Major BUG: GPU-Z causes Windows 10 PC to hard lock

[Jon8RFC]

Just for grins, I've setup an old drive and have a virgin windows10 installation on it, so it's ready exclusively for testing.

Installed Windows, installed steam (did not login yet), updated graphics drivers, rebooted, ran gpu-z, ran steam, logged in, locked up.
So, no games have even been installed on it yet. It's about as clean as it can get.

If you have an equally-powerful graphics card to the 1070 you could loan to me in the interim, I could send you mine for testing, if you think it's strictly that.

It wouldn't be a pure win10 installation anymore, and I don't necessarily want to be wiping this drive over and over, but if you'd like me to toss in the HD5870 and install drivers for it, I can run it in this new computer on the win10 testing drive and see what happens.

And just to rule out steam account-related issues if heuristics scans are more aggressive for some reason--my accounts (one is 14.5 and the other is 5.5 years old) are both entirely clean of any bans of any type for any game, VAC-enabled or otherwise.

 

[W1zzard]

a virgin windows10 installation

i'll try that, maybe a fresh steam install runs the VAC scan right after login

First attempt with a GTX 1050, loads steam perfectly fine. which nvidia driver version do you have instlaled?

Update: GTX 1070 is fine, too

 

[Jon8RFC]

which nvidia driver version do you have instlaled?

425.31

Ok, I did many hours of testing to try to find a rhyme or reason, and I'm left with more questions than answers.

I fired up my old HD5870 machine did a fresh Win10 installation, and surprisingly, it magically no longer seemed to have the issue. I tried and tried, no dice. It often had the lockup issue though, as recent as February 2019, right when I got my new computer.

I pulled the audio card from my GTX 1070 and put it into the HD5870, still no issues.

I wiped the drive, again, for the GTX 1070 and kept the audio card out, and unplugged my printer (which I had forgotten about before), and it still reliably locked up.

I spent a long time playing in CSGO on the HD5870 both with and without the audio card, and still no problems.

I finally decided to do a sort of wonky test. On the HD5870 machine, I loaded up CSGO and left it on the main screen, and had another computer pinging it once per minute to see when it locked up. It hit a lockup at some point, but when I checked on it, it had actually just gone to sleep--I forgot to change power settings. CSGO was still running, so I joined a CSGO deathmatch server, moved around a little bit so as to not be idle, then unplugged my mouse, keyboard, monitor, and used them to shutdown my other computer and had a third, low-power computer pinging the HD5870 once per minute to log when it no longer responds. This went on for over two hours when I finally checked it again, so I thought "maybe it needs a monitor on and connected, I should at least waste some power on that". At some point, it didn't lock up, because when I connected the monitor I was greeted with the following, which I've never seen before in my life:

http://imgur.com/a/GkiSBHh

So, it did have an issue at some point during those two hours, but it didn't lock up the computer. I actually played for half an hour, earlier, and had no problems, so I just gave up and decided to let it sit, at which point it finally had an issue.

Windows firewall is setup to permit CSGO, as it always has been, and pfsense is running the same as it always has, and I've never had problems with CSGO. Taking pfsense out of the equation still permitted the GTX 1070 system to lock up reliably.

This is becoming a nuisance rather than an interesting challenge, because it was repeatable on the HD5870 computer in the past (my new computer refuses to boot with the HD5870, unfortunately), but for some reason it isn't anymore, and ended up with that notification instead. I'm going to now reboot the HD5870, join a deathmatch server again, leave the mouse, keyboard, monitor connected normally, while gpu-z runs in the background, and have another machine ping it for a few hours to see if/when it locks up...maybe the VAC issue earlier was residual from the computer going to sleep. We'll see.

 

[dark12]

I wanted to chime in that I, too, have this issue. It occurs reliably with two monitors, gpu-z on the second monitor, and running Counterstrike: Global Offensive.

