Strange system crashes out of nowhere, help

[Mazer]

First of all, I want to say that my PC has been working fine in the last few months with no problems, I don't know what could be causing this very bizarre crash and it's a little difficult to explain. I will try my best here.

I'm getting this PSHED.dll + 15b0 ntoskrnl.exe crash when I play some games and it's a little random and by that I mean everything works fine on idle or even when playing games normally but then I get this reset or hard crash that makes both of my screens dim dark weirdly with no response back (it makes my PC hard lock where I have to force shut down to get it back up again) I will see if I can upload a recording of this happening it doesn't look good but sometimes it reboots instead of the hard crash. Another thing I want to add here is that I turned from PBO on -10 on all cores to setting back defaults as in PBO auto off as my CPU temps were getting up to 80c+, I did the usually AIDA64+prime95 stress testing AFTER this crashing was happening and so far so good no crashing or weird hardlocking on load. My ram has been tested too with memtest86 for 5 hours now and going only time will tell...

PC specs are here https://pcpartpicker.com/list/YP86bL

I have installed bluescreenview hoping someone could help me find out what it causing this crash? I did see in event viewer that core 12 of my cpu was giving WHEA error but only once it happen and that is when PBO was turned on auto. It never did this before I have no idea if this is the culprit or not. I'm at a loss what is going on here.

 

[tabascosauz]

Bus/Interconnect = Fabric not stable

Has nothing to do with your Curve Optimizer settings - Bus/Interconnect comes from uncore. Thus APIC ID is also not useful. Either:

• Fabric is not stable at that speed on your CPU, period
• Fabric is not stable for your given VSOC
• Fabric is not stable due to your VDDG settings (primarily VDDG_IOD)

Send a Zentimings screenshot to show actually relevant info for this.

Testing CO under full load accomplishes next to nothing as long as it's not so unbelievably unstable it's crashing immediately at the start of test. CO instability manifests at idle because that point in the V-F curve has dipped too low. If you want CO testing, you're going to have to test with something like Corecycler or OCCT in its specific testing mode. Otherwise you can never get clocks as high as they need to go to properly test.

Memtest86 is borderline worthless for "memtesting" unless you're concerned about a stick that is potentially physically broken. It puts zero stress on the uncore side of CPU (Fabric and memory controller) and doesn't work the RAM hard either. Use Testmem5.

 

[Mazer]

Bus/Interconnect = Fabric not stable

Has nothing to do with your Curve Optimizer settings - Bus/Interconnect comes from uncore. Thus APIC ID is also not useful. Either:

• Fabric is not stable at that speed on your CPU, period
• Fabric is not stable for your given VSOC
• Fabric is not stable due to your VDDG settings (primarily VDDG_IOD)

Send a Zentimings screenshot to show actually relevant info for this.

Testing CO under full load accomplishes next to nothing as long as it's not so unbelievably unstable it's crashing immediately at the start of test. CO instability manifests at idle because that point in the V-F curve has dipped too low. If you want CO testing, you're going to have to test with something like Corecycler or OCCT in its specific testing mode. Otherwise you can never get clocks as high as they need to go to properly test.

Memtest86 is borderline worthless unless you're testing for stick hardware failures. It puts zero stress on the uncore side of CPU and doesn't work the RAM hard either.

 

[tabascosauz]

View attachment 345291

okay, assuming this CJR kit is stable at those XMP settings, not gonna dwell on that further unless you run TM5 and it turns up errors.

VDDP is way too high. Should be fine at 0.9V.

VDDG_IOD can come up to 1.05V or so. Try that.

 

[Mazer]

okay, assuming this CJR kit is stable at those XMP settings, not gonna dwell on that further unless you run TM5 and it turns up errors.

VDDP is way too high. Should be fine at 0.9V.

VDDG_IOD can come up to 1.05V or so. Try that.

What caused the memory to become like that from just trying out PBO? I will try those settings and test again, is there anything else I should look out for?

