SOLVED PC hangs so hard power button isn't working

[AyoKeito]

Hey everyone.
I'm using my PC to game and encode H.264 and H.265 videos. I've been doing that since August and it was fine.
Yesterday, i've left my PC encoding again and came back to see that it's completely frozen. It hang up so hard, that power button wasn't working at all (i don't have a reset on my case but i'm pretty sure i wouldn't have any luck with it as well). Power button on the mobo itself isn't working too. I had to unplug my PSU to shut down my PC. To make things worse, it didn't boot up after that. No mobo debug LEDs were lighting up: only RGB (thank god!), HDD and fans working. I've unplugged my PC to disassemble it later but tried to turn it on 20 minutes later - it worked!.. For some time.

Here's how crash happens: first, mouse cursor changes it's state to "loading something" (i think you know what i mean), a second later screen freezes, USB devices disconnect (RGB goes off) and that's pretty much it. No weird sounds, and screen stays on, like i just made a screenshot, lol. That also allows me to see that my CPU isn't overheating, crashes happen even with 70-85C CPU temperature. There are no events in Windows event viewer regarding that crash, and no BSOD dumps.

Here's what i got: I'm pretty sure that this crash is hardware related, probably CPU temperature or something like that, because it's a 100% crash on the first 10 minutes of H.265 encoding (AVX2, runs CPU very hot, 80-90C). H.264 was crashing too, until i've set up a manual Ryzen Master profile 4000 all-core with 1.23V. [email protected] was a no-go. H.265 is unstable even at [email protected]. Games are mostly fine, i've had just one crash in a few hours of playing R6 Siege with PBO before i've switched to manual profile, i'll test more with it later. So reducing CPU temperature increases stability.

RAM is probably OK. I've been running [email protected] for a month and it was okay. I've had one annoying memory error that showed up after hours of testing with Ryzen DRAM Calculator test, and Prime95 large is fine after 30 minutes of testing (will run a longer test later). I also have some low speed RAM in my other PC so i will try that later too. I've tried running my RAM at 3200 (coupled with IMC 1:1) full auto, that didn't help. On a side note, when my PC hangs up, DRAM debug LED lights up. But i still think it's not RAM, but CPU IMC or something related to mobo. When my PC wasn't booting (see below for a "fix"), debug LEDs didn't show anything even with no RAM installed (i guess that means that boot failed even before RAM test started).

BIOS settings were probably fine. I've manually adjusted a bunch of them to what i think is safe, including all voltages on Auto and all AMD CBS settings on Auto too. That didn't help. I'm now in a process of testing with Optimized Defaults.

PSU is probably fine, but i haven't tested it at all yet. I have other PSU to test, tho.

OS has to be fine since i believe no OS failure can prevent PC from booting and\or reacting to power button.

Other problems i've had before this occurred:
One of my old SATA SSD's disappeared from system a few months ago. It was appearing and disappearing with reboots for a few days but now it's completely gone from OS and BIOS for more than a month. Since it was just a 120GB cheapo cache SSD, i've decided to just ignore that for now.
RAM OC was broken with the last BIOS update for my mobo (nothing above 3200 boots, even on full auto). RAM OC was fine before it, but now i have to OC through Ryzen Master (this one still OC's RAM fine). Also, that one update made my previously 1000% stable profile throw one single error in RAM test. But that update was installed in the end of November, and crashes appeared yesterday.

I believe my motherboard or CPU has gone bad.

Also, there's how i fixed my PC not booting after that crash: i've decided to take it apart again and started unplugging things. After i've removed pretty much everything but 1 RAM stick, CPU and 2 M.2 drives, i've decided to turn my PC on to see if it boots. It did! So i've returned my waterblock to it's place to try again - it didn't boot! I've reseated the waterblock again and applied less pressure now, only 1\2 of how long the screws go - now it boots fine. I've checked both CPU legs and socket and they both look OK. Since then, if my PC crashes, i "only" have to switch PSU off and on again. No longer waiting for nothing and clearing CMOS with a jumper to no avail. I'm going to test the BOX cooler if i manage to find the original backplate for my motherboard.

I'm going to take my mobo and CPU to a local shop next week to check which one of them failed, if any, but i'd appreciate your guesses on what's wrong with my system. I got my mobo from a local shop and i'll be ok with replacing it, since i'm not very happy with it anyway; but my CPU is whole other story, since i've bought it from another country. I will change it to 3950X if it's failed but on the other hand i hope it's not CPU because international shipping is not fun.

I'll run pretty much all the test you can suggest and i'll continue swapping parts to be sure it's either CPU or mobo later this day.

Specs:
AMD Ryzen 3900X
ASUS Prime X570 Pro
4x8GB DDR4 Patriot Viper Steel 3733CL17 running at [email protected]
A bunch of HDDs and SSDs, system is running from Samsung 970
Cooler Master Power Supply V750 750W
ASUS 1080TI FE
Windows 10 x64
Custom water loop, no leaks (that would be obvious)

 

[trents]

There are a lot of statements you make like, "Bios settings were probably fine," "PSU is probably fine," "RAM is probably OK." You are making a lot of assumptions without testing the components.

