BSODs on new system

[Ozaron]

Hi folks, I mentioned this elsewhere and thought I had it fixed but nay, this problem is giving me a headache. Possibly soon a wallet ache.

Set up a new desktop for gf two weeks back. Spec as follows;
Ryzen 3700X (Scythe Fuma 2)
MSI Unify X570
Gainward RTX 2060 Super
32GB TeamGroup 3000MHz DDR4
Kingston KC600 512GB SATA SSD
PNY CS3030 1TB NVMe SSD
EVGA Supernova G1+ 750w

All parts new out of box.
Essentially Windows is randomly throwing BSODs. I've received at least 8 seperate stop codes so far while attempting to resolve it. They don't occur during a specific workload, but I have noticed that once the system blue screens, it will reboot or BSOD a second time within minutes after, then boot a third time and be fine for hours.

-My first guess was RAM, but I've pushed the 4 sticks of 8GB through HCI Memtest to 200% coverage twice, and it's reported nothing wrong. I also edited the timings and tried without A-XMP to no avail.
-I've reinstalled Windows, had it installed on both SSDs (seperately, one after other).
-I've given the SATA SSD a new SATA cable, and swapped the port it was connected to. I've disabled it entirely, and alternatively disabled the NVMe drive, but whichever I had running the problem persisted.
-I've also tested temps, at full whack the CPU peaks at ~73c, RAM peaks at ~38c, GPU peaks at ~70c, nothing untoward is going on.
Edit: -Forgot to mention, I've updated to two seperate BIOS revisions, and both have presented BSOD.

Am I cursed to need a new motherboard?

[Domestic_Ginger]

Without reproducing the error It's difficult to pinpoint. Off the bat.

More stress tests? Prime, AIDA
Power saving options?
Check VRM temps
PSU working okay?
Different ram, one stick, in correct slots?
Enough voltage for the ram.

[Ozaron]

I've tried reproducing the "error" but given that two seperate Windows installations give an assortment of stop codes I'm struggling to see how it can not be hardware faulting, with that said I've run;
-two kinds of memory test which came back clean every time
-OCCT CPU test, produced no errors
-F@H workload produced no problems on CPU or GPU while I was running it (and yes, did have workloads for both)
-games, 3DMark, Superposition have proven no problem and it's played for hours, only to crash on desktop.

I've reseated RAM twice and tried fewer modules, didn't help. I've bumped voltage up for RAM also, didn't change anything.
VRM temps are fine.
I've run both normal Windows power plan and the suggested 1usmus power plans, including with and without BIOS options 1usmus suggests (Cool'n'Quiet, etc).

As mentioned, there's no workload that causes the system to BSOD. No structure or timing at all, while I have had it BSOD once in the night maybe during a folding WU and once on the main menu of Signal Simulator, it has also crashed many times on desktop, frequently while left to idle. Actually the most repeatable crash is after running HCI Memtest, both times after completing 200% RAM test coverage, when I stop and close the testing windows, the system has BSOD. However, simply running the test up to 10% and stopping / closing hasn't reproduced the problem.

[Domestic_Ginger]

I have a 1600AF and a b450m mortar. I have had a few probs with bad overclcoks but everything has been fine at stock.

I'm not sure at 200% the ram is stable. It may well be. I was only sure at 1200%; it seemed to error at about 600-800% a few times when I thought it was ok. I'm a bit paranoid so tried running 6/6 threads with prime/memtest.
Try some JDEC numbers for the ram i.e 2400? Something digusting.
Ram voltage is at 1.35V
SOC AUTO on my MOBO was about 1.15 far too high.

Is boosting the problem? Turn it off?
Voltages from the PSU are okay?

I'm sure you've tried everything :/

[Jonj1611]

What were the stop codes, try bluescreenview to find out.

Also while the ram maybe fine it may not be compatible with the board, is the ram on the approved memory list?

[Ozaron]

What were the stop codes, try bluescreenview to find out.

Also while the ram maybe fine it may not be compatible with the board, is the ram on the approved memory list?

Most recent three (since Windows reinstallation) were:
IRQL_NOT_LESS_OR_EQUAL (0X00000000a)
MEMORY_MANAGEMENT (0x0000001a)
SYSTEM_SERVICE_EXCEPTION (0x00000003b)

They affected Ntfs.sys, ntoskrnl.exe

I really don't think RAM sticks are the problem, and manually giving them JEDEC timings would not only make me sad but also don't think it would help.
Supposedly 400% will give you 99.x % error coverage and anything beyond that is just random for memtest, I could run up to 400% but again I don't think this is the problem, and if it were really that rare the system wouldn't be crashing around every 4-8 hours.
PBO's off, only normal PB2 boosting, and while the numbers have been good I think that's due to a steady flow of cold air. Fuma 2 is really good it seems. I could try manually setting the I/O die voltage to 1.05 and taking it off auto. I've not really measured the power supply but the system has had no issues folding on both CPU and GPU simultaneously (albeit the two work units didn't cross over for long, GPU was starved by their WU delivery service) so I could test that but again seems like a long shot.

If I had to make one observation on the cause of crashes, I would note it happens more often when I close applications. Memtest, games, I'm sure now that the folding unit which got interrupted by BSOD had completed and was being sent when the system crashed, also when locking the logged in user, so maybe that helps?