 

[tabascosauz]

What caused the memory to become like that from just trying out PBO? I will try those settings and test again, is there anything else I should look out for?

If you haven't changed anything, then they're just voltage auto-rules decided by the board firmware. Setting or disabling PBO isn't likely to affect any of them, unless your board BIOS happens to be super bugged beyond reason.

Not really much else, WHEA 19 kinda speaks for itself. Fabric has always been hard to test as well, there are good stability tests for CPU, good tests for CO settings, good tests hammering memory controller and RAM, but hard to find stuff specifically good for Fabric. If you can get through a demanding config of TM5 (1usmusv3, it's in the zip with TM5), and no WHEA 19 or other symptoms, then don't worry about it.

Also, as to PBO, there's a lot of granularity beyond just "enabling" it. From your R23 run you were running way too much wattage for that score (150W). You can adjust CO independently of PBO - either set PBO advanced and punch in whatever PPT/TDC/EDC you desire manually, or set CO then disable PBO again. It will hide all the settings, but your CO settings will stick.

 

[Mazer]

If you haven't changed anything, then they're just voltage auto-rules decided by the board firmware. Setting or disabling PBO isn't likely to affect any of them, unless your board BIOS happens to be super bugged beyond reason.

Not really much else, WHEA 19 kinda speaks for itself. Fabric has always been hard to test as well, there are good stability tests for CPU, good tests for CO settings, good tests hammering memory controller and RAM, but hard to find stuff specifically good for Fabric. If you can get through a demanding config of TM5 (1usmusv3, it's in the zip with TM5), and no WHEA 19 or other symptoms, then don't worry about it.

Also, as to PBO, there's a lot of granularity beyond just "enabling" it. From your R23 run you were running way too much wattage for that score (150W). You can adjust CO independently of PBO - either set PBO advanced and punch in whatever PPT/TDC/EDC you desire manually, or set CO then disable PBO again. It will hide all the settings, but your CO settings will stick.

I just did a TM5 test with the 1usmusv3 config file you mentioned, and it past no errors were detected. If I run into to it restarting/crashing whole PC up again, I'll try the VDDP and VDDG-IOD settings you posted hope everything is okay now. I was worried my CPU or PSU were to blame since I hear a buzzing sound coming from my pc I think it's from a fan maybe coil whine it's only happening in one game for me.

 

[tabascosauz]

I just did a TM5 test with the 1usmusv3 config file you mentioned, and it past no errors were detected. If I run into to it restarting/crashing whole PC up again, I'll try the VDDP and VDDG-IOD settings you posted hope everything is okay now. I was worried my CPU or PSU were to blame since I hear a buzzing sound coming from my pc I think it's from a fan maybe coil whine.

The TM5 testing was just to make sure your assumptions about RAM stability are sound (which it sounds like they are, which is good).

The main issue here is still Fabric, testing in TM5 won't fix that. Most of the time, VSOC and/or IOD should solve Fabric stability issues, in that order of priority.

 

[Mazer]

The TM5 testing was just to make sure your assumptions about RAM stability are sound (which it sounds like they are, which is good).

The main issue here is still Fabric, testing in TM5 won't fix that. Most of the time, VSOC and/or IOD should solve Fabric stability issues, in that order of priority.

I went in to the BIOs and changed them. How should I test the fabric in the same game I had problems in?

 

[tabascosauz]

I went in to the BIOs and changed them. How should I test the fabric in the same game I had problems in?

Just do what you normally would do, honestly. Like I said, it's hard to test Fabric specifically. But as an uncore component, it is usually useful enough to use a good memtest (HCI, TM5, Karhu) as it stresses the memory controller and puts an increased amount of bandwidth throughput through the UMC and Fabric. At 3600 (1800MHz) it's not like you're trying to push the absolute bleeding edge of Fabric speeds, anyway.

 

[Mazer]

Well I tried that same game again and it happen to lock up my whole PC I was playing different game before and it didn't do anything like this one does. This is what it looks like on my screen when it freezes my PC to give you an idea:

Now I'm thinking it's my psu or need a bios update..

