Strange problems with newly built PC...

[Warenwolf]

I built recently a second PC for my minions - every component is new, except the PSU:

CPU: AMD Ryzen 5 5600X
GPU: Saphirre 6700XT (running latest drivers, has been tested with older different drivers).
RAM: Corsair 32 gb (no idea of the model)
PSU: Corsair 650W, 80% gold certification, 7 years old. (Updated to 850W brand new PSU)
MB: asrock 550m hdv (chipset drivers from AMD) (replaced with MSI B550M PRO-VDH WIFI)

All drivers are up to date. Bios has been flashed to the latest version.

Anyways, I have tested the setup these two days and have experienced all sorts of crashes, seemingly with no connections.

I have run benchmarks (3D Mark demo) with no issues and then experienced a crash while watching a youtube video or surfing a wiki page. On other hand I played Starfield for hours with no issues and then a crash when being in open area with nothing happening.
I loaded up games from 2012 and experienced crashing after 10 minutes. Temperature and load do not matter.

In all cases the screen goes black and PC tries to restart. Sometimes that restart does not go well and I have to turn off and on the PC. Reason I think the GPU is to blame is because if I toss out the 6700XT and install GTX 1070, the system becomes rock stable.

Any good idea where to start the detective work?
This is the first PC build I am putting together in ages since due to IRL obligations, I rarely use PC outside my work so I am bit out of practice when it comes to error searching.

[CBJ]

The first thing I'd look at would be power. Try different power rails and connector cables, and make sure they're seated correctly, especially as there are (I believe) two of them. Obviously drivers are the other place to look, but I'm assuming you did that already. Making sure you clean out any old drivers, especially for non-AMD hardware would also be a good thing to do. Vulkan in particular gets confused easily if old leftover bits of drivers are left behind, but they can interfere elsewhere too. Beyond that, I think you're looking at an RMA on the card itself, I'm afraid.

[esd]

Also worth checking for hardware conflicts; the northbridge of a previous PC motherboard did not get on with the Nvidia graphics card.

[burger1]

6700xt is pcie gen 4 1070 is pcie gen 3 (mainboard slot should be at gen 4 with your cpu according to the mainboard page)

6700xt pulls more power than the 1070

6700xt uses rebar/sam 1070 doesn't

Any diagnostic lights go on on the mainboard when the system crashes? Might be misleading even if the do.

Using 4 sticks of ram or 2? Turn off docp/xmp,etc... and see if the crashes happen in the faster crashing games.

Try underclocking the gpu memory in msi afterburner. Black screen often seems to be memory.

All the fans on the gpu spin up when the gpu is under load? Vs 1 fan only, etc...

Can install gpuz and see what the gpu memory temperatures do.

May as well press windows key and type memory and run windows memory diagnostic test.

Any hints in event viewer or view reliability history? Ignore kernel power 41.

Any overclocks?

Any funny stuff going on in task manager or device manager?

[Terre]

I use Furmark to test GPU's.

https://geeks3d.com/20230827/furmark-1-36-released/

[Warenwolf]

Thanks for the useful tips people, I have not done this in ages.

So far:

I have tested Furmark: No problems, temps reached 67 degrees when GPU was 97% loaded (1080FHD test)

A friend will come over this evening with PSU which they use for testing at his work (so outputs are known). Frankly I think if PSU was the culprit, the stress testing with things like Furmark would logically triggered the error. In fact - it did not.

Thanks to @burger1 - I used the windows in-built memory test to test the RAM. No issues.

Just for the hell of it - I stress tested the CPU - no issues found there either.

@burger1
No overclocks of any kind, the XMP is turned off.

Event viewer and Reliability Test show me nothing useful...

[burger1]

If your swapping power supplies remember to change out all the cables. People wreck their systems or even multiple hard drives by not using the cables that come with the power supply. So sata power cables, etc... They aren't standardized.

Try the gpu memory underclock in msi afterburner. Might work. Stress tests don't seem to work well always for some reason.

[Alan Phipps]

Probably the only thing I could add has already been said by CBJ. I would do a complete uninstall of your current card graphics driver with all remnants and then a clean install of the latest WHQL driver for it and your OS from the AMD site (not Windows Update).

[Warenwolf]

Well, back at square 1.

