Very specific crashing issue in some games, extremely detailed info provided

[andydsm]

Hey everyone,

 

The time has come for me to officially ask for help, because the last few months, I have completely exhausted all possible ways I've come up with to try and resolve my problem at hand. Before anyone decides to dive into it, I want to warn you beforehand it is highly specific and would require some time and patience just to read my explanation, let alone suggest something afterwards . So this post will be … well .. very long and detailed, in order to explain the problem as clearly as I can, so here goes...

 

Short summary of the core issue:

 

Crashing to desktop without any error, sound, nothing at all, at completely random intervals, while playing a specific list of games (could happen after 5 or 40 mins of gameplay, completely random occurrence). Experience is similar to when you hit Alt+F4.

 

Affected games, discovered so far:

- Watch Dogs 2, Forza horizon 4 (demo), Battlefield 5 beta, Wolfenstein II the new colossus, Anthem (demo)

Unaffected games so far:

- Crysis 3, Battlefield 1 & 4, Far Cry 5, Doom, Dangerous Golf (indie game, using UE4 engine, very demanding)

 

Current rig configuration:

 

Corsair AX860 (PSU)

EVGA GTX 1080 Ti FTW3

Asus Maximus X Hero

Core i7 8700k (Non de-lidded)

16GB kit G.Skill Flare X 3200MHz, CL14

Creative X-Fi Titanium PCI-E (sound card)

NZXT Kraken X52 (AIO)

Samsung 960 EVO M.2 256GB (boot)

Samsung 850 EVO SATA 500GB (games)

WD black 2TB SATA HDD (storage)

Asus MG279Q (monitor)

OS is Windows 10 x64 Pro up to date, all installed software and drivers also up to date, nothing was skipped

 

how the system is set up and tuned:

 

The only manually overclocked component in the entire PC is the CPU. After many hours of testing, I've dialed in the OC at 4.9GHz at 1.27V and just to be 100% safe and stable, I run it at 1.28V instead, cache min/max is set to Auto and not OCd. VCCIO and SA voltages are set to 1.1375V, although HWmonitor reports them a bit higher.

RAM is configured to run at its stock XMP profile and is not OC'd at all.

The 1080 ti is set on slave bios and by using MSI afterburner, power target is maxed to 127% so is the voltage slider, nothing else is changed. Card boosts by itself in the 1950MHz, plus-minus 40MHz range. System is basically rock stable at these settings.

 

Installed additional tuning and monitoring software

 

MSI Afterburner and RTSS - I always use them to see fps and other stats in games

EVGA Precision XOC - set to not autostart, it is only installed when I need to configure something specific on the card

NZXT CAM - set to autostart and always runs, because my Kraken X52 refuses to ramp up its fans without CAM running, only pump RPM goes up.

 

All of these programs (except Precision XOC) were also used and configured the same way, on the previous platform and caused zero troubles on it, I will briefly mention in the next point.

 

A bit of history and the preceding system

 

In late 2018, I decided to swap my previous X99 based system for the current one and at the time I first built the Z370 rig, the graphics card I had was a GTX 980 ti, which I have extensively played on, using the previous X99 platform, with absolutely zero issues and crashes on the titles I have listed as affected (except Anthem demo).

But... here it starts getting weird... When I built the Z370 system and still used the 980 ti I had also no problems with the "affected" games. All issues arose after I simultaneously upgraded 2 things to the new build. One was the 1080 ti and the other was the memory (previously was a Corsair 16GB 3000MHz CL15 kit). I do not have access to any of the previous parts anymore, since they've all been sold, I also have nothing else laying around for testing. All hardware I own is in use in the current system.

Now it gets even weirder … The 1080 ti I bought was used and unfortunately it had an issue, unrelated to the crashing, so the card got RMA'd to EVGA and I got an identical 1080 ti FTW3 as a replacement. After installing the new card, the crashing was still there and happened every damn time, just like on the problematic 1080 ti.

Naturally, I started suspecting the second thing I had swapped as the culprit, the G.Skill memory kit, so this brings me to the next point, the tests and results of troubleshooting the RAM and everything else.

 

Testing done so far to pinpoint the issue, if it is related to hardware:

 

All testing was done, using settings and voltages described above.

