AMD AM4 system instability

[Tech Tweaker]

A bit of background, I built myself an AMD socket AM4 system back around January of this year (if I remember correctly).

And ever since I built it pretty much, it has had this odd quirk where it will randomly lock up or crash either when going into screen saver mode, or when coming back to the desktop when coming back out of screen saver mode.

This issue is persistent across multiple OS reinstalls, multiple different versions of Windows 10 and multiple different pieces of hardware.

Original Build Specs:
AMD Ryzen 5 3600X
Noctua NH-U12S heatsink
Gigabyte X570 Aorus Ultra Gaming Motherboard
2x8GB G.SKILL F4-3600C16D-16GTZR TridentZ RGB
EVGA GTX 1080 FTW3
Samsung 970 Evo 1TB OS Drive
Corsair RM650i PSU

After a few months of fighting with these issues I started swapping hardware, found a Ryzen 7 3800X online for a good price and put that in (I didn't buy it for this actually, had bought it for another build that I didn't move forward with). Issues continued. Then bought another Gigabyte X570 Aorus Ultra motherboard, issues continued. Reinstalled OS with the new motherboard and the 3800X in, issues continued. Swapped out original RAM sticks for a kit of F4-3200C14D-16GFX, issues continued. Tried probably 4-5 different video driver versions with no change. In addition to the GTX 1080 I also tried a couple of 1660 Supers and a 2060 Super, with no change in symptoms. Swapped out the RM650i power supply for an RM750i power supply (stupid, because my rig could possibly be under supplied with power, it only drew 350-375W under load at most while gaming), with no change. Bought a third kit of RAM at some point and tried that, and you guessed it this made no difference either.

Finally, I bought another board from a different manufacturer because after swapping out everything else and finding posts on other forums from users complaining of issues with Gigabyte and/or Aorus motherboards being unstable I was convinced that the motherboard must be the issue. Got myself an Asus ROG Crosshair VIII Hero and a new SSD, a Samsung 970 Evo Plus 1TB drive to do a completely fresh install of the OS. Within a few hours of the fresh install of Windows 10 and reinstalling "most" of my programs, it proceeds to lock up again after I'd walked out of the room to get something in the kitchen.

Next, came to the conclusion that the video card was the issue and tried it with an RTX 3070. Which, while helping my frame rates in games, made no difference to my random lockups when in screen saver mode.

I may have forgotten to mention this, but I can game on this system for hours on end with no obvious issues or instabilities. But will randomly crash or lock up once or twice a week when entering or exiting screen save mode. Temps are reasonable all the time, power draw on the system even at its highest-end/highest-power config maxes out at 425W while gaming. The longest I've managed to go without a crash is maybe two weeks. I've swapped hardware on this computer so often it's barely had time to get dusty, and the thermal paste certainly hasn't had time to break-down.

I am well and truly stumped, I have replaced every part of this computer, save for the case, and it is still unstable. I've basically re-built it three times now probably. I've even considered whether to cut my losses and just build an Intel-based computer instead, though that may not make sense at present with Intel rumored to finally be moving on with a new platform that hopefully won't still be 14nm+++++ with potentially DDR5 memory.

I had made a similar post to this a few months back, but then the site had to do a rollback the next day for some reason and my thread was lost as a result. I well and truly hope that I don't have to type all of this out a THIRD time, this time I waited until I was annoyed enough to retype it again (since it took me over 30 mins to type out the first time and I had no copy of it saved on my computer). This time I saved what I posted here in a file, if they roll back the site again, at least I have a copy of the original post to re-post it if I have to.

 

[Mr.Scott]

It's not a hardware issue. It is OS related and how it manipulates the power saving stuff when you go in or out of sleep mode or screen saver. Try changing to a performance power plan and see if that makes a difference.

 

[Tech Tweaker]

It's not a hardware issue. It is OS related and how it manipulates the power saving stuff when you go in or out of sleep mode or screen saver. Try changing to a performance power plan and see if that makes a difference.

I've tried Power saver, Balanced, and High performance power plans, they all resulted in a system crash or hang. High performance unfortunately results in a hotter-running system, but I kind of expected that.

It was stable for about a week on High performance, then one night I left my computer to go put some clothes in the dryer. Came back, and found my computer had locked up while I was in the laundry room.

I don't understand it really, my old PC never had this problem and it used the same PSU and GPU. Maybe it's just an AMD-specific issue, as my last two builds were Intel-based (Z97 and G33 chipsets) and neither had this issue? Although another difference is that I'm on NVMe storage now, and my last build was a 2.5" SATA SSD.

