UE4 Crashing

[AmazingReflex]

Hi,

I seem to get this crash fairly regular. Game crashes with a message about UE4 crashed, stopped or error (can't remember exactly). It can happen mid race or just at the end.

I never used to get problems with ACC. Has something changed, do others notice more crashing?

I recently upgraded from an RX580 to an RTX 3080. I wonder if that's causing the issue. I could try downclocking the GPU?

 

[AmazingReflex]

Update:

Removed and Re-Installed latest NVidia drivers (no GE experience) and I used DDU to clean first with no network on reboot to stop Windows trying to dwonload it's own drivers from Windows Update.

Re-Installed Fanatec drivers and did not install Fanalab at this stage.

Ran Furmark for 1.5 hours no issues, peaked at 68 degrees

Ran Cinebench R20 for 30 mins multicure test no issues.

Down clocked my MSI Gaming X Trio RTX3080 by 100Mhz using MSI afterburner.

Played for about 30 minutes on ACC no issues.

Made a full backup disk image of C Drive so I can revert back to this state easily.

My plan now is to leave everything as it is for a fair while to see if I have stability. I do notice the fans are picking up a fair bit now whilst playing ACC. I did set my case fans to normal mode so that is speeding them up a bit based on temps and I have the GPU now using the 'cooler / louder' firmware.

 

[AmazingReflex]

Everything seemed OK. I have managed to run ACC multiple times for an hour at a time no issue but I had another UE4 Fatal Error crash after quitting out of ACC last night.

I have downclocked my GPU to boost at 1775Mhz (running 'Low Temperature BIOS') and have my PC BIOS set to defaults with RAM running it's standard XMP profile at 3200Mhz.

I have ACC set to Full Screen, vsync off, no framerate limit. Nvidia CP set to GSync compatible enabled

I am running the latest NVidia driver v461.72.

Struggling here.

I'm thinking of downclocking further to Nvidia FE standard clocks (Boost at 1710Mhz, Mem clock at 19Gbps). My GPU temp is currently maxing out at 65C which doesn't seem high.

I might try rolling back Nvidia driver to v457.51 as I've read a few posts on Nvidia forums complaining about recent updates and v457.51 being more stable.

The thing is I don't remember getting these UE4 crashes when I first got my 3080 back in November. Just seems to be in the last month.

RIght, I'm going to go for the driver roll back option I think. Whilst I'm there I will clear out Nvidia shader cache and Direct X shader caches.

 

[DeakRulf]

I had the same problems (3D Device lost) with several Nvidia drivers with my 2070Super. I also had lots of reproduceable crashes in AM2.
I guess the problem started with 461.xx drivers in combination with one of the last updates, so it happened "suddenly".

So far the latest hotfix driver 481.81 seems to work for me, no crashes for a couple of hours in ACC and AM2.
No guarantee of course...

 

[Deleted member 379375]

Have you looked at the windows event logs also have you considered a faulty harddrive?

 

[AmazingReflex]

Hi,

I didn't realise their was a Hotfix v461.81.

I don't notice any other programs or Windows issues that might point to hard drive failure. I am running on a Samsung 970 Evo Plus NVME drive that is 18 months old with very little use so unlikely I think. Not impossible though.

Last night I did the following:

Ran Disk Cleanup and cleared Direct X Cache.
Rebooted into Safe mode and ran DDU to remove Nvidia driver
Restarted PC and disabled wifi before Windows tried to auto install it's video drivers.
Installed just the Nvidia driver v457.71 (December 2nd)
Downclocked GPU a further 25Mhz (now at -200Mhz)
GPU fan fixed at 70% (this seems to keep GPU temperature below 60C)
GSync still left enabled for full screen (default setting)

I played ACC for 1.5 hours (40 minutes online session and 50 minutes combination of practice and AI racing) = NO CRASHES

I was also running MSI Afterburner and ACC Results Companion throughout. Nothing else other than Steam.

I won't change anything else on the system for a while now and see how this goes. Ultimately I want to revert my GPU back to stock clock and fan control but I will wait for a while to eliminate all the other factors.

I will keep my eye on the Nvidia forums for the new driver releases. I must admit, if it is driver related I am not used to this level of unreliability coming from previously only ever owning fairly old AMD GPU's (R9 280X = 3 years old at time of purchase and RX580 = 3 years old at time of purchase). Is this the accepted norm when buying GPU' that have just been released?

