Correcting parity check finished with errors for the second time

[gelmi]

Due to some hard resets while GPU passing through to VM, standard parity check after booting showed parity errors. I run correcting parity check 2x and every time it showed errors. 

SMART tests on both HDD disks (parity and data) show no errors. I have Asus X370 Pro board where all SATA are going through X370 chipset (no Marvel controller). Attaching my diagnostic logs.

Any ideas how to fix these parity errors?

 

darktower-diagnostics-20180222-1643.zip

[JorgeB]

Run a memtest.

[gelmi]

Do you mean a memtest from EFI booting menu (UNRAID, UNRAID GUI, SAFEMODE, MEMTEST)?

 

[JorgeB]

From the boot menu yes, but you need to use legacy booting memtest doesn't run in UEFI.

[gelmi]

My Flash configuration is to force EFI boot. Memtest in Mutli-threaded mode started spilling errors just after 5 minutes. I recently updated the BIOS and that was the first time I could set 3200 MHz on RAM and boot. I have returned to 2933 and started Memtest in Mutli-threaded mode again. No errors so far. I will let it run for the night and if there are no errors, I will do parity check once more. Does this sound like a plan?

 

[S80_UK]

According to Asus the motherboard rates 2933 and 3200 for RAM are regarded as an overclock - the official nominal rating is 2600MHz - https://www.asus.com/uk/Motherboards/PRIME-X370-PRO/specifications/

For the CPU, AMD also rate the memory interface at 2667MHz - https://www.amd.com/en/products/cpu/amd-ryzen-5-1600

 

Normally there is relatively little to gain from faster RAM clocks since the number of cycles required to access data is generally increased at the higher clock speeds.  

 

You should treat Memtest as a guide.  It tells you that the system runs or fails Memtest at a particular setting.  But even if everything passes, there can still be potential issues with other software, especilly if overclocking. 

Edited February 22, 2018 by S80_UK

[SSD]

@gelmi -

 

There is a self booting USB you can create for the Passmark memory tester. Works with UEFI and also a better test. 

[gelmi]

OK. But after changing to 2933, 12h OK Memtest and 0 errors on second correcting parity check run, would it be safe to assume that 3200 OC was causing the problem?

[SSD]

I'd run the memory test for 24 hours, but yes, this seems very promising.

[JorgeB]

5 minutes ago, gelmi said:

would it be safe to assume that 3200 OC was causing the problem?

Probably, and also safe to assume that a storage server should never be overclocked.

[gelmi]

Great, will update thread when finished.

[SSD]

14 minutes ago, johnnie.black said:

Probably, and also safe to assume that a storage server should never be overclocked.

 

Overclocking is a complex subject and the risks of these types of issues greatly increased if you are pushing the hardware beyond is rated minutes. It is strongly discouraged and could lead to data corruption.

[gelmi]

Understood. I am in process of assembling only NAS with dockers machine to store my data with ECC RAM, but for now I have to share a PC between NAS and workstations (VM). In the mean time, probably I will switch to 2667 MHz just to be on the safe side. Thanks.

[gelmi]

Little update. I was able to pull off 24h RAM test without errors (Mutli-threaded mode) for 2667MHz and 2933MHz. After that I started correction parity check on 2933MHz 3 times. First one found and corrected ~300 errors and following two found 0 errors. I think I will stay on 2933MHz and not to push for 3200MHz with next BIOS updates until I move my array disks into the new storage box.

Thanks for all your help.