Cyberpunk restarting PC

[TougherLoki]

So a bit of time has pass since my previous post. During that time I dug deep into the issue and tested so many different theories. In short, I fixed it - but it wasn't the solution I thought it was going to be. Skip to end for answer.

1) started by digging through the event viewer and noticed a kernel power event
2) moved over to "view reliability history" since it appears to be a cleaner view of the event viewer data. This showed the same kernel power events, but also highlighted other key events such as a hardware failure event. Also, all events are laid out in a nice timeline. Highly recommend using this.
3) ran GPU, CPU, Power, and Mem stress tests using OCCT. All PASSED. ( which is crazy when you read the fix). But the system would still crash in less than a minute into the Heaven benchmark.
4) downgraded from an RTX 2070 Super to a GTX 1070 hoping a lower power card would help. Still crashed.
4) swapped around memory sticks and ran Windows Memory Diagnostic. Passed.
5) since the system was still crashing I got a screw driver and started taking the computer apart. Replaced thermal paste.
6) with the motherboard out I inspected caps to look for anything noticeably bad. Thankfully, everything was good.

THE FIX: Checked the PSU in detail. I have a modular power supply. One of the sockets used for the GPU cable was 'pulled' out about 1mm. I took a closer look at the cable and noticed part of the plastic pin shroud on the PSU side of the cable was cracked. Swapped to a different cable and avoided the bent socket. Boom! It works. I've been running Heaven for the last 30 minutes. I clearly missed this when swapping GPU's since the issue was on the other side of the cable.

So that's it. A ~$20 power cable and a bent PSU socket. Not what I expected at all. What a wild ride.

 

[Alenheim]

Thought it was only me.

 

[Genie52]

Current "fix" for me is to disable auto-saving (not sure about forum rules about links and all so google cyberpunk disable auto saves - hint you will need user.ini file with some text in it). I noticed that all the time my game was crashing when I was either doing F5 - quick save or game was doing auto save. So decided to disable auto save and when I want to save a game I dont do F5 but save in game menu. Yday played all day with no crashes whatsoever.

 

[Restlessdingo32]

I've held off posting here to get a better perspective of my own crash related issues. Anyway, it may not apply to every case but in mine I think the issue was memory related. I cannot say I've had many crashes or issues with this game. A few crashes, yes. A few full system restarts, yeah. The latter has been rare. For a variety of reasons I suspect it's RAM related.

Symptoms of Issue

1. Infrequent Crashes to Desktop: I'd emphasize the word infrequent here. I haven't had a ton of crashes but they do happen occassionally. Pretty much always with a, "I tried to read data here and couldn't.", error. <-- Noticed a lot of people reporting getting these.

2. Extremely Rare Full System Restart: Again, with an emphasis on, "Extremely Rare.". This by itself doesn't rule anything out, really. Regardless, all hardware appears to pass the kitchen sink of stability/stress test software and nothing is obviously faulty.

3. Extremely Rare WHEA errors for CPU/CPU Interconnect in HWInfo64: So, at first glance this would imply a CPU related issue. I am fairly confident this error can also be thrown due to issues relating to system RAM. Plus, this is with a 3900x configured under a stock configuration with PBO completely disabled. Yes, boards like to do funky stuff with the auto voltage on these. Despite this I am 100% certain the chip is solid under this configuration (extensively stability tested).

4. Error 41 in Eventviewer: Someone can feel free to correct me on this.... I believe this error merely indicates the system unexpectedly and abruptly shut down. It doesn't really... mean anything. Not anything useful anyway.

Previously I had a minor undervolt of ~50mv on the CPU. I know for a fact I can increase this offset to ~100mv without having any stability issues (again, extensive stress testing) and without losing any performance (undervolt a Ryzen 2 too much and it will drop clocks, as I understand it). I did remove this temporarily to ensure it wasn't causing any issues with this game. It resulted in no change.

Solution

I pulled my tin foil hat and clown suit out of my trunk, equipped them, took a shower with my underwear still on, grew a wild hare up my ass and decided to "troubleshoot" the RAM. I figured I'd start simple and increased the RAM voltage called for by the XMP profile by ~50mv (from 1.35v to 1.4v). Yes, I also use XMP. Some of us are lazy.

The thought here is perhaps when this particular game bombards the RAM it results in RAM instability. And, as far as I can tell from monitoring, this game does in fact bombard the RAM (at least, I have a strong suspicion it does). In fact, it appears to hit every major component pretty hard. Perhaps in general, or with a sufficient rise in temps, my RAM under XMP is setup to be right on the stability line and this game is doing something with/to it causing the threshold to be crossed.

