Likely causes of memory failure on z840 when memory is good

[mwx]

I am trying to understand why my z840 had a memory failure 5 beeps. I tried to find bad module but they were either all bad or some other explanation and maybe all good. I then tried modules from previous install(all good at replacement) and got the same result. The only other change over the years was adding an evga  nvidia 2060 (195 watts peak) recently with a HP 6 to 8pin adapter. It had worked flawlessly for three months.The  z840 had no other cards installed, but had 16 modules of 16gb modules and 2x Xeon E5-2680 v3. The 1125 PS comes up with fans. Am I pushing system limits and such is likely to happen again or is this a reasonable configuration that should be reliable?

[DGroves]

time after time people complain about memory issues on systems saying they have the correct memory installed when in reality they either have wrong ram, or a improper dimm load order of different capacity modules

first of all, for the "Z" series workstations (old and new models) ram is not ram,.............. in other words a 16gb single rank module is not the same as a dual rank 16gb module

a buffered ECC module is not the same as a unbuffered ECC module

a load reduced (LR) dimm is not compatible with normal dimms

you can not mix single/dual/quad rank dimms within a memory bank, you can however have a memory bank filled with single rank and a different bank filled with dual or quad rank

each cpu (if  it's a dual cpu board) has its own dimm sockets so placing ram in the sec cpu dimm sockets with only one cpu will not work

each cpu needs to have the same dimm's, order/size installed. mismatched configurations will cause issues

i recommend most people go to kingston or crucial websites to make sure the ram they are buying is compatible

and that they also look at the HP recomended ram setups for their workstation model

for the z840 go here:

https://www.bhphotovideo.com/lit_files/113010.pdf

to check power related issues do a "BIST" test on the power supply this is covered in the z840 service manual or simply google z840 BIST test

the video card and the cpu's are within the HP specs for power draw however i have seen cases where smoking or other air related issues can cause cause the dimm sockets to have intermittent contact with the installed dimms using a cleaner made for cleaning the dimm contacts can be used if necessary

https://www.amazon.com/G5S-6-Enhancer-Protector-Surfaces-Integrated/dp/B003A7KHK8

[mwx]

For over a year the z840 worked fine with 16 identical HPE MTA36ASF2G72PZ-2G 2Rx4 filling all memory banks.

System memory fails with Processor 0 only, and reversed with Processor 1 used as Processor 0.

More testing when I get another for test. 

One suspect is using a Tripp Lite 14 gauge plug which always seemed a little loose at the unit. Much tighter now when I used a 16 gauge cord.

[DGroves]

on the z840 the memory controller is part of the CPU, so if swapping cpu's causes the memory error to also swap to that cpu's   dimm sockets then the CPU is most likely at fault

i would visit the intel site an dl their cpu test program

https://www.intel.com/content/www/us/en/download/15951/intel-processor-diagnostic-tool.html

[mwx]

Tested some more and have found cpu1 in proc0 without anything in proc1 boots. Will test another cpu soon, but it does not boot with cpu0 in proc 0 or proc1 to use software test. Besides cpu0 being bad, there could be something wrong in proc1 part of motherboard.

[DGroves]

are both cpu's PRODUCTION "SPEC" revisions? non production cpu's will have issues in these HP workstations

is the cpu model a listed HP approved model?

while some OEM specific models will work, this does not mean all cpu's from the E5-26xx v4 line will wok

[mwx · Accepted solution]

Bent pins on the LGA socket proc0 was primary cause as well as minor bent pins in proc1. I never removed from socket until failure. Bought a used z840, two E5-2280 V4 and a motherboard. Confirmed bent pins the culprit on original as memory all good and all E5-2280's V3 and V4.  Fell off turnip truck trying to get Wake-On-Lan working as I did not know HP has BIOS branding for configuration, and such branding will alter BIOS options displayed, according to support. Soon I will know if that is true. Before support expected me to see a WOL option in BIOS, and my unbranded 2.59 on both systems did not.