Entire day today, according to my minions, the PC was stable (normal use, not gaming but just surfing and such stuff). So in comes the PSU and I link it up. Everything starts normally, no issues.
So after running some more benchmarks, I decided to start up the Starfield and played for two hours, tinkering with new ship.

- Suddenly the PC crashes as it has done so many times yesterday (13).
- Whenever I restart the PC, it crashes again whenever I reach the log on screen or just after I log in.
- I finally change GPU, setting in a old 1070.
- It crashes again (this time I manage to log in and the crash happens when I try to open reliability rapport)
- I log in again, remove the AMD drivers with DDU and reinstall nvidia drivers. This time the PC does not crashes and I can open event viewer, reliability report, etc...

The thing is that I kinda stumped on what exactly triggers the crashes or whether GPU is the real culprit. I have tested the RAM and CPU. And have used the system with old 1070 card before the new 6700xt card was delivered so I would guess that MB is ok? The only thing I am certain is that PSU is not the culprit.

[burger1]

Reload optimized defaults in the mainboard? Try stuff at default values?

You install the chipset drivers and stuff from the mainboards driver page? Might be a win 11 and also a win 10 version?

You reinstall windows when you built the new pc?

Ram speeds supported by mainboard?

For Starfield try capping your fps and/or underclocking the video card in msi afterburner. Other people have issues with Starfield crashing. Can type crash in the search area of the games steam forum. 212 pages of results.

You could technically boot to a windows install usb or maybe even the bios and leave the pc running. If it crashes there then likely a hardware issue. Basically hardware vs software. Could technically install a version of linux but that's nuts.

The gpu have two separate power cords plugged ins vs one pig tailed? Not really pulling enough power for that to matter.

Unplug any game controllers, usb hubs, etc...

Try a lower power power plan if it's set to a high performance one?

Any special software installed? Vs plain windows? RGB stuff, razer software, ryzen master, driver updater software, etc....

Not sure how but the vrms on your mainboard seem ok based on reviews. Mainboard has good reviews.

Win 10 or 11 maybe?

If the windows install is older try a sfc/dism cycle?

Press windows key and type cmd. Right click on command prompt and select run as administrator. Then type sfc /scannow and hit enter. When it's done restart the pc. Then press windows key and type cmd. Right click on command prompt and select run as administrator. Then type DISM /Online /Cleanup-Image /RestoreHealth and press enter. Then reboot the pc when it's done. Run sfc /scannow again.

Antivirus software - unlikely plus usually when launching stuff vs running

Reset the games preferences? Preferences stored in the cloud from another pc being downloaded and applied?

Any voltage offsets, etc... in the bios?

Hwinfo show any hot temperatures at idle?

You can also ask in:

steam/community/discussions/hardware and operating systems

linus tech tips forum

popular - https://www.reddit.com/r/techsupport/

popular - https://www.reddit.com/r/buildapc/

https://www.reddit.com/r/pcgamingtechsupport/

https://www.reddit.com/r/pcmasterrace/

https://www.reddit.com/r/pchelp/

https://www.reddit.com/r/computertechs/

https://www.reddit.com/r/24hoursupport/

https://www.reddit.com/r/buildapcforme/


AMD

https://www.reddit.com/r/AMDHelp/

[Warenwolf]

@burger1

Thanks for many suggestion, will test them when I get home. Note that I have asked regarding help on other fora, I just think that suggestions I get here are more on point despite this not being hardware forum what so ever. Also I am running everything on default values, with everything that can be turned off, turned off just to not complicate issues.

I at least verified that GPU is not the culprit. After plugging in GTX1070 so that I could run some tests while away, before I was done with making coffee same type of crash occured - I had podcast on youtube on, listening on my headphones when same crash happened.
The PC could not restart and had to be started again by turning on and off.
Prior to that (half hour) I run a cinebench (stresstesting CPU again) with no issues what so ever.

Now previously I thought functional build when I slotted in old GTX1070 but I guess that having in 6700xt just triggered the problems sooner.

[CBJ]

That's starting to sound like it could possibly be a memory issue. Try the usual: memory testing tools, and running with various combinations of memory sticks removed to see if you can identify if one of them is the cause.

[Warenwolf]

That's starting to sound like it could possibly be a memory issue. Try the usual: memory testing tools, and running with various combinations of memory sticks removed to see if you can identify if one of them is the cause.