Aida 64 stability test, ran without issues for about 1 hour (temps in high 70C)

Aida 64 Cache only test to eliminate CPU cache instability, no issues after 30 mins test, at min/max BIOS cache settings to Auto, no cache OC.

Prime 95 fft min/max at 1344, ffts in-place ticked, well over one hour, problem free with temps in mid 70 C

Firestrike and Timespy pass without issues

HCI memtest - ran at about 800% with 0 errors (at 1.1375V VCCIO/SA voltages)

 

I've also tried upping the VCCIO, SA voltages one step at a time, at up to 1.25V as well as DRAM up to 1.37V and tried the affected games. Only change was the crash to desktop seemed to happen after a slightly longer average period of time, but either way they always crashed.

 

What I tried reinstalling and uninstalling:

 

Using DDU I removed in safe mode the Nvidia drivers after the first and second 1080 ti installs, no change.

Reinstalled Precision XOC (since it's the only new program introduced to the mix) no change.

Tried games with XOC completely uninstalled, no change

Reinstalled MSI afterburner and RTSS several times, no change.

Tried the games with MSI afterburner and RTSS completely uninstalled, no change.

 

The results so far:

 

I have managed to only get Wolfenstein II the new colossus to work by killing MSI afterburner and RTSS, but the others listed still crash to desktop at random.

The games that are unaffected play perfectly and as expected on this rig, without stutters and slowdowns, the affected games also run perfectly, until the moment they decide to crash.

 

Worthy mentions:

 

Watch Dogs 2 always crashes in a specific way, a few seconds before game "gives up" I notice one short stutter, after which game continues to run normally and then crashes to desktop in few seconds.

Unigine Heaven does not want to launch at all... When I double click on the launch shortcut, absolutely nothing happens and no "stuck" process is present in the task manager, processes tab.

The system runs absolutely flawlessly, doesn't randomly hang, crash, stutter or anything... The sole and only issue is the crash to desktop in some games.

 

 

I believe I went above and beyond to be as thorough as possible in my explanation, so please if anybody with some serious knowledge and nerves of steel decides to chime in, I'd very much appreciate it, since I am at a complete loss on what to do next. 

 

Thanks in advance guys!

[TempestCatto]

This is how to write a forum post, well done.

 

I didn't see any attempt to run everything at stock clocks. Have you tried removing every and all OC and reseting the CMOS? That would be the first thing I'd try personally. I remember having the exact same issue, with a list of certain games (don't remember them anymore, was so long ago now) and for me, my system was all AMD. As it turned out for me, some how the AMD drivers were not installing correctly, no matter what I did. I ended up doing a whole fresh install of Windows (7), and that finally worked. But back to your issue, do try reseting the CMOS and removing all OC's. I don't suspect wholly that's the issue, but since it wasn't tried before, it's worth a shot now.

[Radium_Angel]

27 minutes ago, andydsm said:

Thanks in advance guys!

Reset everything back to default. Turn off *all* OCing. Reset BIOS back to default.

Test system/games again. Report back

[andydsm]

Since I had no stability issues, I didn't try to run everything at stock.

I will do that and report back as you guys suggested.

[andydsm]

I just tested watch dogs 2 on optimized defaults in BIOS, game crashed to desktop after less than 3 minutes of play. 

So the problem is till persisting.

[TempestCatto]

Can you try installing the problem games on your OS drive? I'm wondering if they don't like being installed on a drive other than C (or the boot drive for that matter). And until we can figure this out, let's keep everything at stock clocks and stock BIOS, just make sure we have a clean pallet to work with.

[andydsm]

The exact same layout and drive letters from the X99 system was retained.

Even same installations are still in use, (for watch dogs 2, for example) on the current setup. I am pretty sure that shouldn't be the issue.

 

Reverting to bios optimized defaults can be done in under a minute, so no problem to do that when required.

 

[TempestCatto]

I wasn't aware that you didn't do a full on fresh install of everything when you upgraded platforms. At this point, I feel like that's the problem right there. At least, I've never had a good experience whenever I did it. But I was originally suggesting that perhaps they just don't like being installed on a drive that isn't with the OS. In theory it shouldn't be an issue, but there's always that off chance that some files can't be located by the program because of being on a different directory than the OS.