 

[Pvee]

A bit of background, I built myself an AMD socket AM4 system back around January of this year (if I remember correctly).

And ever since I built it pretty much, it has had this odd quirk where it will randomly lock up or crash either when going into screen saver mode, or when coming back to the desktop when coming back out of screen saver mode.

This issue is persistent across multiple OS reinstalls, multiple different versions of Windows 10 and multiple different pieces of hardware.

Original Build Specs:
AMD Ryzen 5 3600X
Noctua NH-U12S heatsink
Gigabyte X570 Aorus Ultra Gaming Motherboard
2x8GB G.SKILL F4-3600C16D-16GTZR TridentZ RGB
EVGA GTX 1080 FTW3
Samsung 970 Evo 1TB OS Drive
Corsair RM650i PSU

After a few months of fighting with these issues I started swapping hardware, found a Ryzen 7 3800X online for a good price and put that in (I didn't buy it for this actually, had bought it for another build that I didn't move forward with). Issues continued. Then bought another Gigabyte X570 Aorus Ultra motherboard, issues continued. Reinstalled OS with the new motherboard and the 3800X in, issues continued. Swapped out original RAM sticks for a kit of F4-3200C14D-16GFX, issues continued. Tried probably 4-5 different video driver versions with no change. In addition to the GTX 1080 I also tried a couple of 1660 Supers and a 2060 Super, with no change in symptoms. Swapped out the RM650i power supply for an RM750i power supply (stupid, because my rig could possibly be under supplied with power, it only drew 350-375W under load at most while gaming), with no change. Bought a third kit of RAM at some point and tried that, and you guessed it this made no difference either.

Finally, I bought another board from a different manufacturer because after swapping out everything else and finding posts on other forums from users complaining of issues with Gigabyte and/or Aorus motherboards being unstable I was convinced that the motherboard must be the issue. Got myself an Asus ROG Crosshair VIII Hero and a new SSD, a Samsung 970 Evo Plus 1TB drive to do a completely fresh install of the OS. Within a few hours of the fresh install of Windows 10 and reinstalling "most" of my programs, it proceeds to lock up again after I'd walked out of the room to get something in the kitchen.

Next, came to the conclusion that the video card was the issue and tried it with an RTX 3070. Which, while helping my frame rates in games, made no difference to my random lockups when in screen saver mode.

I may have forgotten to mention this, but I can game on this system for hours on end with no obvious issues or instabilities. But will randomly crash or lock up once or twice a week when entering or exiting screen save mode. Temps are reasonable all the time, power draw on the system even at its highest-end/highest-power config maxes out at 425W while gaming. The longest I've managed to go without a crash is maybe two weeks. I've swapped hardware on this computer so often it's barely had time to get dusty, and the thermal paste certainly hasn't had time to break-down.

I am well and truly stumped, I have replaced every part of this computer, save for the case, and it is still unstable. I've basically re-built it three times now probably. I've even considered whether to cut my losses and just build an Intel-based computer instead, though that may not make sense at present with Intel rumored to finally be moving on with a new platform that hopefully won't still be 14nm+++++ with potentially DDR5 memory.

I had made a similar post to this a few months back, but then the site had to do a rollback the next day for some reason and my thread was lost as a result. I well and truly hope that I don't have to type all of this out a THIRD time, this time I waited until I was annoyed enough to retype it again (since it took me over 30 mins to type out the first time and I had no copy of it saved on my computer). This time I saved what I posted here in a file, if they roll back the site again, at least I have a copy of the original post to re-post it if I have to.

I normally use ASROCK motherboards, sounds like one problem I had that a bios update fixed..

 

[DaveB]

Don't use the screen saver or sleep mode. I never use it on any of my systems, when they're on they stay on.

 

[Tech Tweaker]

Well, about three years on and my attempts to resolve these peculiar issues continue.

Tried a GTX 1080, GTX 1660 Super, RTX 2060 Super, RTX 3070, and RTX 3060Ti for Nvidia GPU's. Tried two different Gigabyte X570 Aorus Ultra motherboards, and an Asus ROG Crosshair VIII Hero (X570 chipset). I've tried four different ram kits, two different power supplies, three different M.2 NVME SSD's for the OS drive each with more than one OS reinstall. Disabled all Power-Saving features in the BIOS relating to the CPU, motherboard in general, and PCIE slots but none made a difference (tried turning off all at once, and trying turning off individual settings one at a time).

