Hi All,
A few weeks back i ordered a brand new system to run as a server here in our office.
Hardware specs:
I7-14700K
Noctua NH-U9S
128GB DDR5 memory
Supermicro X13SAE-F motherboard
Right now, i have 5 containers and 4 VM's running, and them the motherboard started to beep every now and then... After some investigation i found out that the motherboard detected CPU temperature, and started to beep as a warning. That worried me, since my avg CPU usage is below 10%.
So i installed glances, to easily monitor the temperatures and find out what was wrong.
There i noticed that the average temperature was around 65C, but every now and then there are temperature spikes, up to 90C. The spikes are usualy very short. Glances updates every 2sec; e.g. temp is steady 60C, 2sec later shows 90C, 2sec later shows 60c again.
For a test, i stole the CPU cooler from an gaming PC, which is twice the size of the Noctua, and i bought an thermal pad, to make sure i did not f*ck up with the cooling paste.
This did improve things, average temperature dropped to around 50 degrees. But still i'm seeing the same spikes, still up to 90C.
Glances also shows an overview with the tasks incl cpu usage per task. There i did notice a few things;
I have a win10 vm running. I initialy gave it 2 sockets, 4 cores, and 16gb memory.
the VM runs smoothly, and in glaces i see an KVM task with an average CPU usage of 30%, but with very high spikes sometimes around 150%. During startup of the VM the CPU usage of that task goes up to 350%.
I gave it 4 sockets with 4 cores and 32gb memory, but still the same CPU usage...
I'm not entirely sure, but i think this maybe the cause of the temperature spikes?
In PVE on the summery page, the vm cpu usage never exceeds 10%, not even during startup.
I also noticed this behavior on other vm's and containers sometimes.
So my question;
the high cpu usage on some tasks as shown in glaces, can that have something to do with how i set up the sockets vs cores etc?
Why does increasing the nr of sockets/cores lower the total CPU usage in the summary page, but glaces still shows high cpu spikes on the KVM task, and high temp spikes?
Hoping for some advice, thanks in advance!
A few weeks back i ordered a brand new system to run as a server here in our office.
Hardware specs:
I7-14700K
Noctua NH-U9S
128GB DDR5 memory
Supermicro X13SAE-F motherboard
Right now, i have 5 containers and 4 VM's running, and them the motherboard started to beep every now and then... After some investigation i found out that the motherboard detected CPU temperature, and started to beep as a warning. That worried me, since my avg CPU usage is below 10%.
So i installed glances, to easily monitor the temperatures and find out what was wrong.
There i noticed that the average temperature was around 65C, but every now and then there are temperature spikes, up to 90C. The spikes are usualy very short. Glances updates every 2sec; e.g. temp is steady 60C, 2sec later shows 90C, 2sec later shows 60c again.
For a test, i stole the CPU cooler from an gaming PC, which is twice the size of the Noctua, and i bought an thermal pad, to make sure i did not f*ck up with the cooling paste.
This did improve things, average temperature dropped to around 50 degrees. But still i'm seeing the same spikes, still up to 90C.
Glances also shows an overview with the tasks incl cpu usage per task. There i did notice a few things;
I have a win10 vm running. I initialy gave it 2 sockets, 4 cores, and 16gb memory.
the VM runs smoothly, and in glaces i see an KVM task with an average CPU usage of 30%, but with very high spikes sometimes around 150%. During startup of the VM the CPU usage of that task goes up to 350%.
I gave it 4 sockets with 4 cores and 32gb memory, but still the same CPU usage...
I'm not entirely sure, but i think this maybe the cause of the temperature spikes?
In PVE on the summery page, the vm cpu usage never exceeds 10%, not even during startup.
I also noticed this behavior on other vm's and containers sometimes.
So my question;
the high cpu usage on some tasks as shown in glaces, can that have something to do with how i set up the sockets vs cores etc?
Why does increasing the nr of sockets/cores lower the total CPU usage in the summary page, but glaces still shows high cpu spikes on the KVM task, and high temp spikes?
Hoping for some advice, thanks in advance!
