I suspect the processor adds the temperatures of all 4 CCDs and then incorrectly divides them by 2 instead of 4.
EPYC Rome ES CPU temperature
- Thread starter chraac
- Start date
-
- Tags
- core temperature cpu epyc
chraac
Member
Thanks for the important picture, may cpu is B0 stepping and the temp in ipmi is not as high as the die average, from you picture I see your CPU also high in this die average temp, so I think it's probably because of the ES stage CPU.
HWinfo sensors: CPU Die (average) is IMHO nonsense. in the IPMI sensor page is nothing hot.
IPMI sensors
Last edited:
chraac
Member
Maybe it's calc by adding all the temp and then divide by some factor
would be helpful to see a 7452/7502 production units CPU die (average) value.
chraac
Member
it's not so easy to find a 7452/7502 with windows os and run hiwnfo
chraac
Member
Yep, I think the 7452 and 7542 are just different in frequency but same in silicon die, so they are same in the physical chip...
chraac
Member
Ok, but EPYC with windows and running hwinfo is hard to find.
Chozo4
Active Member
If it helps... I'm running a Production Sample EPYC 7252 on bare metal here under Windows 10 1809. Noticing a similar trend with the CPU Die (Average) temp accordingly which may point to this not being an ES exclusive issue or even specific to 7452/7502 for that matter. There is barely any load on the CPU at the moment.
I would be happy to lend myself to any further testing needed.
Last edited:
Well, then this seems to be an issue of the SMU which calculates such a high average value.
There's a new feature in works now, which is currently just waiting for AMD's approval: per-core temperature monitoring. If this gets approved (which I expect/hope to happen in the next couple of days), it would give you a better insight into individual temperatures to confirm what's correct and what isn't.
There's a new feature in works now, which is currently just waiting for AMD's approval: per-core temperature monitoring. If this gets approved (which I expect/hope to happen in the next couple of days), it would give you a better insight into individual temperatures to confirm what's correct and what isn't.