My RTX 3070 developed an issue maybe two or three months back where it liked to crash one specific game (Need for Speed: Payback), which I happened to be interested in playing at the time, swapped that out for something else. That seemed to stop the game crashing issue.

One of these crashes caused the entire system to crash, and upon restart my M.2 NVME games drive was entirely undetectable. That one required me to shut down the system and pull out the M.2 drive from the system, inspect it physically (found no obvious damage), and then upon reinstalling it and booting back up it was detected as normal like nothing had happened.

Then put my RTX 3070 in an Intel LGA1200 test bench and ran it through multiple benchmarks and stress tests of 3DMark Firestrike and Timespy, and could not get it to fail a benchmark or stress test even once.

Basically every Nvidia GPU I've tried has run flawlessly in any Intel build I've used them in, and the RTX 2060 Super I had in my AMD socket AM4 rig that the rig was crashing with previously has been running perfectly fine in my X99+6850K rig for somewhere near a year now.

And then I decided to try out an AMD RX 6700XT I had on hand. Card was brand new, aside from being bench-tested on an X99 test bench of mine for a couple of hours when I first bought it. I had that GPU in my computer for something like a month, and my computer didn't lock up or crash once while it was in there. I did run DDU before I installed the AMD drivers for this card, to wipe out the traces of the Nvidia drivers used for the RTX 3070 and any other Nvidia card it might have had a record of in the registry somewhere.

Pulled out the RX 6700XT, wiped the driver traces with DDU and installed an RTX 3060 Ti and an Nvidia driver. With the RTX 3060 Ti in, I proceeded to have another lockup of my system within a couple of days.

Disabling screen saver isn't a solution for me, I did in fact try that at one point and it didn't solve it either. My monitor had turned off when I was out of the room, I came back in and moved the mouse or tapped a key on my keyboard to wake the monitor and it resulted in the computer locking up again. I also prefer to have screen saver enabled anyway.

It seems I have built an AMD sAM4 computer which refuses to work properly with an Nvidia GPU/graphics card in it. Worked fine with an AMD GPU though, for something like a month. And I also have an RTX 3070 which crashes one specific game, but completes benchmarks and stress tests with no issues and could presumably run other games just fine.

I've never heard of an AMD platform computer that couldn't work with an Nvidia GPU, or an Intel platform that wouldn't work with an AMD GPU for that matter. But, I seem to have one of those.

In late 2020 I'd decided to go back to an AMD-based PC because it had been so long since I'd used one for my main PC (last was in 2010 IIRC) and I wanted to see what AMD had come up with since then, and I've regretted it basically ever since I built this thing because it's never been fully stable.

I'm very much thinking of getting myself off of this platform entirely and replacing it with an Intel-based PC, as I'm just getting tired of the random lockups and crashes. I've got enough spare hardware on hand that I can build an Intel 10th gen PC pretty quickly (or 9th gen, or 8th gen, or 7th gen but I don't want to go back that far). At this point, I hardly even care if the Intel platform I switch to is slower than my current one, so far as it doesn't randomly crash or lock up multiple times a month. This Core i9 10850K I have here may be calling my name, I just need to decide which board to put it in.

 

[Kenrou]

I had a very similar issue with wife's rig, specs are very close to OPs post - 3600x + 2x8gb g.skill 3600 c16 + 2070s on a b450 tomahawk max, everything default on bios, turns out the issue was the XMP profile, the motherboard didn't like the primary profile, randomly crashed after being stable for days, but started working normally (and has been ever since) when I enabled the secondary profile, I assumed AMD quirk or looser timings since it's T2 memory instead of the faster T1 they like (not Neo)

Sorry if this was covered and I didn't catch it, but I saw mention of switching memory but not XMP profiles, I don't even know how many sticks actually come with several profiles/if it's standard...

 

[Tech Tweaker]

I had a very similar issue with wife's rig, specs are very close to OPs post - 3600x + 2x8gb g.skill 3600 c16 + 2070s on a b450 tomahawk max, everything default on bios, turns out the issue was the XMP profile, the motherboard didn't like the primary profile, randomly crashed after being stable for days, but started working normally (and has been ever since) when I enabled the secondary profile, I assumed AMD quirk or looser timings since it's T2 memory instead of the faster T1 they like (not Neo)

Sorry if this was covered and I didn't catch it, but I saw mention of switching memory but not XMP profiles, I don't even know how many sticks actually come with several profiles/if it's standard...