 

[tabascosauz]

Well I tried that same game again and it happen to lock up my whole PC I was playing different game before and it didn't do anything like this one does. This is what it looks like on my screen when it freezes my PC to give you an idea:

Now I'm thinking it's my psu or need a bios update..
View attachment 345322
View attachment 345320

Aren't you on 1.I already? That's only the previous release before current. AM4 AGESA has been stable for a long time, there haven't been any real performance changes since like, 2022 (except those negatively impacting performance due to vulnerability fixes). So unless your BIOS just happens to be uniquely bugged, don't expect anything to be different just because newer AGESA.

Not sure what exactly I'm looking at in the photos. Either looks like GPU artifacting or a Windows install that is in dire need of a clean install.

If this bug never happens with RAM on JEDEC speeds, then this is still the same problem.

 

[Mazer]

Not sure what exactly I'm looking at in the photos. Either looks like GPU artifacting or a Windows install that is in dire need of a clean install.

It looks like inverted colors to me all messed up the PC becomes unresponsive have to hold down the power switch when it happens, I did run a sfc /scannow and looked at my OS ssd both are fine so I have no idea what is the problem unless it's just DX12 games that are screwing up my system?

I tried Trepang2 DX11 game - no problems during gameplay. Half life 2 DX9 no issues either. Same error.

 

[tabascosauz]

It looks like inverted colors to me all messed up the PC becomes unresponsive have to hold down the power switch when it happens, I did run a sfc /scannow and looked at my OS ssd both are fine so I have no idea what is the problem unless it's just DX12 games that are screwing up my system?

I tried Trepang2 DX11 game - no problems during gameplay. Half life 2 DX9 no issues either. Same error.

View attachment 345323

Which games? You didn't specify. Half Life 2 is hardly a demanding game, idk about Trepang. Still doesn't seem to rule out GPU/power issues.

Like I said, if you drop down to JEDEC speeds and run it for a while, should give some ideas as to whether you are dealing with 2 different issues on your hands.

It's not a particularly high power draw rig, but GX2 is also an incredibly middling quality PSU.

 

[Mazer]

Which games? You didn't specify. Half Life 2 is hardly a demanding game, idk about Trepang. Still doesn't seem to rule out GPU/power issues.

Like I said, if you drop down to JEDEC speeds and run it for a while, should give some ideas as to whether you are dealing with 2 different issues on your hands.

It's not a particularly high power draw rig, but GX2 is also an incredibly middling quality PSU.

Oh sorry, I had forgot the game it's new called backroom break it's poorly optimize as in very limited options the game doesn't even have a way to switch off the very awful TAA. I finished it just before it crashed my PC. Always crashed it within 5 mins or sooner with a reboot when you break walls or anything physics related. A bad game overall.

The only other game that it crashed doing that was metro exodus just before the benchmark it was loading. Like I said it's very strange only happening during transitions (credits/loading) I can install other games like HFW to see if it does crash. Robocop doesn't crash so that one is okay was playing for good 10 mins.

I am thinking of either switching out the RAM and or the PSU at this point (even though I keep spending more and more on this as of now once a month) The RMe or RMx I can buy next week and RAM too.

@tabascosauz So after last night I came back to play metro exodus and not even being able to play it long and my system rebooted itself WHEA error same one again

 

[tabascosauz]

Oh sorry, I had forgot the game it's new called backroom break it's poorly optimize as in very limited options the game doesn't even have a way to switch off the very awful TAA. I finished it just before it crashed my PC. Always crashed it within 5 mins or sooner with a reboot when you break walls or anything physics related. A bad game overall.

The only other game that it crashed doing that was metro exodus just before the benchmark it was loading. Like I said it's very strange only happening during transitions (credits/loading) I can install other games like HFW to see if it does crash. Robocop doesn't crash so that one is okay was playing for good 10 mins.

I am thinking of either switching out the RAM and or the PSU at this point (even though I keep spending more and more on this as of now once a month) The RMe or RMx I can buy next week and RAM too.

