Nawrot thanks for your info.
Im planning to get a PSU for testing. But It might take a while. Ive limited Unreal Editor to 60fps and Realtime OFF with no success. I dont have any Unreal based software rather than the Editor itself and Twinmotion (with other games Halo, Apex, Flight Simulator, all of them on Ultra Settings, no problems in multi hour sessions).
Then you have some hardware issue. Depends on prices of parts and labor it may be cheaper to give PC to some service shop. You may potentially (with some bad luck) replace all parts before you find faulty one, se servcie may be cheaper.
If furmark runs fine…
Clean windows install, download and install the graphics drivers. Reboot.
Then install the engine and test.
Eliminate software before jumping to a hardware conclusion when all the tests claim the hardware is ok…
On that note…
Have you ran a check disk (chkdsk) and a sfc /scannow to rule out hard drive corruption?
It’s very possible though you’d see it in your windows log as a critical error…
Just go to an expert computer repair shop.
It will save you time and effort. They have spare parts to test each part.
So they can test which one is faulty.
I still believe it’s a hardware issue. Faulty power supply, ram, and GPU are causing a mysterious shutdown even they are new. Based on my experience in repairing.
I hope you’re not using a pirate version of Windows 10 or 11 (KMS activation). Windows can detect it, and windows do something weird in the background. Also, the old version of Windows 11 has many issues…
But if it is a software issue you can see it in “windows event logs” as a “critical error”.
Yesterday had a 1 hour AIDA64 Test. CPU max 83C and 80C for GPU. Stressing out all components.
Nothing wrong at all.
I might be with you on the software side, Im going to test disks next!
Thank you for the input.
Well my event viewer shows the KernePower related ID 41 critical error everytime the shut down happens. But theres no info rather than the “windows closed unexpectedly”, so I cant see much there.
Thing is, dont know a good high end hardware shop on my city, but im gonna look for it to see if I can find it.
Lately Im blaming the MoBo, its a Workstation one. Could it be that the MoBo is unable to deliver the power requested by unreal on certain processes? Particularly the proceses that rely on sudden GPU power spike?
Is there a limit for the PCI slot power delivery?
Thanks for your reply.
If AIDA and the other tests don’t cause the power draw to increase it’s highly doubtful that unreal does.
You can always try to purposly overload the system by running furmark alongside any CPU test of your liking…
Maybe the combination of the 2 tests could give you failures…
Gpu will be at peak temp after 15m.
Cpu probably needs a full hour.
Set furmark or whatever gfx test to auto close after 30.
And mind you, to be able to run both tests you will likely need to set the priority of the cpu test to the lowest possible in task manager.
If you get failures like a shut down, then yes the power draw increase is the issue here.
(For what it’s worth, I suggest ordering a new 1000w psu or even better from evga. At this point I have to reason you lost way more than the $150 it costs just in stress / pain of watching tests run…)
Not really.
The only non self powered item is your m.2
Everything else is usually hooked right into the psu.
The latest 3080ti/3090 have 3 dedicated psu cables.
Sata/ide has its own psu cable.
The mobo usually has 2. The atx connector and a supplemental 6 (or 8? Pin) on the top left corner.
If you buy new stuff for whatever reason, remember that DDR5 is new and you can’t put more than 32GB into anything / expect it to perform where it should.
In bios you can maybe limit the draw. But that’s usually done on the card with overlooking tools, so I doubt that could be an issue or even possible for most bios…
What you can change is the Gen that it works on. From auto to a specific gen.
Usually that’s just a recipe for forgetting you did it and cuss ages when you update the system… (could this be what’s hapepning?)
Yes there is, but GFX has those power cables for exactly this reason.
So mobo should be able to deliver all power needed, GPU gets most if not all power trough cables.
BTW. I just had a thought about GPU and power. How do you connect those extra power connectors for GPU. Do you use Y split cable from single power line or TWO different cables from TWO different PSU rails? Try connecting GPU to two separate cables that are not used to power anything else. This is cheapest solution that MAY work.
I am Toms Hardware fanboy, however this is best google result i seen:
They have some suggestions there, maybe something will work.
You also can uderclock CPU and GFX to see if that helps locating culprit (just to test).