Thanks for that suggestion, I had a look in my BIOS last night but my current sticks only have one XMP profile. I have seen other sticks that offer two, but the memory I currently have in my system only has one profile for whatever reason.

I went in though and manually set the Command Rate to 2T timings, I'll see if that makes any difference.

Just checked my memory settings in CPU-Z, it's showing Command Rate at 1T timings. I wonder what I changed, or why the change didn't make a difference to the reported Command Rate?

Edit: checked again, found that Command Rate seems to be able to be set in two different locations for some reason. I've set it to 2T in both, but CPU-Z still reports the CR as being 1T, so I'm kind of stumped. Either the setting has no effect or the software is misreporting the setting.

 

[Zerileous]

The easiest way to see if RAM is the problem is going to be disabling XMP and running it at the default settings for a bit. If the problems go away then you can try to isolate the setting that needs to be adjusted.

I believe Kenrou was referring to the XMP profile using T2 instead of T1, not that his issue was corrected by changing the CR to T2.

Is your memory on the QVL for the motherboard? The motherboard manufacturer maintains a list of memory that was tested to work at XMP on a given board. GSkill also has a compatibility list by motherboard reflecting your own testing. You can check both places.

 

[Kenrou]

That's the thing, as far as I know both Intel and AMD used T2 until Ryzen made T1 "popular", Intel still uses T2, Ryzen PREFERS T1 (the aforementioned NEO sticks, also Flare X). I remember trying to get T1 to run stable on my old 6700k was a PITA, but voltage cures (almost) everything

Someone that knows more about memory and how this can affect stability please chime in here

 

[Tech Tweaker]

The easiest way to see if RAM is the problem is going to be disabling XMP and running it at the default settings for a bit. If the problems go away then you can try to isolate the setting that needs to be adjusted.

I believe Kenrou was referring to the XMP profile using T2 instead of T1, not that his issue was corrected by changing the CR to T2.

Is your memory on the QVL for the motherboard? The motherboard manufacturer maintains a list of memory that was tested to work at XMP on a given board. GSkill also has a compatibility list by motherboard reflecting your own testing. You can check both places.

I honestly basically never even look at the QVL lists for RAM on motherboards, given that basically none of the RAM kits I've used in systems in the past were on the QVL lists of my old motherboards and I'd never had problems on any other system in the past related to memory. I'd kind of adopted the impression that the QVL list didn't matter much, since past the first year or two of a motherboard being out I'd found that manufacturers don't really update the QVL list so I've often ended up with memory kits newer than the motherboard and/or newer than the last time the QVL list for a board was updated and so they'd never been tested for whatever board I owned at the time.

My current kit is F4-3600C16D-32GTRS, G.SKILL Trident Z Royal Silver, it isn't on the QVL list. I did not buy them for their looks or exclusiveness, I bought them because they were the only single-sided 32GB dual stick kit I could find which used Samsung B-Die chips when I was looking.

My previous kit was F4-3600C16D-16GTZR, G.SKILL TridentZ RGB, which isn't on the QVL list.

My original kit was F4-3200C14D-16GFX, G.SKILL FlareX, which isn't on the QVL list but lists directly on the sticks themselves that they are AMD Compatible memory.

I'd bought all of these because they were Samsung B-Die memory and because I'd read threads on other sites about people having good performance and/or compatibility with AMD/Ryzen platform systems.

Oddly though, while they aren't listed on the QVL for this board on Gigabyte's website, they are all listed as compatible recommendations for this board using G.SKILL's online RAM Configurator. I know the speeds of these should be compatible with Ryzen 3000 and Ryzen 5000, but know they wouldn't be compatible with Ryzen 2000 or 1000 as they topped out around 2800-2933MHz IIRC.

I kind of went with Command Rate, since I don't have a second XMP profile and because I remember reading something years ago about certain motherboards and/or CPU's not liking the 1T Command Rate. Figured I'd try that, since I think it's one of the few things I hadn't tried.

 

[Kenrou]

F4-3200C14D-16GFX, is the kit I'm using on this system right now (going on 4y now), OC as you can see on my sig, as far as I know these were the 1st to have the tag "Ryzen/AMD compatible" and using T1 on the XMP. Although they are tagged as B-Die, I think they are low quality or are on the edge (they were cheap even at the time they came out), because in terms of overclocking they have been less than stellar so far, that coupled with the "newness" of Ryzen 1 and 2 might have been the tipping point for compatibility with most mobos

 

[Trap05]

Is your bios current?