And did you test that SSD that dropped out? Maybe the SSD is fine but some motherboard component it interfaces with is not.

Is the CPU overclocked?

Do you have the latest bios installed?

Have you tried running the system at platform defaults? There should be a hotkey in bios to automatically set all bios parameters to default, even the RAM, which would be at 2400 mhz. I think you need to do this to get sort out whether the problem is a bios setting or a hardware component.

 

[AyoKeito]

I didn't test the SSD, and i can only do that with an external USB adapter. But i will do that when i get home. 3 other SATA devices are working fine.

I've tried CPU in Default mode, with PBO and with manual overclock. I get more stability as i reduce frequency and voltages, so it's currently underclocked and running at lower voltage and frequency than in Default mode right now. But it's unstable in any case. I've tried Default yesterday with Default CPU settings and it crashed almost immediately on that H.265 encode.
I have the latest BIOS installed.
I am running the same H.265 encoding with BIOS default settings right now, but i will only know the result when i get home in 8 hours or so.

I just came back and here are some more details:
Setting "Optimized Defaults" in BIOS doesn't help.
I've also removed dead SSD from the PC and attached it via external USB: it's certainly dead, not detected in CrystalDiskInfo and detected as a "USB Device" in Device Manager. It's also showing up in Disk Management, but it's telling me i have to initialise the disk which instantly throws an error saying "Device is not ready". On a side note, BIOS is reporting normal voltages (i know i can't really trust software monitors): +12V is at 12.076V, +5V is at 4.949V and +3.3V is at 3.280V

Now, i don't have time to do the PSU swap right now, but i'm going to run another RAM test overnight: i want to use both memtest86 and prime95. Which should i leave running for longer? I suspect that Prime95 will eventually "overheat" my CPU into the crash even in large FTT mode, so i think i'll leave it running for 4-5 hours and leave memtest86 running for another ~13 hours.
For tomorrow, i'm going to swap RAM and PSU and see how it goes.

 

[trents]

Report back when you have done more testing. Also, not sure that your motherboard is the best choice for doing heavy lifting with the Ryzen 3900x. It's a mid range board.

 

[Vishera]

Report back when you have done more testing. Also, not sure that your motherboard is the best choice for doing heavy lifting with the Ryzen 3900x. It's a mid range board.

If I were to look anywhere if overclock reset doesn't fix anything, it'd be at the CPU still being starved for voltage because the VRMs can't keep up. A niche and slightly unlikely scenario, but it's worth keeping in the back of the mind.

 

[EarthDog]

4 phase 50A parts...the vrm. That could be it...

P95 blend tests the most memory iirc...

 

[AyoKeito]

I'm now 99% positive it's a PSU.

First of all: i've reassembled my PC on my desk as an open stand, used my old rusty 650W PSU that make strange noises while it works and it allowed me to encode for 30 minutes straight, that's when i got impatient.

I've hooked up my default PSU and PC didn't boot. It got stuck at DRAM QLED and i've tried 3 times. I've tried my old PSU again for sanity check and it worked. It also works right now while i make that post.

Another thing, and that's where i leave my other 1%: i'm using Thermaltake Sleeved Extension cables for all motherboard and GPU connections. The one that was going to 8 pin CPU is burnt. Plastic melted in some places. I suppose that's why i wasn't getting any QLEDs while 8 pin was hooked through it. When i got rid of it, i got at least DRAM QLED from my default PSU. Anyway, that means that both the extension cable and default PSU are now dead. I'm going to get a proper platinum Corsair PSU tomorrow and i'm not going to try my luck with more tests on old PSU.

Now, the only question is if there's any damage that burnt 8 pin cable did that i still don't know about... All the other extension cables are fine, but i think i'll get rid of all of them (the remaining 24PIN) except the GPU ones.

Thank you everyone!

 

[notarat]

I'm now 99% positive it's a PSU.

First of all: i've reassembled my PC on my desk as an open stand, used my old rusty 650W PSU that make strange noises while it works and it allowed me to encode for 30 minutes straight, that's when i got impatient.

I've hooked up my default PSU and PC didn't boot. It got stuck at DRAM QLED and i've tried 3 times. I've tried my old PSU again for sanity check and it worked. It also works right now while i make that post.

Another thing, and that's where i leave my other 1%: i'm using Thermaltake Sleeved Extension cables for all motherboard and GPU connections. The one that was going to 8 pin CPU is burnt. Plastic melted in some places. I suppose that's why i wasn't getting any QLEDs while 8 pin was hooked through it. When i got rid of it, i got at least DRAM QLED from my default PSU. Anyway, that means that both the extension cable and default PSU are now dead. I'm going to get a proper platinum Corsair PSU tomorrow and i'm not going to try my luck with more tests on old PSU.

Now, the only question is if there's any damage that burnt 8 pin cable did that i still don't know about... All the other extension cables are fine, but i think i'll get rid of all of them (the remaining 24PIN) except the GPU ones.

Thank you everyone!

Thermaltake. They'll get you everytime! (or die trying to get you)

Other than the impeccable TT Armor case of yesteryear, everything I've ever bought from TT has actively tried to kill me. I think they were established by the Nazi's near the end of WWII as the V3 vengeance weapon, but that's just my opinion...