Elden Ring DLC crashes my PC

[Mike_J_Smith]

Hi,

I've had my PC working in it's current configuration for coming up to a year now and havent had any issues.

I started playing the Elden Ring DLC a few weeks ago and again, no issues until a week or so ago and now the game crashes my PC every 30mins-1 hour. It's quite random, sometimes it'll happen once then I'll reboot and it won't happen again all night. Other times it'll be every 10 minutes or so and I'll get annoyed and stop!

The crashes are abrupt, pc cuts completely, black screen - no BSOD or just the game crashing. I have to turn off the power for a minute or so before it'll reboot.

I've not really played any other games recently and I have no issues during normal use.

One thought I had when building the PC, which now has a 7900XT, a AMD 5800X plus a lot of fans / watercooling equipment, was that the PSU wasn't enouugh. It's a Phantek 750w 80 Plus Gold PSU.

I was going to replace it, but thought I'd see what happened. It's worked fine for a while, but maybe it's feeling the strain now? I don't know.

Any thoughts?

 

[Mike_J_Smith]

It should be enough, but one way to test if your PSU is tripping is to just downclock and/or power limit and/or frame cap your card. If reducing it by e.g. 30-40% stops the crashing then that could be a clue, unless e.g. it is related to temperature.

Thanks.

Last night I played with the resolution down to 1080p and all the settings set to low, and low and behold had no issues whatsoever.

I also ran a stress test using OCCT which showed my CPU using 140W, the GPU sitting at around 260W, but with peaks up to 350W. So with all the fans going full pelt, RGB, water pumps, I do wonder if it's possible that the PSU struggles. Then on the other hand it worked fine during the stress test and I set all fans to 100%, so maybe not?

Temps I think were always OK as the system is watercooled an the water temp doesn't go over 36-38.

I'll have a play with the card / CPU power settings and see if I can get it stable. I also moved the PC a few times recently, so may open it up and make sure all the cables are set well.

 

[Mike_J_Smith]

GPUs can have very high peak current depending on workload. Games vary their graphics workloads considerably. This can effectively increase the peak power over very short durations by up to around x2.5. The rating for both GPUs and PSUs is given as continuous power. Unfortunately PSU manufacturers don't generally specific the maximum inrush that their units can handle which makes it hard to be sure that a PSU will cope under all circumstances.

If a GPU demands a peak current that a PSU cannot supply, there will be a rail voltage drop and a typical result is that the GPU no longer communicates properly on the pcie bus. This has the sort of symptoms described. Check the SEL to see if there are any logs which might correlate.

The event viewer has a critical event 41 Kernel Power at the time of the crash, and multiple ones over the past few days and they also look like they are at the time of the crashes. From google that just means the PC shut down unexpectedly / was not shut down cleanly so doesnt really reveal a great deal.

 

[Mike_J_Smith]

An update on this, after playing around with the settings I came to the conclusion it was a PSU thing. Even more when I started playing Black Myth and it kept on crashing. If I undervolted the graphics card / turned down settings then it would be OK, but that didn't feel like a great solution. I ended up getting a 1000w Phanteks and so far have had no issues at all after that was installed.

 

[Mike_J_Smith]

I was going to say 750W should have been enough... but if you had an "OC" 7900XT it could pull 350W.

And "lots of fans + water" probably means the 5800x was on PBO. Which could mean it and the motherboard pulling 150W. Before adding up drives and stuff like sound cards and gfx card power spikes that's 500W.

I'd still be intrigued to know if undervolting the 7900XT and running the CPU at stock would have also 'solved' the issue.

Exactly, the card can pull 330 according to Aquasuite monitor, plus 120+ for the CPU. I saw the pump could be 17+watts. I have 10 fans but forgot I fitted 3 more behind one of the radiators just to see if they would fit and what it would be like, so I had 13 fans and had set a really bad cooling profile that tried to keep the water temp to 35, which for some reason was putting the fans really high, like almost 100%. I’ve fixed that now and removed the extra fans.

So a random spike would probably push the PSU too far.

Yes, undervolting seemed to solve things, but I felt that was a bit of a knife edge that wasn’t worthwhile dealing with.

Anyway, the new PSU is in and seems to be working nicely so I’m hoping that’ll last me a number of years.