share Posted March 20 using the watch command as it reboots and spins up 15 windows VM's, it is indeed turbo boosting threads up to 3.3Ghz, seems to be an issue with netdata and not the cpu scheduler. Thank you for helping me figure this out. 1 Quote Link to comment
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.