Gigabyte boards with ecc ram support

Hello,

I wanted to give my home/office server some ECC memory. Gigabyte states it supported and tested so with this in mind i bought a nice dimm to start testing.
When putting in the dimm the motherboard sees it as ECC, Truenas recognizes ECC.
After reading much about ECC memory support on AMD i also wanted to test if it actually posts the errors.
I changed timings on the memory and i can see these timings being used in Memtest. However, after testing for multiple hours on different timings it will boot on i am not able to see errors within Memtest.

I wanted to check if i doing everything correct, is my ram correct, do i have to extend my tweakings or are there other things i could test?
I only changed CL memory timings from 22 to 16, it will not boot if putt lower.
When i set a higher mhz for the memory it will not use it. Memtest does not see a higher MTs (at the moment i am unsure why).
It will not boot if i change the mhz to much, so it does seem to do somethiung, memtest is just not able to see it. It does see the setted timings.

I know there is a way to short circuits by connecting some pins but that seems like a step to far from me. I hope to get errors by changing settings.
Maybe there are people experienced with this exact motherboard or other Gigabyte Motherboards and tested this? I would love to hear some specific Gigabyte results.

Hardware i am using:
Motherboard: Gigabyte X570S UD
Memory: Kingston KSM32ED8/32HC (32GB 2Rx8 4G x 72-Bit PC4-3200 CL22 288-Pin DIMM)
CPU: AMD Ryzen 5 PRO 5650GE

Other motherboards i have and could test this on:
Gigabyte B550M AORUS PRO-P
TUF GAMING B550M-PLUS WIFI II

Would these have a higher change of posting errors, does anybody know?
Good to get some info aobut this before i rebuild all pc’s only to test ecc.

I really hope to get ECC working. Any advice is welcome, i already read every forum post about ECC on AM4 here, this is just a desktop board and nothing really server related to this so i guess i will have to settle with the fact it just won’t post errors.

I had this post all filled with links to the products but i could not post this, sorry for having half info in my post.

The best answer Wendell’s videos show is Linux kernel dmesg or journalctl showing data about EDAC like:

Nov 08 15:43:30 haxx kernel: EDAC MC0: Giving out device to module amd64_edac controller F19h_M50h: DEV 0000:00:18.3 (INTERRUPT)
Nov 08 15:43:30 haxx kernel: EDAC amd64: F19h_M50h detected (node 0).
Nov 08 15:43:30 haxx kernel: EDAC MC: UMC0 chip selects:
Nov 08 15:43:30 haxx kernel: EDAC amd64: MC: 0:     0MB 1:     0MB
Nov 08 15:43:30 haxx kernel: EDAC amd64: MC: 2: 16384MB 3: 16384MB
Nov 08 15:43:30 haxx kernel: EDAC MC: UMC1 chip selects:
Nov 08 15:43:30 haxx kernel: EDAC amd64: MC: 0:     0MB 1:     0MB
Nov 08 15:43:30 haxx kernel: EDAC amd64: MC: 2: 16384MB 3: 16384MB

This is a Gigabyte B550 board, but I have seen the same on a Gigabyte X570 board.

I run Debian, and there’s a package called edac-utils which has a command-line utility (confusingly called edac-util) for reading the sysfs at /sys/devices/system/edac/mc/uevent and one (called edac-ctl) for configuring the labels given to results, adding your mainboard and model.

edac-ctl confirms I’ve had no uncorrected errors or corrected errors:

k3n@haxx:~$ sudo edac-util 
edac-util: No errors to report.

I don’t know where those perf counters will be in Windows.

K3n.

2 Likes

Thanks for your reply!

I will do some more testing soon and if anything usefull happens i will report back :slight_smile:

Let us know if the TUF supports ecc. If you decide to test it. :slight_smile:

Any results?