It happened reliably on an old q9550, HD5870, Win10, two monitors and CSGO. It also happens reliably with an i7-7700, GTX 1070, Win10, two monitors and CSGO. Just one monitor seems to not be an issue, and I have no problems until I run gpu-z while utilizing two monitors and occasional lock ups which I can't reproduce, but it will always lock up with CSGO. I can't recall if it happened on the q9550 system with Win7.

Made an account to say I have been busting my brain over this for the last month.
Spent countless hours messing with overclocks. Formatted Windows 10. Even replaced my Vega56 with an RTX2070.
Nothing has helped.
I am using triple monitor. CSGO on main screen. GPU-Z on second screen.
Hard locks. Can't turn off num-lock. Can't reset. Must hard power off.

 

[lexluthermiester]

Can't reset. Must hard power off.

Your reset button literally doesn't work?

 

[dark12]

Your reset button literally doesn't work?

Correct.
I need to hold down the power button to turn it off.

 

[Jon8RFC]

I'm glad we have another reliable system to test with!

My test with the (previously reliably locking up) Hd5870 ran last night for about 4 hours, and had no issues. No pop-up either, which was likely spilling over from when the computer went to sleep during the previous test.

Unfortunately, I can't identically-reproduce the situation before, which was dual-monitor on the hd5870. I'm inclined to think that's coincidental, but still would like to test it in the same manner. I used the two dvi outputs on the graphics card, and recently tossed all but one dvi cable. Since this isn't a simple problem, being nit-picky is important since the graphics card knows which outputs are in use. I have just one dvi cable now and will try to find/borrow another.

I have a backup (and crazily paranoid) theory that this is an external, network-borne infection which re-infects the targeted computer even if windows is reinstalled, and it low-level monitors graphical output data for specific triggers (hence why it's not streaming constant video and making any noticeable network impact), before calling home. My old computer has simply "gone dark" for two months, so it's of no use anymore, which could explain why it has no problems anymore. I don't want to dedicate 2 months of using my slow computer again to (dis)prove this theory...and it's a power hog and heat demon.

 

[dark12]

I'm glad we have another reliable system to test with!

My test with the (previously reliably locking up) Hd5870 ran last night for about 4 hours, and had no issues. No pop-up either, which was likely spilling over from when the computer went to sleep during the previous test.

Unfortunately, I can't identically-reproduce the situation before, which was dual-monitor on the hd5870. I'm inclined to think that's coincidental, but still would like to test it in the same manner. I used the two dvi outputs on the graphics card, and recently tossed all but one dvi cable. Since this isn't a simple problem, being nit-picky is important since the graphics card knows which outputs are in use. I have just one dvi cable now and will try to find/borrow another.

I have a backup (and crazily paranoid) theory that this is an external, network-borne infection which re-infects the targeted computer even if windows is reinstalled, and it low-level monitors graphical output data for specific triggers (hence why it's not streaming constant video and making any noticeable network impact), before calling home. My old computer has simply "gone dark" for two months, so it's of no use anymore, which could explain why it has no problems anymore. I don't want to dedicate 2 months of using my slow computer again to (dis)prove this theory...and it's a power hog and heat demon.

Interesting theory, but I am gonna stick with a GPU-Z bug in multi-monitor systems. CSGO itself might be involved.
I can consistently get my rig to hard lock by playing csgo and leaving gpu-z sensor tab open on my other screen.
Sometimes it takes 2 seconds, sometimes 3 minutes. But it never fails.

 

[Jon8RFC]

I can consistently get my rig to hard lock by playing csgo and leaving gpu-z sensor tab open on my other screen.
Sometimes it takes 2 seconds, sometimes 3 minutes. But it never fails.

If you reboot, run gpu-z, THEN run and login to steam, but not open csgo or any other game, about how long does it take to lock up? For me, it varies, but it's typically around 40 seconds, and always under 2 minutes.

 

[dark12]

If you reboot, run gpu-z, THEN run and login to steam, but not open csgo or any other game, about how long does it take to lock up? For me, it varies, but it's typically around 40 seconds, and always under 2 minutes.