Run the memtest86 (default 4 passes, will test further tonight) without it finding any errors...
Couldn't be anything easy, could it? Sigh....

[burger1]

Not sure how big the ihs/dye is on the 5700x but could try slightly loosening and retightening the cpu cooler to get a different pressure profile. Bigger the cpu the greater chance of issues. Maybe try to not loosen it enough to need to repaste the cpu just to get a slightly different pressure on the cpu. Cpu cooler heatsink might be being pushed down by springs so might not make much difference. Amd cpus can also get popped out when messing with the cpu cooler so need to be careful.

Try using a different video output method like hdmi vs display port, etc... if you have an extra cable handy.

Try using normal pc speakers? Maybe front jacks (usb,audio,etc...) a little flakey? Can technically strip pc down to pc case, mainboard, ram, cpu cooler, gpu. Probably won't help.

Can try 1 stick of ram but it might work just due to it nerfing performance vs being a dual channel ram issue. Single channel (1 ram stick) will nerf your entire system causing lower gpu and cpu performance.

Ram voltage correct in the bios for your ram model?

Any mainboard software installed like asus software suite, etc... that might be tweaking your pc or causing issues?

Reseat the ram and gpu again?

Did you ddu the system when swapping video cards? 1070 crash due to an unrelated issue?

Does the pc crash in safe mode with networking? Disables a bunch of stuff and programs. Can technically run programs with some tweaking but might be stuck watching youtube, web browser.

Psu voltages look ok in the bios. +/- 5%

Could just be a setting somewhere. Rebar on/off maybe?

May as well run a hard drive scan to see the smart data for the drives. Crystal disk info or if you have a samsung drive can use samsung magician. At that point you will have run all the normal tests.

Try an older bios that supports your cpu? Need to read bios notes before doing stuff. Sometimes can't go backwards or there's caveats. Mainboard might not have a flash bios button so might need to be extra careful.

Lots of case fans? RGb hub?

[Alan Phipps]

This is a long shot - I had fitted new additional memory, a pair of DDR4 DIMMs of identical spec/make/type to the pair already installed and rated and set for 3600 MHz operation at which the system had been quite happy. I then started having several in-app failures. I thought the new RAM DIMMs might be less stable than the old ones and so dropped the system memory speed to 3200 in BIOS and I have not had any issues since.

[Warenwolf]

This is a long shot - I had fitted new additional memory, a pair of DDR4 DIMMs of identical spec/make/type to the pair already installed and rated and set for 3600 MHz operation at which the system had been quite happy. I then started having several in-app failures. I thought the new RAM DIMMs might be less stable than the old ones and so dropped the system memory speed to 3200 in BIOS and I have not had any issues since.

They are running at 3200 and are of type specified by motherboard maker to be compatible (although those lists are not the written in stone).

I have also tried most of Burger1 suggestions (where relevant) and no effect either.

Try using normal pc speakers? Maybe front jacks (usb,audio,etc...) a little flakey? Can technically strip pc down to pc case, mainboard, ram, cpu cooler, gpu. Probably won't help.

Have already stripped everything. Only the rodent and keyboard are connected. Everything that is disabled has already been disabled or disconnected

Can try 1 stick of ram but it might work just due to it nerfing performance vs being a dual channel ram issue. Single channel (1 ram stick) will nerf your entire system causing lower gpu and cpu performance.

Can test but that is something a memtest would trigger too I think. Also the problems would be triggered more often perhaps than they are?

Ram voltage correct in the bios for your ram model?

Yes
Any mainboard software installed like asus software suite, etc... that might be tweaking your pc or causing issues?
Nope

Did you ddu the system when swapping video cards? 1070 crash due to an unrelated issue?

Yes ddu is used whenever I take out a GPU and put in the another. 1070 crash might be due to to totally unrelated issue but it also might be related to it being gen3 PCIe and using less power. thereby triggering less often the problem

Does the pc crash in safe mode with networking? Disables a bunch of stuff and programs. Can technically run programs with some tweaking but might be stuck watching youtube, web browser.
This I have to test

Psu voltages look ok in the bios. +/- 5%

Yes
Could just be a setting somewhere. Rebar on/off maybe?
Turned off. Also other stuff like POB and XEM.