So far, no crashes and no system restarts.

I'd revisit the statements, "Extremely rare.", and, "Infrequent.", at this point. Given my issues were both of these things it's a much more difficult problem to troubleshoot. I am changing things around to prevent X from happening every so often. It's not a change you can make and instantly recognize as a fix. In short, I am not 100% confident this is my issue or whether the solution fixed it.

I did engage in an ~2hr CP gaming session the other day and had no issues. I intentionally fit a long drive around the city at absurd speed (180+ on teh highway... the not making any turns soon speed) in there too. This appears to be one of the better ways to make this game slap your system around. Again, no issues.

So my problems don't seem to be showing themselves anymore. If someone is having issues with similar symptoms they might want to look into tweaking their RAM configuration (XMP on/off, frequency, voltage, timings, various other settings related to or with influence on RAM function).

 

[kandynsky]

Try to set your power plan to like this, it seems much more stable like this. No crash in 2 hours.

Are you still able to play without issues? Let us know if that worked.

 

[Arcafonk]

ID41 just say someone pulled the plug out, or that your computer just was reset while doing stuff. Nothing to investigate there.
What says whocrashed ?

 

[kandynsky]

ID41 just say someone pulled the plug out, or that your computer just was reset while doing stuff. Nothing to investigate there.
What says whocrashed ?

That's the only event that shows anything related to this problem. There's no other crashes or BSOD indicated in EventViewer as it seems. The game just makes the PC restarts. Reason? Unknown. All the plausible reasons were already tested. What's left? We're still waiting but everyone seems to be having the same exact issue with different rigs and in different parts of the game.

 

[claudiu111]

i5-4690k / 16GB RAM 1866 / 5700 XT / CX 750w

Not sure if this was posted before as I just saw few people on Reddit having the same problem and couldn't find anything here. Sorry if it's duplicated.

As the title says, Cyberpunk 2077 is making my PC restart every time I try to play it. I'll list what I've checked so far:

- Temps OK
- Usage OK (no struggling while running it at all, just few drops in FPS when looking at some specific lights)
- Stress tested everything, not a single restart.
- All programs closed EXCEPT some security software I need to be running all time (but doesn't seem to flag anything)
- Some suggested limiting CPU usage in the Power Plan Settings, didn't work.
- Some suggested installing on the SSD. Already was.
- Event-Viewer... Getting Kernel-Power ID 41 and it seems that everybody relates that to PSU problems. Another player with the same issue tested 3 different PSUs, 2 of them brand new. Same problem.
- Reinstalling, verifying files...

Don't know what else to do at this point. No other game does this and I've seen worse temps/usage while gaming. Anyone else has any suggestion?

(also just found out that "RTX" config is set to "true" in "bumperSettings.json" under "Cyberpunk 2077\r6\config" even though I don't have RTX. Does that changes anything?)

Post automatically merged: Dec 14, 2020

Just tested some configuration on "memory_pool_budgets.csv" as it was shown on Reddit that it could improve the performance and it actually does... For the less than 5 minutes I get to play until the PC restarts. Nothing changed until now.

Anyone else? Please?

Hi. Do you get an instant reboot or a BSOD before?

 

[Arcafonk]

That's the only event that shows anything related to this problem. There's no other crashes or BSOD indicated in EventViewer as it seems. The game just makes the PC restarts. Reason? Unknown. All the plausible reasons were already tested. What's left? We're still waiting but everyone seems to be having the same exact issue with different rigs and in different parts of the game.

The thing is on each BSOD you've got a 41. Basically, BSOD are 41.

Install whocrashed and tell us what he says, so at least we can have some detail about the mem dump.

 

[kandynsky]

The thing is on each BSOD you've got a 41. Basically, BSOD are 41.

Install whocrashed and tell us what he says, so at least we can have some detail about the mem dump.

There's no BSOD as with all of the other players with the same problem. I repeat: there's NO BSOD. The PC just restarts without a notice and no error message. The logs are empty because there's no BSOD. The PC just restarts. We've checked that already.