 

[Zerileous]

For me running B-die (4x8) on AM4 (3700x) I was able to reach 3600 at 1.4v vRAM, CL 16 16 16 38 45 420 1T (sorry I'm reading my chicken scratches from a couple of years ago and I didn't right down exactly what timing was what, but you should get an idea from the value range) You shouldn't need 2T to get 3600. Your IMC may need a little extra voltage. You can also just search B-die timings or look at something like the ryzen DRAM calculator - but off-the-shelf settings never work for every system, they are just a starting point.

Honestly, 3200 is just as good as 3600 unless you're benching. I still think you should try it at stock first to see if this is even the issue.

 

[MisterEd]

Original Specs:
CPU: AMD Ryzen 7 1700X
MBD: ASUS Prime X370-Pro Motherboard
RAM: Corsair Vengeance CMK16GX4M2A2666C16R 16GB (2x8GB)
GPU: ASUS STRIX-GTX1060-6G-GAMING
SSD: ADATA XPG SX8200 Pro (2TB M.2 NVME)
SSD: Intel 660P (2TB M.2 NVME)
PSU: Corsair RM750i

Current Specs:
CPU: AMD Ryzen 7 3700X
MBD: ASUS Prime X370-Pro Motherboard
RAM: G.Skill Flair X F4-3200C16D-32GFX 32GB (2x16GB)
GPU: ASUS ROG-STRIX-RTX3060TI-O8G-V2-GAMING
SSD: ADATA XPG SX8200 Pro (2TB M.2 NVME)
SSD: Intel 660P (2TB M.2 NVME)
PSU: Corsair RM750x

Note my motherboard is first generation Ryzen that I pre-ordered in March 2017. The new RAM I have is not on the QVL list. In fact G.Skill does not even list 32GB kits faster than 2933MHz for my motherboard.

A couple years ago my computer started crashing when I ran a benchmark or played games. When it crashed my screen went black and the fans went to max speeds. I had to unplug the computer and plug it back in before I could power it on again. It turned out the original power supply a Corsair RM750i was at fault. I proved that by replacing it with an old Antec TruePower New TP-650. I RMA'd the RM750i and received a RM750x in return. No more problems.

BTW, that the crashing started when I still had my original configuration. I was testing my computer in preparation of upgrading the CPU, RAM & GPU two years ago.

 

[Tech Tweaker]

F4-3200C14D-16GFX, is the kit I'm using on this system right now (going on 4y now), OC as you can see on my sig, as far as I know these were the 1st to have the tag "Ryzen/AMD compatible" and using T1 on the XMP. Although they are tagged as B-Die, I think they are low quality or are on the edge (they were cheap even at the time they came out), because in terms of overclocking they have been less than stellar so far, that coupled with the "newness" of Ryzen 1 and 2 might have been the tipping point for compatibility with most mobos

Well, that is interesting to know at least one person is running one of the same kits I tried at one point without issues.

Is your bios current?

Not quite, I was on the F35 BIOS and checking Gigabyte's site earlier I discovered there was an F36 version as well as a couple of what I would consider Beta bioses for fixing specific issues. So I downloaded and updated to F36 and spent 30 minutes redoing my bios settings (Fan settings, enabling Re-BAR and 4G Decoding, disabling CSM, enabling Secure Boot, and so forth). F36 didn't exist when I updated this motherboard in August of 2022 to get Resizable-BAR to work.

For me running B-die (4x8) on AM4 (3700x) I was able to reach 3600 at 1.4v vRAM, CL 16 16 16 38 45 420 1T (sorry I'm reading my chicken scratches from a couple of years ago and I didn't right down exactly what timing was what, but you should get an idea from the value range) You shouldn't need 2T to get 3600. Your IMC may need a little extra voltage. You can also just search B-die timings or look at something like the ryzen DRAM calculator - but off-the-shelf settings never work for every system, they are just a starting point.

Honestly, 3200 is just as good as 3600 unless you're benching. I still think you should try it at stock first to see if this is even the issue.

Well, I ran mine with 4x8GB a couple of years back with the regular XMP profile (CL16-16-16-36 1T @1.35V). It ran basically the same with that RAM configuration as it does with 2x16GB. I'd swapped to 2x16GB because I'd read somewhere that Ryzen sometimes has issues running 4x8GB, but my 3600X and 3800X didn't seem to be affected by that though, as they didn't respond any differently after I switched the RAM config.