 

[AmazingReflex]

UPDATE:

Reminder, I've rolled back my Nvidia driver to v457.71 (December 2nd).

I did 1.5 hours practice mode on ACC last night plus 30 minutes tinkering with Assetto Corsa = NO CRASHES

I will do some races over the weekend and see how it goes. I've re-done my HDD backup image to capture the current state of my install.

I guess if I get through a whole week of stability then my next step would be to revert my GPU back to stock clock (currently -200Mhz on core).

 

[AmazingReflex]

Things are holding up at the moment.

I did an online race for about an hour on Friday = NO CRASHES

I did various AI races on Saturday and Sunday = NO CRASHES

I'll give it until next weekend and then I'll restore the GPU core clock back to standard (currently -200Mhz) and reset the GPU fan control back to automatic (currently fixed at 70%).

 

[AmazingReflex]

I had no crashes for a week so I have now reverted the GPU back to stock core clock and the GPU fan is now back in automatic.

Since reverting the GPU back up to to stock clock I have done hour long AI races and several online races with no issues.

Still early days. I need to see what the next step is.

I've already proven even with no over clocks using the v461.72 Nvidia driver gave UE4 crashes. Once I removed that driver and installed the older v457.71 (December 2nd) driver I now have stability.

I think I will re-apply the overclocks to the CPU & memory and see if the performance increases are worth it. If it's not much of a significant gain I might leave it back at stock.

The final step I suppose would be to try latest drivers again at some point but I'm a bit put off because of the experience I have had.

 

[AmazingReflex]

I did some testing last night with ACC and AC to compare FPS with and without overclocks on CPU and memory.

Basically, I could hardly notice any difference. Could have been 2-3 FPS either way.

I've returned it back to stock clocks on CPU and memory.

Stock = 4.7Ghz all cores and memory at 3200Mhz
OC = 5Ghz all cores and memory at 3600Mhz

 

[AmazingReflex]

Update:

My UE4 crashes seem to be a thing of the past. I have played for multiple hours with no crashes. Offline and Online.

I believe my issue was the v461.72 Nvidia driver for my MSi Gaming X Trio RTX3080. Once I removed that driver and installed the older v457.71 (December 2nd) driver I now have stability. My PC and GPU are at stock BIOS settings with just the memory XMP standard at 3200Mhz.

At some point I will try the latest Nvidia driver but not for a while as I'm quite happy with current performance.

 

[Sascha Reynders]

I'm so glad I stumbled upon this thread before de-tuning my entire system... I started experiencing the exact same problem more and more frequently these past couple of weeks and thought it had something to do with my overclocks (CPU, GPU & RAM). Rolled back to the 457.30 driver and all seems fine now, still need a few more hours of driving to confirm but so far, so good.

 

[AmazingReflex]

I'm glad you have benefited from this. I have spent hours re-installing Windows and changing things incrementally.

It's interesting you have gone with the v457.30 driver and not the one I am using v457.71. I just looked on Nvidia's site and they only show the last ten versions so perhaps that's why?

If you want access to every previous driver go to Guru3D. Funnily, my stable version v457.71 doesn't seem to be there only v457.51

Videocards - NVIDIA GeForce Windows 7 | 8 | 10

Videocards - NVIDIA GeForce Windows 7 | 8 | 10

www.guru3d.com

There seems to have been a few more new versions recently. Latest is v466.11.

What version were you running when you had the same problem as me?

 

[Sascha Reynders]

It's interesting you have gone with the v457.30 driver

What version were you running when you had the same problem as me?

I went for 457.30 'cos after reading this thread I found a couple of posts on various websites (the GeForce forum amongst others) recommending this driver as one of the last stable and performant ones. I'm still on a 2080 ti so I don't absolutely need the latest drivers. And I'm quite happy with the 457.30 so far, got a slightly higher Timespy score than previous drivers.

I was on 461.09 when I got the UE4 fatal errors.

 

[thepharcyde]

Anyone still getting crashes, as I am. Latest Nvidia driver RTX3080FE / i5 9600KF overclocked to 4.9ghz.

 

[Stepy]

Same issue for me. "Fatal Error File unknown line 198. Unreal engine is exiting due to D3D device being lost. Error 0x887A0006 - Hung"

I get this alot as well so annoying....

 

[flanked9148]

its an old thread but i thought this might help

my audio settings were at 32bit 384khz sample rate
turning it down to 32bit 192khz fixed the fatal error at start-up