Is your main power source 220v-260v or 110v-150v? (power outlet)
If your main power source is 110volts-160volts…
Theoretically, PSU power output is lower than it should be.
The solution is using a UPS power supply or A.V.R.
What country do you live in?
Thank you again.
I had my 2080 GPU directly to the PCI slot, it burned it out.
Then, when the 3080 got here, same thing, only with unreal.
Now I have a Fractal Vertical mount with Riser Gen 3 cable. Had to dig arround for a Bios that gave me the option to lower to gen 3 (I had it on gen 4 previous). Anyway, im still having same issue, nothing changes. Only my debt.
My GPU has 3 cables that are connected independently all the way to the PSU.
A bit confused here… what other way is there to connect a GPU to a motherboard?
I think maybe you should start to try and get support from the maker of your motherboard?
I just bought a voltage regulator and just had a shut down again, while editing a material. It was fine for 4 hours of Unreal Material editing, but again, back tothe begining.
I live in Bogota, Colombia, most PC shops I know are Youtube fix type. So will be my last resort.
Sorry i ment no riser cables or PCI extensions where used at that time.
I have the same problem as you before when I operate my computers in 110v. It’s very frustrating… 
Unreal will try to push your GPU to the limit since it knows you have a high-end GPU.
In the top right corner of the Editor, there is an input field for console commands. Try to enter t.MaxFPS 30, which will limit the fps to 30. So that the GPU doesn’t run at 99% all the time.
Or C:\Program Files\Unreal Engine\4.1\Engine\Config\Windows\WindowsEngine.ini
add this line at the bottom
t.MaxFPS=20
This will limit UE fps in the startup.
Then limit your viewport fps to 30fps
Let me know if it will work.
Thank you Striker.x.
Im going to limit fps as you sugested.
But, yesterday I saw some regulators that would take voltage to 220v from a 110v outlet (with 100W power delivery) . You think this could work?
What did you do when you had that same 110v issue?
Use a 110v psu.
America has the same voltage. Never been a problem…
Hi
I’m having a similar issue with a 2080TI machine.
The system shuts down with a ‘clunk’ (no BSOD) with a blank screen, although the LED strips on the RAM and case power indicator remain on. I have to unplug the PSU before restarting.
It definitely seems to the a GPU issue, and thought it was heat, but I ran a stress test on CPU and GPU and RAM (each separately and together) and it went fine. In the stress test, my GOU temp went to 80C for 60 minutes, but it routinely crashes at 60C in UE. Also tried GPU benchmarking and that was ok also. But in UE5 is is very unstable, even with UE’s default projects ands levels.
I’m going to try dropping the GPU power and limiting the FPS, and see if that works…
Did you resolve your issues?
Hey!
Ive been abstent from this post while performing some test. As you may have read above, like you, I tested each part of my system with various test programs. All perfect, even AIDA 64 60mins, not a single issue until UE engine base programs ran.
My suspicions where true. Ive got my hands on a new MB and the problem went away. Its been a month since I changed to a Asus Pro Art X570 Creator and been working on UE 4.27 and 5 with no problems. It seems that at some point where UE asked too much power from the CPU (instantly) the MB could not deliver and shut itself down. This new board has an extra 12V 4-PIN connector that seems to give the CPU that extra push in order to clear that power spike UE may need during certain CPU demanding tasks.
The only thing that bothers me in your scenario, is that, part of you system remains on after the failure occurs. Its similar to when my old GPU fried. “Normally” the system went down at once, but THAT time the video went first but fans kept working, few seconds in RIP GPU. Take into account that before that I had max out the power limit AMPS of my GPU via Afterburner.
Hope something of this helps you.
Thanks!
That was actually really helpful.
The issue has been effecting two machines, my older unit and the brand new one that I replaced it with.
I replaced the PSU in the older unit, and spread the GPU power across two buses, which I did in both. (Both had previously had all the GPU power coming from one bus.)
I also used Gigabyte’s Aosus App to limit the GPU power availability to 95%.
So far, this seems to have made a big difference, but I guess I’ll see how it goes over the next week or two.
Fingers crossed!
Cheers
