So I just picked up a few Poweredge 2950 servers and I have them set up with Windows Server 2012. I installed Core Temp to monitor CPU temperature. One server has a CPU operating temp of the following according to CoreTemp:
CPU 0:
47C
47C
41C
41C
CPU 1:
57C
51C
54C
51C
In OMSA I see a System Board Ambient Temp of 23C and the fans are somewhere around 6900RPM, which leads me to believe everything is OK… else the board would ramp up the fans. I have the servers in a 42u rack and the room they are in is roughly 65F( 18C ). So I'm not sure if I should be concerned.
The reason I am asking is because the specs I found on the e5430 Harpertown say that the maximum operating temp is 67C … however CoreTemp says the max is 100C… My question is: Are these readings normal? I'm concerned that I'm reaching the top end of the CPU temp.
Best Answer
I've been confused by different temperature terms myself, and to be honest I'm still not always sure how to interpret such information. What I do know is that TCASE is not the same as core temperature. Core temperature is also called Tjunction or Tj, and even that isn't really one sensor but the highest value from multiple sensors inside the core placed on "hotspots". Here are some in-depth explanation for Core 2 Duo CPU's.
Intel's definition of TCASE is:
The problem is that there's no sensor actually measuring TCASE, and the best one can hope for without placing a sensor there is an approximation. Most motherboards have a "CPU temperature" reading, and from what I understand that is usually a sensor placed in the CPU socket combined with an estimation calculation trying to find TCASE. Their accuracy are highly questionable though. You can find a discussion about the topic here and further information on google.
The details of all this is for the more than normally enthusiastic, the important thing to understand is that you're comparing apples and oranges. Core temperatures will be higher than TCASE, under load significantly. Core Temp refers to Tj Max, that is maximum allowed core temp Tj, even though I wouldn't be comfortable bringing my cores close to this value.
Software that report CPU temperatures seldom specify which values they actually present. My rule of thumb is that if you're presented with one reading it's most likely some TCASE estimation, but if you're presented with the same number of readings as cores, it's probably core temp/Tj's.
As a side note, I have a PowerEdge 2950 myself, and from what I've been able to figure out, the BMC doesn't throttle the fans in relation to CPU temperatures, but rather from some calculation based on what devices are installed, and maybe ambient temperature. I've done stress tests of the CPU that raised the core temps, without seeing the slightest change in fan speed, while mounting a SCSI card connected to nothing raised the fan speed notably.
I don't think there's much reason for concern, depending on how much load that produced those temperatures. I would clean and reapply thermal grease between the CPU and heatsink on CPU 1 if it stays consistently ~10 degrees above CPU 0.