About to pass out here. Early day tomorrow.
This bug bothers me so much because when it happens, it often deletes all of my config settings when my pc comes back up.
I understand the importance of being able to recreate the problem though.

 

[Jon8RFC]

I've done hours and hours of testing tonight, and even got my old HD5870 in my new computer (insta-lock, as well) and my new GTX 1070 in my old computer (zero lock ups, angrily). Pulling the graphics card and using strictly the integrated graphics of the i7-7700 also produced quick, reliable lock-ups, on my real Win10 installation, or on a fresh Win10 installation. Frustrating, and it officially has nothing to do with nvidia vs amd dedicated graphics.

Made a little progress, eventually, after doing a variety of things. Steam/VAC does fingerprinting all the way down to which sata port you're using. The best idea I can come up with is that certain hardware IDs (I guess a combo of the hardware itself and hardware configuration, such as which sata ports or IRQs) tied to certain steam accounts get flagged for more scrutinizing scans. Everything I'm doing on my old computer, is a drastically different setup than it was back when I actually used it--different/fewer PCI-E slots are in use, just one hard drive instead of 6 I think, different output ports on the graphics card, and...a different sata port than I used 2 months ago.

I decided to reformat, pull the graphics card from my i7-7700, and it still locked up. Next, on a whim, I used a sata port that I have never once used on this computer, other than before I used steam--all I used it for was to dban wipe the hard drive it came with--sata4. All of a sudden, no lock ups. I moved it back to sata0, and I got lock ups. Moved it to sata1 (where the dvd drive is) and got lock ups. Moved it back to sata4, no lock ups. Then--here's where it gets interesting--I kept my other two drives unplugged, and simply moved my real Win10 drive to sata4, assuming that I would get no lock ups...it DID lock up. I then unplugged it, and plugged the fresh Win10 back in to sata4, and all of a sudden it WAS reliably locking up again. This is f-asterisk-ing ridiculous. I reformatted, reinstalled Win10 on sata4, and it still locked up. I tried my storage drive's port (this drive was in my old computer and is the current storage drive on my new computer, unchanged), sata2, and it locked up. So, my next step was the below...

I'm posting this on a fresh Win10 install on the same testing drive I've been wiping over and over again, on my newer i7-7700, on sata3 (which I haven't used yet at all for testing purposes, and typically holds my old HD5870 computer's Win10 drive until I'm positive I need nothing else from it--which is NOT the current Win10 drive), with the GTX 1070 physically uninstalled, with the hard drive partitioned differently, on a brand new steam account, with a different windows account name setup, and it has not locked up. I've exited steam, closed gpu-z, opened gpu-z and started steam and no lock ups; I've rebooted and done the same--no lock ups.

I'm about to ruin this in an attempt to prove what's happened before. If I try my real steam account on the test hard drive right now, it may or may not work, but that part isn't quite relevant. What I suspect will happen next, is if I plug in my real Win10 drive to sata3 and try my real steam account on there, it will absolutely lock up even if it worked on the test drive. After that, I will then try the test Win10 drive again, I expect that my real steam account will lock up, even if it did work previously on the test Win10 drive; I will then try the newly-created steam account and I suspect that it will begin to lock up as well. Coming up in an edit below...

EDIT:
The real Win10 drive and real steam account did lock up on sata3. I plugged back in the test Win10 drive to sata3 and now the real steam account is locking up, while it wasn't before. The test steam account, however, is still logging on just fine. It has no associated VAC-enabled games on it, so maybe that plays a role. ...I've now added CSGO and ran it once, briefly. Quit and logged on again, no issues on the test steam account.

Also, just for the sake of testing, gpu-z 2.11.0 and older reliably locked up the old HD5870/q5550 machine which seems to (now, but not prior to March) run fine with versions newer than 2.11.0 at this point. I thought maybe I needed to set some type of flag on the server side of things with Steam since I used older versions on that computer with random lock ups in the past two years, but I couldn't get that to work as expected.