May as well run a hard drive scan to see the smart data for the drives. Crystal disk info or if you have a samsung drive can use samsung magician. At that point you will have run all the normal tests.
Done that

Try an older bios that supports your cpu? Need to read bios notes before doing stuff. Sometimes can't go backwards or there's caveats. Mainboard might not have a flash bios button so might need to be extra careful.
Can't do that. Caveats and risks as they say
Lots of case fans? RGb hub?

[burger1]

No bluescreen messages just a black screen?

Avoid using display port if you can?

Other options might be thermal expansion/contraction. Pc part heats up a bit then stops working.

Hard drive temperatures ok? SSD getting very hot?

Have you tried maybe a -500 gpu memory and core underclock in msi afterburner? Set power limit to 70%. Make sure you hit apply.

Black screens are often pc or gpu memory.

Cpu cooler mount/back plate too tight? Is it bending/crushing circuits in the mainboard around the cpu? Mostly of concern to mainboards with fewer layers.

Disabling windows reboot on critical errors unlikely to help and might be dangerous.

Pc in a cabinet or some place the psu can't circulate air?

Unlikely to be the power bar.

Pc use air or water cooling? If air can try the pc on it's side. Gpu sagging? Slot slightly broken? Be careful and make sure not to break stuff.

Turn off windows fast start up and shut down your pc and start it again. Don't think it will help but it's a bad setting anyways since it can cause issues to persist through shutdowns.

Power and sleep settings/additional power settings/choose what the power buttons do and uncheck the box next to turn on fast startup.

Can disable turbo I guess. Set maximum processor state to 99. Going from 100 to 99 is a big step down.

Settings > System > Power & sleep > Additional power settings > Change plan settings > Change advanced power settings > Processor power management > Maximum processor state

Undervolt gpu in msi afterburner a bit? Undervolt cpu in the bios voltage offset?

A bad overclock might give your results but you aren't overclocking.

On reboot does the memory train again and no messages?

Do the diagnostic lights starting sequence get stuck on anything during reboot.

Can run userbenchmark and link results webpage but it's quite unlikely to help.

Have a beep speaker maybe you can plug into the mainboard to get beep codes ? Not sure they still have beep codes.

Unplug reset and power front panel buttons from mainboard help? Jump start the mainboard by the pins? Some mainboards have a start button on the mainboard.

Any noticeable breaks in the main power connection going to the mainboard? Meaning mainboard connector metal pin on the mainboard slightly broken.

If no mechanical hard drives does the pc turn off if you tilt it back and forth a bit. Something making intermittent/insufficient contact?

Load on any of the mainboard headers exceed 1 amp total?

[Warenwolf]

To update this thread, if nothing else to help those that encounter same problem.

The PC started crashing in the safe mode, regardless of what component I had installed (changed GPU and RAM).My gamble is that MB is the culprit so I ordered a new one. To be honest, I am not willing to experiment more with that because in olden times faulty MB could take your other components and I am playing it safe.

I was supposed to get new MB yesterday but delivery services are notoriously unreliable around here even when operating within 3 hour delivery window so perhaps this evening I can see whether my theory about faulty MB is right.

Meanwhile I installed the 6700xt card in old rig and so far no issues (although it is limited by PCIe being a generation older - not that I notice)

[Terre]

To update this thread, if nothing else to help those that encounter same problem.

The PC started crashing in the safe mode, regardless of what component I had installed (changed GPU and RAM).My gamble is that MB is the culprit so I ordered a new one. To be honest, I am not willing to experiment more with that because in olden times faulty MB could take your other components and I am playing it safe.

I was supposed to get new MB yesterday but delivery services are notoriously unreliable around here even when operating within 3 hour delivery window so perhaps this evening I can see whether my theory about faulty MB is right.

Meanwhile I installed the 6700xt card in old rig and so far no issues (although it is limited by PCIe being a generation older - not that I notice)

My fingers are crossed for you. (Metaphorically, as else I wouldn't be able to type this.)

[burger1]

The weighting is probably on the mainboard. Might be the gpu slot or something? Maybe you just got lucky putting in the gpu that one time? Buying mainboards is the hardest component to buy in terms of reliability. They never seem to have great review stats.That mainboard seemed ok reviews wise.