Suspect z820 Thermal Sensor causing one of two CPUS to misbehave

[TeaspoonVFX]

I'm running a dual z820 which has now become very loud.

I've run both HWMonitor and HP Performance Advisor to get diagnostics on the system, and found out some interesting things:

1. The system thinks that CPU 1 is running at 92 degrees Celcius. And CPU 0 is averaging about 35 C.  When I open the case and check it out.  CPU 1 is actually cold to the touch.
2. CPU 1's fan is running at about 4900, while CPU 0's fan is strolling along at 1100.
3. CPU 1's clock speed is locked to 1.1Ghz.  CPU 0 is running at its proper 3.5 GHZ

So, my conclusion is that the system is misreading the temperature of the CPU, revving up the CPU 1 fan to hyperspeed, and ratcheting down the CPU clock speed in an attempt to protect it from overheating.

Oh, and if I do run something that pushes the system to try and use all the procs on both processors, the system shuts down.

All in all, I currently have a loud workstation, barely functioning and responding at half speed, and intermittently shutting down -- which is unproductive and annoying.

So -- if my assessment is sound, my question is:  where do I look to find the thermostat for CPU 1 so I can either replace it, or perhaps the cable simply needs to be re-seating.

Any insight would be fantastic.

[MSLEEE]

The CPU itself include the thermal sensor. Please check the followings:

1. First of all, update the System BIOS to v3.88.

2. Reset the CMOS.

3. Check if the thermal grease is put properly between processor and heat sink.

4. Check if all of fans are running fine.

5. When you run the CPU intensive progranm, check if the second CPU's fan speed and temperature. Aloso check the ambient temperature.

6. If you add any add-on cards from the original configuration, please remove them for testing purpose.

7. If the above does not help, then call HP Support.

The system shus down itself. It means that there is a thermal issue with your system. If there is an ambient temperature issue, the IO cable should be replaced. If there is any fan (rear/front/memory/CPU) running fast or no running, the fan may be bad. If you do not use the liquid cooling unit and thermal grease is put properly, the processor may be bad.

[TeaspoonVFX]

Thanks for the reply:

1)  BIOS is up to date

2) I've reset the CMOS

3) In order to check the thermal grease, wouldn't that necessitate taking the heat sink off the CPU, which in turn, would require putting on new thermal paste?  I can definitely do that, but, I don't think the CPU is ACTUALLY overheating.  I think the system just THINKS its overheating.

4)  All fans are not running fine... the CPU fan on the "overheating" CPU is running at 4x the speed of the other CPU...despite the fact that the other CPU is physically hotter.

5) It there a monitor for checking the ambient temperature, or are we talking about a physical thermometer placed inside the case?

6)  No additional add-on cards. 

7)  I may have to buckle down and get through to HP.  The box is 5 months out of warranty, so we'll see what happens.

THanks for taking the time to answer!

[MSLEEE]

Waht is the second CPU's temperature when you run the PA? The PA shows the System Ambient Temperature as well as CPU temperature and fans' speeds.

[TeaspoonVFX]

I've run into a problem where PA (I assume that is Performance Advisor), it throwing a .NET error when I run the Workstation Performance.  Troubleshooting that at the moment.

But here are the results from last night.

so from here, you can kind of see my predicament.

Ambient Temp is a cool 23C

CPU0 is a mild 54C

and CPU1 is a scorching 92C  (which isn't the case on physical examination.)

CPU 0s fan is 795 RPMs

and CPU1 is screaming at 4195 RPMs

Finally, if you look closely, the odd numbered cores (i.e. on CPU 1) are barely clocking.  They might spike here and there, but there average is 0-5%.

[MSLEEE]

My recommendation is to remove the second heat sink and put new thermal greases. Then check next step.

[TeaspoonVFX]

Just to be clear...are you saying to remove the heatsink on CPU1?  The one that is misreading at 92C?  Or remove the heatsink on CPU0 -- the working one?

[MSLEEE]

I mean to remove the heatsink on CPU1 that misleading at 92C. If possible, please swap the heatsinks on CPU 0 and CPU 1 as well.

[TeaspoonVFX]

Heya, 

Based on a conversation with HP tech support, I ordered and replaced the Front IO Assembly -- which generally contains the thermostat.  But to no avail.  (It did fix a USB port problem though)

So, I did what you recommended and swapped the heatsinks.  In the process, I noticed that they are liquid cooling units.  Once swapped, the problematic temperature reading is on the other CPU.

A replacement Liquid Cooling Unit is ordered and should be here tomorrow or Monday.

I will post with the result.

[MSLEEE]

As long as I understamd, the fron IO cable includes the thermal sensor but it's for the ambinet temperature. The processor itself includes the thermal sensor inside of the processor package. The probability of the problem is like this sequence " CPU heatsink contacts > CPU heatsink (if it's air heatsink, it's very low possibility the problem is caused by heatsink. If it's fan or liquid cooling heatsink, it's suspected > Processor > System board  > others (BIOS..)".