@Naki
I have sent an email to the VAC team, but received no response. I'd expect a very tier-1 type response from normal steam support, and they'd never escalate it since I can't provide the right and useful techno-babble. If I can get useful info to W1zzard, I bet they can really get to the bottom of it, and either have steam fix their bug, or have Steam/VAC people help W1zzard figure out a work-around. I've seen that type of collaboration between Factorio developers and AMD, which ended up being Ryzen's 2nd or 3rd "our CPUs are flawed, here's another BIOS update to alleviate it", so I wouldn't be the appropriate one to speak to steam people, even if it's steam's fault.

EDIT 2:
It's Steam/VAC account-related, primarily. The test account works fine on the real Win10 drive and with everything hardware-related back to normal. There must be more aggressive heuristics set to scan on certain steam/VAC accounts, and it's bumping heads with GPU-Z. I've sent a lengthy help request through Steam, and linked to this thread. Hopefully, they can shed some light on this. Good call, @Naki

 

[Naki]

Contacted Steam tech support already? If not, please do!

 

[Naki]

@Jon8RFC -- great! Thanks. (I mean your EDIT 2: part.)

Likely you are not the only person with such problems in recent months.
Even if Steam/Valve is already aware of the issues, if more people post support tickets likelihood of sooner actions/possible fix is higher.

 

[W1zzard]

gpu-z 2.11.0 and older reliably locked up the old HD5870/q5550 machine which seems to (now, but not prior to March) run fine with versions newer than 2.11.0 at this point

this behavior is expected. for some reason vac decides to scan all mapped hardware memory, which causes the crash. i implemented a workaround which should have solved the problem, and works 100% in my testing.

The test account works fine on the real Win10 drive and with everything hardware-related back to normal. There must be more aggressive heuristics set to scan on certain steam/VAC accounts, and it's bumping heads with GPU-Z

just to clarify, you tested on the same machine/installation with two steam accounts, one works fine the other does not?

 

[Jon8RFC]

@Jon8RFC -- great! Thanks. (I mean your EDIT 2: part.)

Likely you are not the only person with such problems in recent months.
Even if Steam/Valve is already aware of the issues, if more people post support tickets likelihood of sooner actions/possible fix is higher.

Well, they weren't able to assist, but they passed it along to the VAC team. The email address the gave to me was the same one I used to send my initial email.

just to clarify, you tested on the same machine/installation with two steam accounts, one works fine the other does not?

Correct.

I just booted up, ran gpu-z, logged in to the brand new test account and it's been fine for about 7 minutes; quit both applications, ran gpu-z again, ran steam and logged in to the test account again and it's fine after more than a minute. Then, I quit both applications again, ran gpu-z again, ran steam again and logged in to my normal account, and it locked up after about 39 seconds from the point I saw the big, main steam window open.
After forcing the computer to power off, I booted up, ran gpu-z, ran steam and logged in to the test account and it's fine after more than 2 minutes. At least it's narrowed down to something being triggered server-side to run locally by steam/VAC.

EDIT:
I associated the test account with my phone number by adding the test account's steam authenticator. It hasn't locked up...yet. Wondering if it might.

EDIT2:
Still no lock ups with the test account.
I've tested while concurrently running MSI Afterburner (I don't have an MSI card, but use it since it does what I need), RealTemp, CPU-Z, HWMonitor, Nvidia Inspector. No lock ups, if that helps narrow anything down about which things are examined by GPU-Z, for future reference.
GPU Monitor gadget does not cause lock ups (also used this gadget app for it).

 

[Xoxoe]

This is incredible. IN-CRE-DI-BLE! I was just about to format my PC and i found this topic literally before doing so!

I bought a new GPU 3 days ago and since then I'm having huge problems with my PC. Screenshot included (critical errors, errors and warnings, its in Polish)

30 instant crashes in 3 days of testing (no BSOD, every single is marked as KERNEL POWER 41(63), almost no details)- I lost my mind and started testing.