@tabascosauz So after last night I came back to play metro exodus and not even being able to play it long and my system rebooted itself WHEA error same one again

You can go a bit higher on VSOC up to a limit of 1.2V, but especially beyond 1.15V we are definitely getting into diminished returns and VSOC can actually work against stability at that point

I think I've suggested like 3 times now that you run at JEDEC speeds to see if the problem goes away lol

Not a whole lot of point in replacing RAM, Fabric is an uncore component of the CPU and your kit itself is clearly stable. PSU maybe, GPU possibly, but neither would address the WHEA 19 issue, only a potential separate issue.

And like I said in the first post:

- you might be able to solve it with VSOC
- you might be able to solve it with VDDGs
- you might not be able to solve it at all, and it's just a terminally bad CPU that needs RMA

 

[Mazer]

I think I've suggested like 3 times now that you run at JEDEC speeds to see if the problem goes away lol

You mean 2133mhz or 3200mhz?

 

[tabascosauz]

You mean 2133mhz or 3200mhz?

Whichever the default profile/speed without XMP is......seems unlikely that a CJR kit defaults to 3200CL20.

 

[Mazer]

Whichever the default profile/speed without XMP is......seems unlikely that a CJR kit defaults to 3200CL20.

Still having reboots, I think I will change the PSU since it's making a bad noise the fan sounds like its going bad and hopefully it gives better voltages since the GX2 is pretty meh as far as PSUs go (hey it was free!)

 

[tabascosauz]

Still having reboots, I think I will change the PSU since it's making a bad noise the fan sounds like its going bad and hopefully it gives better voltages since the GX2 is pretty meh as far as PSUs go (hey it was free!)

Sounds like a plan.

If a PSU doesn't stop the reboots (I hope it does), then the CPU is definitely due an RMA. Have you stress tested the GPU lately so that it pulls its full TDP?

 

[Mazer]

Sounds like a plan.

If a PSU doesn't stop the reboots (I hope it does), then the CPU is definitely due an RMA. Have you stress tested the GPU lately so that it pulls its full TDP?

I did a full 3dmark time spy test and superposition 1080p extreme along with some games and 100% at full TDP according to hwinfo.

 

[synapsed]

@Mazer Hello!
Man, I strongly encourage you to test your system for at least a few hours or maybe a few days using a known-good PSU.
Please, if you do not have a spare PSU at hand, that you know for sure it's good, just call a friend or something and borrow one for some tests.

I've been bugged with a very similar issue as yours, when my system would crash out of nowhere, even when idle, with a higher frequency of occurence when the computer was colder, like in the first hour of running.
Most of the times (but not always) it showed me a bluescreen with the message CRITICAL_PROCESS_DIED.
From the moment I switched the old PSU to a new one, the random crashes went away 100%.
For the record, I also use a Zen3 platform (X570+5800X3D) on W11.

 

[Mazer]

Please, if you do not have a spare PSU at hand, that you know for sure it's good, just call a friend or something and borrow one for some tests.

I have two crappy ones (coolermaster masterwatt 750w and some ating 600w thing ew) from other builds I don't trust them with something like a RTX 3000 series I might burn my place down trying them you know. I will have to wait till the new one RMx 750w comes in.

Okay well I capped my FPS to 60 and now in metro exodus it's stable from what I can tell it actually didn't restart from setting a fps cap. I will continue to check this out more with some other games.

 

[Mazer]

Got a new PSU a RMX 650w and it looks like that fixed the problem. It wasn't the 5800x going rather the old PSU was no good also found out that CPU pin wasn't fully attached to the connector

 

[Mazer]

Update: After trying out the new PSU I started to play those games again and guess what? It crashed to reboot again but this time it stayed off and I believe it might of damaged the PSU somehow. I did try it with the paperclip and the fan spins up but I think really it's a bad CPU, I have everything set to defaults with no XMP and I haven't had a crash since. I will switch out the RAM and if that still doesn't fix it, I will have to buy a new CPU I can't take this much any longer really.