Well, I'm at system defaults on the memory now, boy going from 3600 MHz to 2133MHz just doesn't feel right. Missing out on some of the speed I paid for and all.

Original Specs:
CPU: AMD Ryzen 7 1700X
MBD: ASUS Prime X370-Pro Motherboard
RAM: Corsair Vengeance CMK16GX4M2A2666C16R 16GB (2x8GB)
GPU: ASUS STRIX-GTX1060-6G-GAMING
SSD: ADATA XPG SX8200 Pro (2TB M.2 NVME)
SSD: Intel 660P (2TB M.2 NVME)
PSU: Corsair RM750i

Current Specs:
CPU: AMD Ryzen 7 3700X
MBD: ASUS Prime X370-Pro Motherboard
RAM: G.Skill Flair X F4-3200C16D-32GFX 32GB (2x16GB)
GPU: ASUS ROG-STRIX-RTX3060TI-O8G-V2-GAMING
SSD: ADATA XPG SX8200 Pro (2TB M.2 NVME)
SSD: Intel 660P (2TB M.2 NVME)
PSU: Corsair RM750x

Note my motherboard is first generation Ryzen that I pre-ordered in March 2017. The new RAM I have is not on the QVL list. In fact G.Skill does not even list 32GB kits faster than 2933MHz for my motherboard.

A couple years ago my computer started crashing when I ran a benchmark or played games. When it crashed my screen went black and the fans went to max speeds. I had to unplug the computer and plug it back in before I could power it on again. It turned out the original power supply a Corsair RM750i was at fault. I proved that by replacing it with an old Antec TruePower New TP-650. I RMA'd the RM750i and received a RM750x in return. No more problems.

BTW, that the crashing started when I still had my original configuration. I was testing my computer in preparation of upgrading the CPU, RAM & GPU two years ago.

Interesting, another person that had issues while using a Corsair RMi series power supply.

I'd considered the PSU being the problem, but discounted the theory when I had the problem with two different power supplies. I'd considered that mainly due to age (mine is ~8 years old now), but also because it was designed for use with an Intel 4th Gen platform (something about being compatible with low-power states/power saving features like C3, C6 and C7). Although, I'd tested with an RM650i and an RM750i because they were both readily available, perhaps testing with two RMi series power supplies is part of my problem. Not enough variance between the two when they're from the same PSU series, and they're just built to accommodate different capacities.

 

[Kenrou]

Well, that is interesting to know at least one person is running one of the same kits I tried at one point without issues.

Well, I wouldn't say without issues, my old 6700k REALLY didn't like them, but ran fine on wife's 3600x and this one (optimized for AMD and all)...

 

[freeagent]

Since moving to AM4 myself, I have surmised that the easiest ram to run on Ryzen, specifically Zen 2 and 3 is Samsung B-Die. I have 3 different sets, they all run together, apart, fast slow, tight, loose, no problem. I have some Adatas that are not B-Die, nor are they on QVL, and I cannot save settings once power is removed from the system. Really weird. Had to make a profile. I normally dont bother with QVL neither, but it seems to have some relevance these days. I cannot speak for Intel though, but I wanna build one shortly

 

[MisterEd]

The G.Skill Flair X F4-3200C16D-32GFX 32GB RAM I have has Hynix DRAM. Since I am using a Zen 2 CPU that may be why I am able to run 3200MHz RAM. I think I remember people with Zen 1 CPUs could at best only run 2933MHz RAM.

BTW, I pre-ordered my Corsair Prime X370-Pro motherboard and AMD Ryzen 7 1700X CPU in February 2017. When I ordered the RAM there wasn't a QVL list yet. I thought I was being conservative when I ordered Corsair Vengeance 2666MHz RAM. Unfortunately, even this RAM was unstable at XMP so I had to run it at 2133MHz. It must have been 6 months before there was a BIOS update that allowed me to run the RAM at the XMP speed of 2666MHz. Note D.O.C.P sets XMP.

 

[Niku-Sama]

i've never been able to run any g.skill ram at its rated XMP profile on any of my AM4 boards and it goes way beyond the command rate settings.
Most of the time i wind up giving the ram and memory controller a little more voltage as well as fine tuning the major timings. Another timing that goes largely missed is TRFC.
Adjusting TRFC higher, or longer (more ns) depending on how your bios displays it, was key to making things stable from my experience.

Had this happen on Asrock and MSI motherboards. something to look into. its free to try and safe.