Dell R730XD: CPU temps are different, seems to be a socket issue

[phil1c]

Just interested if anyone has seen similar or what else I can try: I have an ebay r730xd that I'm migrating my current setup into. The CPU temp between CPU1 and CPU2 is consistently different, on early boot maybe only 4C difference but usually 7-10C different and trending higher depending on system load. I've seen this behavior using Ubuntu and Unraid as the OS (Unraid is the permanent OS, Ubuntu was for early testing). I've tried the following to isolate the issue:

1. Swapped CPUs between sockets.
2. Swapped coolers between sockets.
3. Repeated reapplications of thermal paste on both CPUs.
4. Inspected and juggled fans around.
5. Inspected for any blockages in airflow.

In all tests there was no improvement: CPU1 remained the hotter CPU and the temperate delta between the two CPUs didn't improve. At this point, I'm left thinking there is something wrong with the socket itself.

Has anyone seen this before and/or has any other ideas of what else to troubleshoot? I don't want to get a new motherboard if I don't have to but the temp of CPU1 triggers the fans to run too loud for apartment use but the temp of CPU2 would result in acceptable fan noise so I'll take any advice.

Thanks

 

[i386]

or has any other ideas of what else to troubleshoot?

Are you using the air shroud? (google shows pictures of that system with airshrouds)
Are all fans present? (google pictures has systems without fans #5& #6)
Are all drive bay populated with ssds/hdds/dummies?

 

[fp64]

Had same problem with HP ML350 G9. I made it worse by adding 3x MI50 and 1x MI60 with 1.1KW external psu. Solved it by using (mostly) duct tape.


Applies to HP ML350 G9:

1) Install full complement of fans
2) Go to bios and change cooling from 'minimal' to 'adequate' ie one level up => noise
3) Get duct tape.
4) Plug any openings due to missing mount screws or other externals
5) side cover leaks air just about everywhere. Use duct tape to plug the lot - yes it is ugly.
6) rear panel: use your hand for sensing where the exiting air is @ ambient temp. Block these using duct tape mercilessly to redirect the airflow.
7) Looks terrible. Will not impress your friends about yr tech sophistication, but works ok.

With ambient temp 23C, the temp diff between cpus is 6C (was 10C). GPU max temps went from 80 to 56.

 

[phil1c]

Are you using the air shroud? (google shows pictures of that system with airshrouds)
Are all fans present? (google pictures has systems without fans #5& #6)
Are all drive bay populated with ssds/hdds/dummies?

I have the 730xd with the 4-drive midplane so no air shroud as the tray takes the place of the shroud. Also, I am using the low profile CPU coolers for the same reason.
Yes: all fans are present and functional
Yes: all bays are populated with drives (10/12 bays up front) or trays (not blanks).

Had same problem with HP ML350 G9. I made it worse by adding 3x MI50 and 1x MI60 with 1.1KW external psu. Solved it by using (mostly) duct tape.

Applies to HP ML350 G9:

1) Install full complement of fans
2) Go to bios and change cooling from 'minimal' to 'adequate' ie one level up => noise
3) Get duct tape.
4) Plug any openings due to missing mount screws or other externals
5) side cover leaks air just about everywhere. Use duct tape to plug the lot - yes it is ugly.
6) rear panel: use your hand for sensing where the exiting air is @ ambient temp. Block these using duct tape mercilessly to redirect the airflow.
7) Looks terrible. Will not impress your friends about yr tech sophistication, but works ok.

With ambient temp 23C, the temp diff between cpus is 6C (was 10C). GPU max temps went from 80 to 56.

I think I'll give this a shot as well and see what happens. Good suggestions.

 

[phil1c]

In case any other panicked dell server noob shows up here with similar questions: even something as seemingly small as equal distribution of harddrives can impact temp. I only had 10 drives and filled the bottom two rows and top left two bays - right in front of CPU1. I moved the second from the left drive to the top-far-right bay and now the temps are balanced.

 

[Hrast]

even something as seemingly small as equal distribution of harddrives can impact temp.

Just to re-enforce the lesson, I learned the exact same thing after moving disks around when I got my r730xd as well.