Random Crashes (hard lock)

[homerdog]

My new computer sometimes completely locks up, everything freezes except audio (from youtube for instance) continues to play for a while. I have to hard reboot. It's only happened while I'm gaming but I'm usually gaming. It's hard for me to even know how to begin troubleshooting this. I'm testing memory and I removed all the tweaks I made in the BIOS (only very minor changes, slight undervolting) which doesn't seem to have made any difference. Temps are fine. Does anyone have any tips on how I can identify what might be causing this?

 

[Malo]

Tried setting memory to JEDEC instead of XMP?

 

[BRiT]

Regardless of what settings you do end up using, be sure to run MEMTEST overnight from a USB Boot Device to make sure everything is nicely dialed in. Hopefully that is what you're doing now. Also, one of the MEMTESTs does NOT support ECC memory, so if you have that you need to run the other version. There's Memtest86 and Memtest86+ (fully opensource).

OpenSource Memtest86+ @ https://memtest.org/
Freemium Memtest86 @ https://www.memtest86.com/technical.htm

 

[Pete]

You could also try running with just one mem stick at a time.

 

[pharma]

Check your Event Viewer to see if there are errors.
Also try scanning/fixing your disk for errors (chkdsk c: /f /r /x) from dos prompt (as an Administrator).

 

[orangpelupa]

Try

Disable xmp

Manually set memory voltage. For example, my msi b550i motherboard consistently put much LOWER voltage in auto mode. Even when I manually typed the exact voltage, it's consistently 10mV LOWER than what I typed.

If you use display port. Change to use hdmi

 

[homerdog]

Thanks everyone, I've tried most of these things but I'll keep trying.

If you use display port. Change to use hdmi

Why? I have 1 DP and 1 HDMI monitor. The DP one is 144Hz, HDMI is 75Hz.

 

[orangpelupa]

Thanks everyone, I've tried most of these things but I'll keep trying.

Why? I have 1 DP and 1 HDMI monitor. The DP one is 144Hz, HDMI is 75Hz.

Some gpu (uh.... My rtx 3070 LHR) have issues with DP. On Nvidia forums people also has been reporting a bunch of issues.

The weird thing is that the reports are with the LHR version. Not with the regular version.

 

[homerdog]

Some gpu (uh.... My rtx 3070 LHR) have issues with DP. On Nvidia forums people also has been reporting a bunch of issues.

The weird thing is that the reports are with the LHR version. Not with the regular version.

Thanks. I have a 6700XT so I guess this doesn't apply to me.

 

[Albuquerque]

Only happens while gaming? You've already MEMTESTed all the things?

Power supply is going out.

 

[homerdog]

Only happens while gaming? You've already MEMTESTed all the things?

Power supply is going out.

Still not 100% sure about the memory but I turned off XMP and it still crashed. I ran memtest for a while and it found no errors but I didn't let it finish the whole 2 passes. I'll do that tomorow.

For now I turned XMP back on and set the max turbo from 5.1GHz to 5GHz.

As for the PSU, it's brand new and supposed to be decent but I have an old Seasonic S12II 520W that I can try. Gonna be a pain in the ass.

 

[homerdog]

Is there a way to run a memory test on a GPU? For technical reasons I assume not but I thought I'd ask.

 

[Albuquerque]

There's no easy way to do a "memtest for GPU" insofar as I'm aware.

For context, I've run into a very similar problem in my past. My 3930k CPU would run Prime95 at 4.5GHz (a +25% overclock) for hours on end without issue. My heavily overclocked 1080Ti GPU would run the furry donut test for hours on end without issue. I'd run all eight sticks of overclocked DDR3 in quad-channel mode through MemTest86 for hours on end without issue. Never could pin it down to any one problme at all. Then I'd play video games and, depending on the game, sometimes it was stable for hours or would die in minutes.

It eventually came to me when I ran Intel BurnTest and the fuzzy donut thing at the same time. Only a matter of a minute or less and the machine would hard lock.

My trusty 1KW power supply I'd used for many years was apparently no longer so trustworthy. An updated (and actually lower rated) PSU swap later and I was back in business.

 

[homerdog]

Interesting. I ran the Windows memory diagnostic overnight and when I woke up the computer had frozen on Pass 1 (of 5) at 21%. It's never found any errors any of the times I've run it including this time, but this is the first time it crashed before Windows was even loaded. I guess it could be the PSU but I assume it barely uses any power during the memory test.

I will swap out the new 750W Corsair for the old 520W Seasonic and see what happens. This is not as much of an ass pain as RMAing the memory since I don't have any old DDR5 to use while everything is shipping back and forth. It is still unclear to me how I will even know if I've fixed it since it can be stable for hours or it can crash within minutes regardless of what the computer is doing.

 

[pharma]

Is this a pre-built PC or did you assemble all components? If the latter it might help by reassembling everything from scratch.

Sounds like it might be PSU or bad memory stick(s), though I have had freezes happen before that were related to OC'd memory not getting enough voltage.
See what happens if you remove some memory sticks.

 

[homerdog]

Is this a pre-built PC or did you assemble all components? If the latter it might help by reassembling everything from scratch.

Sounds like it might be PSU or bad memory stick(s), though I have had freezes happen before that were related to OC'd memory not getting enough voltage.
See what happens if you remove some memory sticks.

I built it myself. First I'll swap the PSU and if it still crashes I'll start removing memory sticks. Though I don't think the memory is lacking voltage since this happens in bog standard non-XMP 4800MHz mode as well.

 

[homerdog]

Without any special testing equipment is there any way I can tell if the PSU is faulty? My PSU tester will basically say that it works but maybe something bad is happening under load. I really hope it's the PSU. If not I'll be messing with the RAM and possible swapping in my GTX970

 

[Albuquerque]

Wishing you luck

 

[Pete]

Without any special testing equipment is there any way I can tell if the PSU is faulty? My PSU tester will basically say that it works but maybe something bad is happening under load. I really hope it's the PSU. If not I'll be messing with the RAM and possible swapping in my GTX970

Wild guess, but a multimeter checking for voltage droop on whatever rail the GPU hammers (if that’s even how PSUs fail)? Maybe something like MSI Afterburner reports voltages and could record them to see if anything wacky happens when you crash (temps, wattages, voltages)? Things to do while you’re swapping components.

 

[homerdog]

Thanks.

Not sure if my old PSU has dual 8 pin PCIe connectors, which my GPU requires. It has at least one 8pin and one 6pin. Gotta get some kind of adapter for that. It also lacks dual EPS connectors for the CPU but I guess only one of them is necessary for the system to function.