[Jonj1611]

Might be worth trying another sata cable, I went through 3 "new" ones before which all turned out to be faulty and had all sorts of errors

[DanceswithUnix]

Sounds similar to what I saw with the 2600X I briefly owned. Windows just blue screened occasionally, but thankfully Linux clearly stated in /var/log/messages that there were L1 cache errors being corrected constantly. Occasionally I would get a two bit uncorrectable error, and bang.

I generally keep a Linux live USB stick kicking around for this sort of thing. You might see some error messages that actually point a finger at the errant hardware.

... or it is the PSU or VRM iffy. Those are just a sphere-ache.

[ik9000]

I know you said you'd tried two bios but have you updated to latest bios? my board shipped with original bios and moving to latest one made a big difference.

[Ozaron]

Sounds similar to what I saw with the 2600X I briefly owned. Windows just blue screened occasionally, but thankfully Linux clearly stated in /var/log/messages that there were L1 cache errors being corrected constantly. Occasionally I would get a two bit uncorrectable error, and bang.

I generally keep a Linux live USB stick kicking around for this sort of thing. You might see some error messages that actually point a finger at the errant hardware.

... or it is the PSU or VRM iffy. Those are just a sphere-ache.

I don't have a Linux stick (or, to be fair, the Linux-fu, though I'm not completely new to it) to figure this out but that kind of distinction would be uniquely helpful in a situation like this. At least I could point a finger at what to try replacing.

I know you said you'd tried two bios but have you updated to latest bios? my board shipped with original bios and moving to latest one made a big difference.

Yeah, I moved to the latest BIOS immediately before installing Windows, then after a few days of BSOD issues moved backwards one revision to see if that had been the problem. It wasn't.

Might be worth trying another sata cable, I went through 3 "new" ones before which all turned out to be faulty and had all sorts of errors

That's impressively poor luck, but I figured that having wiped the drive, deleted its volume and offlined it in disk manager then it could no longer cause problems, so I'm ruling that SSD out.


Have since received further error codes, PAGE_FILE_IN_NONPAGED_AREA, two MEMORY_MANAGEMENT (maybe my fault, was tweaking RAM) and KMODE_EXCEPTION_NOT_HANDLED.



Edit: I know there's been talk about the sensitivity of Ryzen and the latest Core CPUs to older PSUs, do you think this page could have any relevance to my problem? Or does this also seem like a shot in the dark? I'm currently without any spare parts to play with and having an alternate PSU is clearly preferable over replacing CPU or motherboard...

Edit edit: The memory modules I own specifically are not on the QVL for this motherboard, but plenty of other SK Hynix AFR and Samsung B-die (I have 2 of each, though they arrived mixed already in the physical packaging) are included on the QVL, including a few similar kits by TeamGroup themselves. This should still only affect the XMP configuration and its optimization..

[kalniel]

It sounds like the kind of errors you can get while overclocking on a light load - on heavy load your mobo is supplying enough juice so you don't see crashes during torture tests and/or your CPU isn't turboing as much.

Make sure to disable PBO and any other tweaks that try to clock the CPU higher under light loads (you mention 1usmus which makes me think you might have been trying that). Get the CPU and RAM running at lower speeds first without turboing so much and see if that gives you stability. If so, then start tweaking what's required for the turbo to become stable - probably more volts somewhere.

[DanceswithUnix]

Make sure to disable PBO and any other tweaks that try to clock the CPU higher under light loads

Reset BIOS to defaults could be a good move if you haven't already tried it and should nail things like this. I have had to do that after a BIOS update in the past, and on my recent upgrade needed to do that to move from a 2200G to 3600 as I missed a setting I had tweaked.

[Ozaron]

It sounds like the kind of errors you can get while overclocking on a light load - on heavy load your mobo is supplying enough juice so you don't see crashes during torture tests and/or your CPU isn't turboing as much.

Make sure to disable PBO and any other tweaks that try to clock the CPU higher under light loads (you mention 1usmus which makes me think you might have been trying that). Get the CPU and RAM running at lower speeds first without turboing so much and see if that gives you stability. If so, then start tweaking what's required for the turbo to become stable - probably more volts somewhere.

Haven't turned PBO on or messed with CPU overclocking, honestly PB2 is just fine as is, only mentioned Usmus for the power plans (which I'm not using right now). Only change I made to CPU voltages was actually an offset, minus 0.025V. I think that's the smallest increment it yields in the BIOS.

Reset BIOS to defaults could be a good move if you haven't already tried it and should nail things like this. I have had to do that after a BIOS update in the past, and on my recent upgrade needed to do that to move from a 2200G to 3600 as I missed a setting I had tweaked.

Have tried this, but for the sake of being thorough I'll do it again next time the machine says no.

Had it run all night with F@H running and didn't get a crash, the last change made was reducing RAM speed to 2666MHz and loosening timings a bit.

[CAT-THE-FIFTH]

Have you tried only two sticks of RAM instead of 4??

[Jasp]

Hi Ozaron,

Just looking through all the notes, you mentioned windows power plans and the custom power plan you downloaded, are you using the RYZEN power plans that come with the AMD chipset drivers? you should have two options AMD RYZEN balanced or AMD RYZEN high performances, they alter the boost behaviour for the chip.

[ik9000]

don't rule out possibility of bad driver. I had similar problems on win7 on my machine when I was trying to find a usb3.1 driver that would work. rock solid on win10 though and on win7 with just generic usb3 drivers

edit on my new machine, not as sig: x570 ace + 3900x

SFAIK the old power supply incompatibility problems were to do with boot and wake from sleep rather than general running. That said I didn't look into it very closely, just skimmed over it in passing while trying to decide which mobo to get (before getting it wrong).