Using non-matched CPU's?

[ember1205]

What happens if you install two CPU's that are the 'same' but are not necessarily of matched stepping? What sort of issues might occur? What might the symptoms be?

I'm wondering this because it seems like there are an awful lot of "lots" for stepping within certain CPU models. 

[DGroves]

as far as i know, for the z800 HP policy is that cpu' must match

issues for mismatched steppings can range from no boot to subtle things like random crashes

[ember1205]

Thanks for the response. I'm more focused on technical issues that could arise as opposed to being "out of policy" as I've had a machine that was actually exhibiting random hard crashes (machine would just straight-up power off with no warning) and then would often fail to boot afterward. I actually wasn't aware that the CPU's in this machine were not stepping-matched until very recently and am wondering if installing a matched set of CPU's would resolve the issues I had previously since I have swapped every single part in this machine that I could and didn't see anything resolve it. Since I don't currently HAVE a set of matched CPU's, I would need to get a set of them to swap them out.

[DGroves]

there is a test that can be done to rule out the hardware of most crash issues

1 remove cpu #2 and sec cpu ram,  and test/run the system with only single cpu, does it crash? if not sec cpu or sec ram banks is cause

2. swap in sec cpu in place of first cpu, again run/test system using various cpu/computer stress apps

is system still stable? if so, swap in sec cpu ram and retest, stable?  issue is with sec cpu socket or mismatched cpu's

3. install matched cpu's (speed does not matter use cheapest ones you can buy) does system crash? if so issue is most likely with motherboard or bad ram slots in sec mem bank

[ember1205]

Thanks, but the tests aren't overly feasible for me to get definitive information from. The primary issue I have with any sort of testing like this is that I do not have any correlation between the crash and anything that has led up to it. The last time the system hard-crashed, it had been running for about six months without any issue.

The other piece of this is that I have read lots of community guidance that says you need matched stepping, but I have also found comments that claim Intel doesn't require that. In fact, with mis-matched stepping, the CPU's supposedly drop down to a lower level of functionality supported by both. I can't find that article, though, so that's why I'm looking for something referenceable that will guide me through how to get this squared away.

[DGroves]

if your crashes are 6 months or more apart then hardware issues are not the cause. i would look at software related issues or things like the power grid have you ever looked at the windows crash logs and crash dump files?

[ember1205]

This is one of two systems that was originally running XenServer. I would migrate workloads onto and off of this machine. Made no difference as to preventing crashes or not.

It now runs ESXi with newly built workloads (and only a couple of them). Crashes persist. So, with the exception of both base operating systems being hypervisors, there is nothing common for software across the two.

It has always been connected to a UPS that provided filtered power (just like the other) and has very recently been changed to a SIN wave output UPS. One crash did occur within about two days of a very brief power blip that had zero impact on the other machine.

[ember1205]

Further details...

It seems that the actual stepping of the two CPU's in question is actually the same. Using "cat /proc/cpuinfo" from a booted linux kernel shows they are all the same microcode and stepping value.