PowerEdge R250 CPU clock lock at .20 GHz

CwH4225,

Would you confirm if you are seeing any errors in the logs, and if so what are the specific errors?

Also, would you power down the server, remove the power cables, then hold the power button down to drain the flea power. Once that is done would you check the power supply cables, specifically the one labeled P3/MB and verify it is connected correctly, with the retention clip facing inwards and locked to the plastic notch sticking out? If it isn't, would you reverse that connection and then retest it?

Let me know what you see and if this helps.

@DELL-Chris H​ 

Not showing any errors in the logs. Also not seeing any cables labeled P3/MB. I checked all the connections to make sure the retention clips are in the correct direction locked to the notches.

@DELL-Chris H​ 

Also a simple restart doesn't resolve the issue. The server has to be completely powered down and powered back up for the cpu clock speeds to go back to normal. Not sure if that helps narrow anything down. 

It should be located as seen below. 

When you locate it, can you verify it is facing the correct way, that there isn't any damage or bent pins, and also try reseating the connection? 

Lastly, is the server completely up to date, I know you stated the BIOS is, but is the rest of the server?

@DELL-Chris H​ 

ok I will look for those connections. Windows server 2022 is up to date, and I just seen there was another BIOS update released on the 13th of this month so I will update that also. Is there any other updated I should be looking for?

I would also consider updating CPLD, iDrac, raid controller, etc. The iDrac in particular, as it cant be too far from the BIOS or you can start having communication issues witht he two. 

@DELL-Chris H​ 

Ok found the connectors and they are connected correctly. Pins looked fine. Updated BIOS to 1.9.1, updated iDrac to 7.10.30.00, CPLD and raid controller were up to date. Simulated a power outage by unplugging the battery backup unit. Server stayed powered, but cpu clocks again locked at .20 GHz.  

Thank you for confirming that for me. The next thing I would suggest testing would be if the issue occurs outside the OS. I suggest you download the Dell Support Live image here. You will then want to boot to that and repeat the test, then let me know what you see. 

@DELL-Chris H​ 

Ok I got a chance to do some more testing on it today. I haven't tried the dell OS iso but I did boot windows server 2022 into safe mode and simulated the power outage. In safe mode the CPU clocks do not down clock and it acts perfectly normal. 

Glad to hear you isolated it. What I would look at is the system profile setting, specifically if it is set to OS managed, and not max performance.

@DELL-Chris H​ 

Well disregard that. Not sure why but in safe mode the clock speeds were not showing accurate after the simulated power outage. Ran a restart and the CPU was indeed locked at .20 ghz after it restarted out of safe mode. One thing I noticed is when the CPU is down clocked and locked at .20, if I just run a reboot, the entire start up is super slow. Even the initial DellEMC startup screen where it's loading the firmware interfaces, driver, ect. This leads me to believe it may not be OS related?

I would still test it outside the OS, as well as you may want to try it off of the UPS, just to eliminate them as variables. 

@DELL-Chris H​ Sorry for the late reply, just now able to do more diagnosing on the issue. I eliminated it being an OS issue as the clocks still lock at .20 on the dell .iso image. I did take the server into our main server room where we have high end pure sine wave APC backup units. With the server hooked to the APC unit the clock do not downclock at all with a simulated power outage. I then tested the server on a brand new out of the box Cyberpower 1500AVR backup unit and the issue showed back up. The cyberpower unit uses simulated sine wave instead of pure sine wave so only thing I can come to is the the R250 server isn't seeing stable output from the cyberpower units during a power outage and is dropping the clocks. If this is the case I would still assume that when power was restored and output is stabilized it should still pick the clock speeds back up but its not?

CwH4225,

I would think it would recover the clock speed after recovery, but unfortunately I have not come across this issue before, short of the cable issue previously discussed. What I would suggest is accessing the BIOS, and see if the issue still occurs when you change out settings within the Processor Settings page, such as Logical Processor, VT, etc.

Let me know what you see.

Hello

i'm taking the liberty of intervening in this discussion, as i'm seeing almost the same problem.
3 months ago, I set up a cluster of 3 news servers, 2 r250s and a t430.
they're up to date on the bios/idrac & co side
the 3 are connected to an APC UPS "Back-UPS RS 1200G".
every 2 weeks I noticed a sudden increase in the cpu load of the 2 r250s, which was not resolved by a reboot, I had to completely switch off the servers for the problem to disappear.

at first I thought it was a problem with a VM (it's a virtualization cluster under proxmox), but no.
then i looked at the proxmox side, everything was normal.
I finally noticed yesterday that this problem appears every time my UPS management software (apcupsd) runs a self-test on the ups.

so obviously it's not an OS problem (since I'm on linux), nor a hardware problem specific to the one CwH4225 has.
there's a problem with the r250s that seems to occur when there's a micro-power cut (this is probably what happens when the UPS self-tests).

my 2 r 250 are affected, but not the t430.

Regards