On 13/03/2012 10:28, Volodymyr Kostyrko wrote: > Matthew Seaman wrote: >> On 13/03/2012 08:59, Volodymyr Kostyrko wrote: >>> The only other weird thing about this server is: >>> >>> dev.cpu.0.temperature: 37,0C >>> dev.cpu.1.temperature: 37,0C >>> dev.cpu.2.temperature: 35,0C >>> dev.cpu.3.temperature: 35,0C >>> dev.cpu.4.temperature: 43,0C >>> dev.cpu.5.temperature: 43,0C >>> dev.cpu.6.temperature: 38,0C >>> dev.cpu.7.temperature: 38,0C >>> dev.cpu.8.temperature: 38,0C >>> dev.cpu.9.temperature: 38,0C >>> dev.cpu.10.temperature: 37,0C >>> dev.cpu.11.temperature: 37,0C >>> dev.cpu.12.temperature: 33,0C >>> dev.cpu.13.temperature: 33,0C >>> dev.cpu.14.temperature: 34,0C >>> dev.cpu.15.temperature: 34,0C >>> >>> And it's consistent - cores 4 and 5 always are hotter then any other. >>> This can be something with scheduler, however this started before any >>> actual load. Though numbers are normal I had never seen something >>> alike... >> >> Two cores per socket, and 8 sockets on the board? If so, that looks >> absolutely fine to me. The average temperature is 36.8C but 43.0C is >> still well within spec. That difference of just over 6 degrees is not >> really significant and probably entirely due to different airflow >> patterns over the different CPU sockets. If you swap the CPU package in >> that socket with one of the other ones, you'll find the hot spot stays >> put. You might be able to even things out by rerouteing cables, but >> really it's not worth the hassle and won't make any perceptible >> difference to performance. > > Nope: > > CPU: Intel(R) Xeon(R) CPU E5620 @ 2.40GHz (2394.05-MHz > K8-class CPU) > FreeBSD/SMP: Multiprocessor System Detected: 16 CPUs > FreeBSD/SMP: 2 package(s) x 4 core(s) x 2 SMT threads > > So the difference is about one physical core with two SMT threads. >
Which explains why the numbers go in pairs -- there's only 8 physical cores. Even so, I don't think there's any great problem there. Different cores in the same package can have different temperatures -- that's perfectly normal, and due to the physical properties of the CPU package and the local environment rather than any difference in processing load between cores. Cheers, Matthew -- Dr Matthew J Seaman MA, D.Phil. PGP: http://www.infracaninophile.co.uk/pgpkey
signature.asc
Description: OpenPGP digital signature