My whole PC became totally unstable, I spend 3 days checking everything (changed PSU, tested RAM, changed all slots, cleared everything, unplugged all SATA drives, cleared Windows, updated all drivers including BIOS, tested old and new gpu on HDMI, DisplayPort and DVI, changed power cables, tested new power source from other room, changed monitor, unplugged everything and plugged again, I believe I tested everything and i just wanted to make a final move- format my PC and if that doesn't help, just send my MOBO for warranty). Im literally losing my mind, but PC is like a temple to me.

And now I'm here and I understand... I tested my new GPU with GPU-Z since start and I used Steam/CSGO for testing aswell, all the time!! I passed all tests (OCCT, Cinebench, Heaven Benchmark 4.0, Furmark, even Furmark+OCCT at the same time) and tried to play GTA V once for an hour- zero problems and crashes.

This is incredible and almost unbelievable. Amazing coincidence, because:
1) I have minor problems with electricity at home and at the same day i installed new GPU one lighting bulb exploded in my room (bad installation I guess), and I got worried about PSU...
2) sometimes my monitor and shows the message "OUT OF RANGE", and according to many guides it's because 20pin displayport cable, which transfers voltage and can sabotage the entire computer with this voltage (come on, I already paid and ordered a new one!)
3) I had some BSOD's before and it was displaying the same error named KERNEL POWER 41, but they were random and very rare, like one per month... That's why I thought it's mayby this new GPU and it's power demand, but...
4) I plugged my old GPU again (gtx660 which works perfectly fine) and started testing- there was no diffrence, it was crashing all the time! Doesnt matter if its my new or old gpu, still crashing.

I lost my mind, but finally I'm here. It wasn't my PC. It wasn't me or even my electricity problems. It was just a faulty testing method- CS-GO, Steam and GPU_Z. If you need any informations let me know, I'll visit this topic. I just felt I need to register and tell you my story, it's very important to fix this problem and tell everyone to avoid this testing method, if anyone get's KERNEL POWER 41 error.

Thanks,
XO

PS. Sorry for weak english grammar. I didn't spend much time fixing this text, just wanted to write this asap!

PS2. IMPORTANT NOTE! I noticed that when I ran GPUZ first and then CS:GO and wanted to minimalize the game, there was no GPUZ running in the back. Simply application was down and I had to run it again, I was like wtf is happening and couldn't believe I forgot again, but it wasn't me.
One more important info- sometimes I could play a while with GPUZ running and this can be misleading, but I noticed that using maximum power in NVIDIA panel and changing WINDOWS power plan to maximum made crashes even more often. And last interesting fact- uncapping CS:GO fps (fps_max 999) made the game even more unstable.

PS3. Dunno if this matters but I have installed two csgo anticheats, one for Faceit and one for ESEA, and if I remember correctly Esea was very intrusive into OS and system boot.

PS4. VERY IMPORTANT! A couple of times after crash my graphic drivers went nuts- lost almost all resolution's and couldn't choose my refresh rate (I have 144hz monitor). I had to reinstall drivers or switch to the other DP port. However, I'm not 100% sure its only due to GPUZ crashing- mayby my displayport cable was involved in this.

 

[Jon8RFC]

• Fixed crashes caused by Valve Anticheat

Is the changelog in reference to this particular bullet-dodging bug?
If so, I am so incredibly curious to hear more about the cause and solution, and if Valve was helpful!

EDIT:
Yes, this seems to have resolved the issue, thanks for the fix! I'm very curious to hear about it.

 

[W1zzard]

Is the changelog in reference to this particular bullet-dodging bug?
If so, I am so incredibly curious to hear more about the cause and solution, and if Valve was helpful!

EDIT:
Yes, this seems to have resolved the issue, thanks for the fix! I'm very curious to hear about it.

Yeah that's what it fixes. Never heard from valve, simply worked around what their anti cheat does

Thanks for confirming