ROG STRIX X670E-E GAMING WIFI DDR5 memory green screen hell without constant retraining

[fatalframes]

Hello, I’ve run into a problem which has become increasingly annoying regarding the constant requirement to retrain my memory modules. Now the below listed memory combination from G.skill is on the ASUS verified memory list (I triple checked), also in combination with my CPU. While I’m perfectly cool with my mainboard training the DDR5 once I find it mind boggling that when I turn the retraining off (so my computer shuts down with values that have been reported to work) and cut the power next time I boot up my system with the stored values I immediately get a green screen due to memory failure. I’m using the latest BIOS (1807), and I’m pretty sure I didn’t have that problem before I updated (please, if your only contribution might be that I should then just roll back to the previous BIOS version, I don’t consider such replies helpful).

Mainboard: ROG STRIX X670E-E GAMING WIFI

CPU: AMD Ryzen 7 7800X3D

Memory: G.Skill F5-6000J3040G32GX2-TZ5NR (2 x 32 GB DDR 6000)

Bios version: 1807 (the most recent one)

OS: WIndows 11 Pro

Now, going by what I read to even make it on the verified memory list ASUS must have performed extensive testing on this particular memory module combination (2 x 32GB), so I can even make use of the “tweaked EXPO” feature. And I can do that just fine, only that then it becomes a hard requirement to keep the memory training enabled even without cutting power.. Because if I don’t keep it on I will instantly get yet another green screen caused by memory failure (if I keep it on booting takes “forever” going by todays standards, especially considering my main boot drive is a 990pro, but ya, w/e).

So the question I would like to ask: Are my memory modules to blame, and should I get them replaced or is this simply an AMD problem I will have to live with, until they get things sorted? I don’t get any green screens once I have overcome that hurdle, but on the other hand, on verified memory modules I think I should be reasonably able to expect not to see a green screen everytime I do not retrain the memory, cause I should be reasonably safe to assume that the values the motherboard obtained by training the memory should be stable with the values it obtained from doing so, without having to constantly retrain (cause either the values are stable, or they are not, right? I mean that’s what my monkey brain tells me, and even to my highly evolved human intellect this sounds logical enough).

So again, should I replace my memory modules, or is this just an ASUS/AMD/DDR5 thing I will have to live with until finally they come out with a BIOS that is actually stable? I mean I do watch popular youtubers (like JayzTwoCents and others), and they say if you go AMD you just have to live with these “quirks”, and that memory configurations that worked just fine with a previous BIOS version are suddenly completely broken with the next iteration for no discernible reason. But ya, I’d rather be safe than sorry, which is I came up here. I mean, I put down 500 bucks for the mainboard alone, so normally I would say (even though mainboard prices are absolutely crazy these days) that I should be able to have a more or less stable experience, especially considering these boards have been out for quite some time (but then again: from what I read earliest BIOS versions used to kill the CPU, so I guess that’s why the AI overclocking features they still advertise with are no longer actually in the BIOS, I mean literally nowhere. If you watch the tutorial on how to OC without the AI suite ¾ of the things shown on screen are not in the current bios version anymore).

Thanks for reading, and hopefully offering some constructive advice 😊

[Silent_Scone]

Hi @fatalframes, I'd recommend condensing your post if you expect customer service to respond quickly, there's a lot to digest there for what can be said with fewer words.

Few things to note:

If the system is crashing then the memory overclock is unstable. Subsequent memory training routines can expose such things if signal margins are tight.

Have you enabled Context Restore? This will retain the last "successful" retrain parameters. Note, if the overclock is conditional this function won't help much, you will need to tune the memory.

[fatalframes]

@Silent_Scone 

hah, yea, I guess you're right, I keep forgetting we live in the age of TikTok, where the average attention span is 2-3min at best, so you have my thanks for still taking the time. To be fair, I did want to shorten it, but there appears to be no edit function. So I guess if noone responds except for you I will do as you suggest. But to get to your post. 

When I enable context restore bios automatically also enables "power down enable" (I'm quite sure it didn't do that in v1709 either), which I always put back on "Auto" (cause I don't see the logic behind it activating itself, haven't tested yet if this wouid prevent a green screen even if cutting power).... However, even if I put it back on "Auto" (the "power down enable" option) and don't cut the power completely my system will boot up just fine until I eventually cut the power (typically do overnight, cause I don't need fancy standby RGB lighting while sleeping). 

Next day, when I turn on my computer with the settings memory training deemed stable the day before (and which remained stable throughout the day) this will instantly lead to a green screen (formerly known as blue screen). If I retrain the memory and then enable context restore it will again work just fine until next time I cut power (I think we can exclude battery failure cause all my bios settings are still in place, so it's something else). Again, I don't recall these issues with bios 1709, and this goes hand in hand with what I've read on the interwebz (and seen in youtube clips made by far more knowledgable ppl than I, aka Jayz2Cents, who reported the exact same RAM issues when updating BIOS on AMD boards, so this is not a new phenomenon).

So again: Considering my memory is on the QVL list for this particular motherboard + CPU would you suggest I return the memory, or is this simply an AMD/DDR5 issue some BIOS update will hopefully address? I mean, once it's trained my computer is completely stable.and I can boot as many times as I want without seeing a single green screen.... until I cut power of course which leads to abovementioned green screens (until I retrain the memory again, and my computer remains completely unstable until I completely turn it off o/n)

[Silent_Scone]

Updating the UEFI involves updating the AGESA firmware. Occasionally, this can impact overclocking.

Regarding retraining, the thing to do would be to adjust Shutdown mode in Armoury Crate for Aura Sync in order to turn off the RGB when the system is in standby. Alternatively enable ERP Ready within the UEFI.

[fatalframes]

That kinda fixed it @Silent_Scone , though it is not really a solution (is more a temporary measure, but I thank you all the same). But ya, if I don't cut power all the way I don't have the problem anymore.. Still, obviously I would prefer to be able to turn the PC off all the way to conserve power, but ya, I just hope that eventually a new bios will fix this, and I still don't quite get why "power down enable" is always enabled when I enable context restore (quite sure that wasn't there in 1709). So ya... or maybe I find a way to condense my original post over the weekend when I have more time (kinda difficult during the week when I have to work and all ^^), so again my thanks for the brief explanation and the suggestion, which at least serves as a temporary "hotfix" ^^

[Silent_Scone]

Both need to be enabled. As already mentioned, if you’re experiencing instability occasionally after a retrain then the memory overclock is conditional. Using MRC settings which retain training parameters is the “hot fix” in this case. With overclocking, all these things help.