 

[Guest 4478408]

undervolt of ~50mv on the CPU. I know for a fact I can increase this offset to ~100mv without having any stability issues (again, extensive stress testing) and without losing any performance (undervolt a Ryzen 2 too much and it will drop clocks, as I understand it).

indeed. ryzen 2 wont get instable when undervolted, but it will lose performance. or rather: i dont fully understand whats going as it will show high clockspeeds in hwinfo, but still produce lower benchmark results.
i messed around a LOT with my cpu when it was new, anything more then -25mv would degrade performance for me. i was not able to find any meaningfull way to oc it neither exept for pbo, and again though my motherboard has ample settings to tweak pbo, the default pbo settings gave me the best results.

as for ram, yes to my experience its VERY common for xmp not working stable without further tweaking. usually i have to add a tiny bit of voltage.
if you are into performance opimization, this guide might interest you

 

[Arcafonk]

The PC just restarts

...a mem dump exists anyway. If the computer crash or restart with no action of the user, memdumps are automatic since windows XP (unless you ticked "i don't wanna mem dump, I wanna save 100kB on my HDD/SSD").

Simple fact : BY THE TIME YOU HAVE A id41, YOU HAVE A MEM DUMP.
Id41 tells "unexpect power failure", put hex code after that (which is an exerpt of the mem dump) and that's it. The PC restarts, it lost its power. WHY it lost its power resides in the memory dump, 70-80% of the time.

Read it, post the info there. Event log is basics of IT, you have a problem that needs second step after basics of IT.

If you only see 41 and no other problem, I doubt it's a cpu powerfailure. Intel CPU goes on stop mode like that, but this write something in the eventlog AND you should have beep codes. AMD "underclocks" (more like puts NOP everywhere, to resync interupts), before frying (since the power is stll high).

I'd smell (pinch of salt) interupt over/underflow from bus desync AKA someone put a huge last gen GPU on a slow mobo (June 2014 for that i5 => which LGA 1150 mobo you have with this 1.5 yo GPU ?), overflows buses, can't feed the cpu, which confise interupts. Then the computer lost its ram.
Whocrashed could say "ram issue" for bus problem (I allready had trouble with my buses which needed a kernel debug - result was shame on me : faulty virtual drive drivers from Win XP, kept in 7 with no harm, before upgrading to Win10).

There is 3 styles of memdump : the id41's arguments, the minidump (which is automatic) and the full memdump that you need to activate in windows before a crash. By default, you should have a minidump. Post the most recent one if you don't want to help yourself. In win10, it is there : C:\Windows\minidump

 

[Guest 4478408]

ntel CPU goes on stop mode like that, but this write something in the eventlog. AMD underclocks before frying

yes but no. big enough voltage drop the cpu wont recover. also psu or motherboard could just reset or shut down the system if their self diagnosis found voltage or temperature parameters to be out of boundries.

I'd smell (pinch of salt) interupt over/underflow from bus desync AKA someone put a huge last gen GPU on a slow mobo, overflows buses, can't feed the cpu, which lost interupts.

hm, maybe you know much more about computers then i do but this makes no sense whatsoever to me.

i had motherboards refuse to boot up at all with gpu that very MUCH older (570x refused to start with a 280). but if it starts there should be no reason why a new gpu on an older motherboard would be instable. nothing has changed about the specifications of pcie if it comes to voltage etc, and the older motherboard also could not be forced into a faster pcie mode by a newer gpu. it simply would run at its maximum speed and thats it.
but of course. a new graphics card in an old system VERY often causes instabilities. but this is usually because the psu can support it, not because the motherboard fails to feed the cpu.

 

[ezdine]

Are you still able to play without issues? Let us know if that worked.

Just got 2 crash today in 10 min...but i was in photo mode and using the hotsampling with crazy resolution for hours so maybe that's why. But since i made this changes, this is the first time. I'll do some more test to confirm if it's work or not because with randomn ctd, the fixes sometime are diffents for each of us.

 

[Guest 4478408]

so start troubleshooting for real. underclock your memory, see if problem persists (unlikely if app ctd), next underclock your gpu and see if problem persists etc.
try to limit the possible source of the crashes. like: if underclocking your gpu removes the cashes, you know its almost certainly either your gpu that gets instable at that frequency, or your psu that cant provide enough power.

there are plenty of great troubleshooting guides on the net. best start with google for a few minutes, youll probably find much better help then in a gaming forum

 

[Arcafonk]

@motorbit 20 years ago, I was supposed to build computer from a pile a sand... Then unemployed, then well, let's activate cheat mode and go to code

Clocks of mobo have an effect on the couple CPU/GPU. If the clock doesn't follow or isn't clean for a various set of reasons (cap failure, for exemple), then your computer will desync any interupt and hard stop like that. Imagine unplugging the ram while the computer is ON : it just don't remember anything and go like "well, ok, I know nothing, I have to boot then".

