Score:0

High cpu temperature and temperature difference

ph flag

On a cluster I am working on there is a node which is showing high CPU temperature.

The node has 2 Intel(R) Xeon(R) CPU E5-2680 v4 @ 2.40GHz.

The sensors command from lm-sensors is showing that one CPU is at around 70°C and the other at 90°C. The load is 100%. It is in fact overloaded but the load can not be reduced. The temperature is highly correlated with the load. The current frequency is higher than the max frequency. max : 2400000 cur: 5280000 So I do not think that there is throttling.

Is the temperature diffrence a sign of cooling issues ?

The intel documentation is showing that the temperature case is 86°C from what I understand it means that the lifespan of the CPU at 90°C will decrease.

It is almost a week with these temperatures, should I look into solution (reduce CPU speed) to decrease the temperature of the CPU ? The node will probably run other intensive CPU jobs in the future.

Score:0
cn flag

Running a CPU at those temperatures is within specs, but it will most likely degrade the longevity of your components. You should definitely look into scaling up, both horizontally and vertically, to reduce the load. If on-premise you could also check if there are more efficient cooling options.

Thomas avatar
ph flag
Do you have any idea of why there is a 20°C gap between CPUs ?
Erik Norman avatar
cn flag
There could be several reasons for that. It could be that the internal load balancing favors one CPU over another (need to verify this), the other could be that the thermal paste on the hotter CPU wasn't applied correctly or has degraded over time.
Erik Norman avatar
cn flag
If you find the reason, I'd be curious to know what it was.
Thomas avatar
ph flag
In fact I already checked if the load was the same on the CPUs and it is the same. So I was already thinking about the thermal paste. I will try to apply new thermal paste on the hotter CPU.
Erik Norman avatar
cn flag
I use Conductonaut by thermal grizzly (unless you have an aluminimium contact area between CPU and heatsink). It really makes a difference.
mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.