AMD Processor ECC Memory Support: Why So Hinky?

[A Computer Guy]

According to online sources 5 = Single-bit ECC, 6 = Multi-bit ECC

If you motherboard UEFI/BIOS supports error injection MemTest86 pro can be used also it seems they have a piece of DDR4 hardware that can help with that as well. (https://www.memtest86.com/ecc.htm)

 

[adata]

Perhaps others will find this informative:
On a platform with ECC implemented (CPU + Chipset + Motherboard + RAM) this is what can be observed in the Windows Hardware Error Architecture log in the event viewer:

There are many kinds of errors that may appear in the WHEA-Logger, but this screenshot is all for correctable memory errors. Ideally, they would not be so frequent. There are numerous possibilities that could be the cause, but the ECC on this particular platform is doing its job preventing it from crashing.

Window Server 2022 Standard (16-core)
ASUS TUF GAMING B550-PRO
AMD Ryzen 5 5600
(4x) NEMIX 16GB DDR4-3200 ECC UDIMM

The most likely reason for this system having such frequent ECC events is that it has all 4 memory slots populated, giving the memory controller the most amount of work to keep them all in sync. Also I have not adjusted anything in BIOS. Everything is on Auto.

On this particular platform, PassMark MemTest86 does not detect any errors during a full 4-pass screening. If it did, I would make the effort to manually reduce the speed, increase the voltage, adjust the CAS timings, etc. I could also switch to 32GB modules, which would probably reduce the frequency of ECC events because it would be less effort for the memory controller to keep only 2 modules in sync instead of 4.

I'm currently fine with it the way it is since it never crashes. I only have to reboot once a month on Patch Tuesday.

 

[A Computer Guy]

Perhaps others will find this informative:
On a platform with ECC implemented (CPU + Chipset + Motherboard + RAM) this is what can be observed in the Windows Hardware Error Architecture log in the event viewer:
View attachment 384141

There are many kinds of errors that may appear in the WHEA-Logger, but this screenshot is all for correctable memory errors. Ideally, they would not be so frequent. There are numerous possibilities that could be the cause, but the ECC on this particular platform is doing its job preventing it from crashing.

Window Server 2022 Standard (16-core)
ASUS TUF GAMING B550-PRO
AMD Ryzen 5 5600
(4x) NEMIX 16GB DDR4-3200 ECC UDIMM

The most likely reason for this system having such frequent ECC events is that it has all 4 memory slots populated, giving the memory controller the most amount of work to keep them all in sync. Also I have not adjusted anything in BIOS. Everything is on Auto.

On this particular platform, PassMark MemTest86 does not detect any errors during a full 4-pass screening. If it did, I would make the effort to manually reduce the speed, increase the voltage, adjust the CAS timings, etc. I could also switch to 32GB modules, which would probably reduce the frequency of ECC events because it would be less effort for the memory controller to keep only 2 modules in sync instead of 4.

I'm currently fine with it the way it is since it never crashes. I only have to reboot once a month on Patch Tuesday.

I could never get 128GB Nemix to work 100% properly regardless of scaling voltages and speeds and ultimately ended up switching to Crucial/Micron instead which worked perfectly in both of my 5950x systems. (see post below, in particular Micron DDR4-3200 ECC UDIMM 18ASF4G72AZ-3G2F1)

What did you have to do to get 128GB ram working on Zen3 CPU?

Thank for this fast answer :) The problem is I have performance issues... cpu seems saturated by pics sometimes, resulting in rollback in game for players. I think cpu is enough strong to supports, but think the ram is too long. It's about 14 ark servers. Here, I saw working 3200mhz with 128gb...

www.techpowerup.com

You shouldn't settle for it throwing that many errors because either a dimm is going bad or it's on the edge of stability for some reason.