RAM disappearing on Gigabyte WRX80 (2 months spent on this!)

Hello! Thank you for taking the time to read this! I have been working on this new build for 2 months now but just keep running into issue after issue. This is meant to be an upgrade to my existing Unraid build.

Current problem: I have 2 issues I currently see.

  1. When I finally get booted into Unraid, after 1-3 hours one of my RAM sticks disappears, dropping the capacity from 128gb to 96gb. This is causing all sorts of issues with stability.

  2. After turning the system on from a state of no power (for an extended period of time), the boot sequence gets stuck on the “GIGABYTE” logo screen with the code “A0” in the bottom right. I can get past this screen if I remove the installed GPU. After a full boot without the GPU and can reinstall it and boot fine. I am trying to fix this to avoid plugging my GPU too many times.

Parts list:

CPU: AMD Ryzen Threadripper PRO 3955WX (Bought New)
Motherboard: GIGABYTE GA-WRX80-SU8-IPMI (Bought Used)
Ram: 128GB 4x32GB DDR4-3200 PC4-25600 2Rx8 ECC Unbuffered Server Memory by NEMIX (Bought New)
CPU Cooler: Noctua NH-U12S TR4-SP3 (Bought New)
PSU: CORSAIR HX Series, HX850, 850 Watt, 80+ Platinum Certified, Fully Modular Power Supply (Reused from old build where it was bought new)

I am at wits end with this one. Anything that can be done to help would be greatly appreciated!

This is a continuation from another thread where I got earlier problems resolved: [Gigabyte ga-wrx80-su8-ipmi not posting]

Could be a bad seat, have you tried reseating the CPU and RAM
Could be it has a slightly bad seat and it has to warm up before the pins mush correctly

1 Like

Thanks for the idea! I have reseated the RAM many times as I initially thought the slot population pattern was the issue. The CPU has been reseated twice already. Would there be any way of knowing if this would be worth the hassle of dismantling? I have been diligent in following the installation guide video posted by AMD.

Also, this may be a noob question but why would the CPU heating up and then fitting properly cause hardware to dropoff? Shouldn’t it be the other way around?

unlikely expansion and contraction are taken into account when they design the pcb and heat spreader glue.

more likely the ram is training and one stick is failing as it cant match the other rams timings or is throwing to many errors that are being corrected.
in this case try booting with less ram installed. (1-2 sticks) with the gpu installed, let the ram train and see what happens.

also try running the ram at stock no xmp enabled. 2133. if you can boot with all the ram showing, try 2933 and if that works then its likely the fastest that ram and cpu will run.

A0 on a gigabyte appears to reference ide initialisation. so may be looking for an ide hard drive to boot from and failing when it cant find it.
do you have any ide devices attached?
if not disable ide support in eufi/bios (switch to ahci). and or disable csm support.

Use ECC Registered DIMMs (RDIMMs)

I use 256GB (8x32GB) DDR4-3200 PC4-25600 2Rx8 ECC Registered DIMMs with no issues on my WRX80. They’re DDR4-3200 native, so XMP does not apply.

Preferably use a known manufacturer like Samsung or Micron. I don’t know about Nemix.
Mines are genuine Samsung DIMMs.
Use Samsung’s DIMM search utility to find exact part numbers:

Mine is M393A4G43AB3-CWE

Thank you all for the ideas!

I double-checked the bios and couldn’t find any RAM tuning options. I would imagine this is due to using “server” RAM and a “server/workstation” board.

I think Kish might be on to something. I ordered Samsung ECC Registered DIMMs which are scheduled to arrive 11/30-12/2. Hopefully, that is it. I’ll post a follow-up either way once those arrive.

Ram: 128GB 4x32GB DDR4-3200 PC4-25600 2Rx8 ECC Unbuffered Server Memory by NEMIX (Bought New)

I can’t say this is for certain your issue, but for me Nemix brand RAM has been very hit and miss. Back when I had DDR3-era hardware I had a two sets of RAM from them that were totally problem free. But I just rebuilt my virtualization servers and the original kit of RAM that I purchased was what I believe is the exact same kit that you have and I had to return it for throwing errors in Memtest86. I also returned a set of DDR4 non-ECC RAM from them in a past build for the same reason. Their price point is attractive but at least in my experience it feels like a matter of luck whether they are any good or not, moreso than with the big name DRAM manufacturers.

About Nemix, you can add me to the “not really sure” column, as built a couple of boxes with Nemix DDR4, and get occasional faults.

Update: The new RAM arrived and has been installed. From the first couple days of use, the RAM issues appear to be resolved. I’m now getting a PCIe error but this is at least progress.

For anyone searching this in the future, the RAM I order was: Samsung M393A4K40DB3-CWE 32GB 4Gx72 ECC & REG Server Memory