That is just a theory. The optimization of this engine makes me think that CDPR have gone very low level in their drivers. They migth have recoded some kernel parts, for streaming. I have seen here and on TW3 lots of overflow problems aswell. Some texture flikers. They have and had a lot of dup item bugs. My confidence about how their programmers handle overflows is breached (since years).

(just theory, we need those minidumps anyway. It must be cheesy, since his ram is "slow", there is no reason it should desync. I just don't know if the ram is slow "for that mobo" or if it's topspeed "for that mobo". In the second case, which happens with older mobo, then there migth be something here. You have users that overclocks their stuff with lowest tiers 50 eddies mobo - having lowest of lowest tiers chineese caps, leading to clock signals that don't look square anymore. Put a high perf application with that, that say "oh with that CPU/GPU/ram I can go that speed in theory so let's accelarate" and... poof).

I've got lots of cheesy theories lke this. Maybe cleaning his computer (not if there is "some dust",but if "well, yeah, there is 4 years I should do that"). This maybe creates harmonics in clock signals by having less adapted resistances in the mobo (esp ram slots). Solution : unmount, clean, remount. So yeah, I hope it's not something like this. His CPU's tech is 5.5 yo. Mobo's tech could be 6.5 yo. If there was never any cleaning then someone have to find alcohol, a screwdriver and a vacum cleaner (wouldn't recommend to change thermal paste though). I had id41 on a dusty 8yo computer that I solved with a vacum cleaner, as strange as it is. I say to myself that his GPU is 1.5yo tops, so it should be clean.

About AMD cpu, i edited : they don't underclock, they (must) put NOP everywhere. The frequencies stay the same, but the amount of power the CPU asks is lower since NOP ask not much transistors grid flipping. Until a threshold where it desync and fries (car analogy : push the clutch and the acceleration pedal at the same time, then see how many time it lasts).
I guess they does this because they use less power than Intel, so to underclock they should rely on a much more precise power signal than intel. So instead on counting on PSU signals, they rely on "let's take shortcuts then". They must "clutch the CPU" with NOPs. Intel's ingeneers are paranoid about their CPU. I wouldn't say the approach of the problem is better here or there, it's just funny how two rival team choose different tech (I think Intel did that on their Celeron too).

 

[NatakuARN]

If you have fast startup enabled in windows try disabling it. I solved a lot of system issues with that one setting.

 

[Guest 4478408]

The optimization of this engine makes me think that CDPR have gone very low level in their drivers. They migth have recoded some kernel parts,

i really do not think windows 10 allows that...

About AMD cpu, i edited : they don't underclock, they put NOP everywhere. The frequencies stay the same, but the amount of power the CPU asks is lower since NOP ask not much transistors grid flipping.

this is true and the reason why its always neccesary to validate performance when undervolting ryzen cpu. frequencies might look great but performance still might suffer.
issue with cpu crashing on current drops is that they happen to fast and to deep for the cpu to adjust for it.

 

[kandynsky]

There is 3 styles of memdump : the id41's arguments, the minidump (which is automatic) and the full memdump that you need to activate in windows before a crash. By default, you should have a minidump. Post the most recent one if you don't want to help yourself. In win10, it is there : C:\Windows\minidump

The most recent one, because I don't want to help myself:

Hope to hear your results soon.

 

[Arcafonk]

@kandynsky Activate them and recrash. IF you have an id41 you should have a minidump.

How to Enable Minidump Logs in Windows 10

BSOD (Blue Screen of Death) is one of the worst errors in Windows. BSOD (Blue Screen of Death) error is that error which occurs after system error or

www.faqforge.com

The default setting must have been lost due to a software like ccleaner, or files may have been cleaned off by ccleaner or the like. Temp deasctivate it if it "cleans" permanently.

Also, give me the name of your mobo, please.

Post automatically merged: Dec 30, 2020

issue with cpu crashing on current drops is that they happen to fast and to deep for the cpu to adjust for it.

CPU doesn't crash. I bet on memory or bus issue with 90% certainty. For some reason, if he launch the game, he loose his ram so he reboots (if restarting like this, cpu must be forced to be in start mode).
If it was a CPU problem, he would bsod.
He did not tell he had beeps so the issue doesn't directly comes from the mobo aswell.

Now if that's a 30x multiplug on a lamp line sucking each their parts of amperes everywhere, I can't also tell (or a computer tower placed around electromagnetic magnet of speakers).