 

The other thing I'd suggest it testing your PSU.

[andydsm]

Yeah, honestly the only thing I've not done was reinstall windows, since I find it an incredible hassle to get everything just right as it was. I've done countless of platform migrations on Win 10 between different PC's, even from Q6600 to sandy bridge never had an issue. In my case the fresh install was done on the X99 system, so I am seriously hoping that isn't the problem and in all honesty it shouldn't be.. Hopefully...

 

PSU is a warranty swap from 2017, haven't seen any failure symptoms from it, could you suggest how to test it specifically?

I think the stress tests load it up as much as the PC potentially can.

[TempestCatto]

As a last resort, do a fresh install. It's gonna suck, but it may very well solve the issue.

 

Using one of these. I doubt it's the issue, but it's something I always test when troubleshooting a situation like this. https://www.amazon.com/HDE-Power-Supply-Tester-PCI/dp/B005UZHB6G

 

[andydsm]

Well Tempest, thanks for your input.

Hopefully there is some easy fix for my problem, without having to reinstall of the OS. As a very last resort, I'll have to man up and do it, if nothing else works.

 

[Evanair]

If you have an extra drive lying around, you could install the new OS to that and disconnect the other drives. This would allow you to test all of the non drive hardware and if the problem doesn't fix itself, then you can reconnect all drives and not lose your install.

[andydsm]

Was thinking about doing this, but I have no spare drives unfortunately.

If no other ideas come through, I'll have to ask to borrow one from a buddy.

[andydsm]

I am starting to suspect my issue is related to faulty graphics card memory, does anybody know how to stress-test VRAM and fill up the entire 11GB of memory?

 

Thanks!

[TempestCatto]

1 hour ago, andydsm said:

I am starting to suspect my issue is related to faulty graphics card memory, does anybody know how to stress-test VRAM and fill up the entire 11GB of memory?

 

Thanks!

https://www.raymond.cc/blog/having-problems-with-video-card-stress-test-its-memory/

 

Give that a look. I found it in an old thread where someone needed to test their VRAM.

[0x1e]

4 hours ago, andydsm said:

I am starting to suspect my issue is related to faulty graphics card memory, does anybody know how to stress-test VRAM and fill up the entire 11GB of memory?

 

Thanks!

I wouldn't think it was the GPU Vram.
You would get some kind of artifact's as the driver interpret the corrupted data incorrectly.

You said you removed all of your OC's in your later posts. Does this include XMP?
To me it sounds like Windows is dumping corrupted data from RAM. It's not a critical corruption so you won't get a BSOD.

I would remove XMP and run stock JDEC specks for now with additional voltage.
Your new RAM has tighter timings which would put more strain on the CPU.

Now you have either exposed a weakness in your IMC or you have cells that are faulty on your RAM.
This can also be a sign of a faulty CPU (although you said that it worked perfectly before),

Either way you will need to fix this sooner rather than later or you will be forced to reinstall Windows.
With each time your programs crash you have opened an opportunity for corrupted data to be written to the disk.
That's my two cents.

[andydsm]

Tempest thanks for the link, I did find it as well, but these apps are all very old and don't work. For example EVGA OC scanner can ill only 4GB of vram to test, so I am looking for something else.

 

0x1e, I can confirm that by loading optimized defaults all OC is reset, including the XMP, so all is bone stock that way.

Indeed, with the previous Corsair kit, I had no issues whatsoever. After running HCI memtest with zero errors, I can't find any other recommended tools for RAM troubleshooting, can you please tell me some other software or a way to test if the RAM is faulty or, God forbid the IMC is dying ? Both components are under warranty, so I'd love to be able to find concrete evidence if one them is failing and present that for RMA purposes.

 

Thanks to all of you guys Hopefully we'll get to the bottom of this!

[andydsm]

Adding another game to the "crash list" Apex Legends, however this game crashes with an error.

I am providing a screenshot of it here. Please if anyone has any idea what is causing this to please chime in. Thanks

[andydsm]

Seems like the problem has been finally solved!

A friend of mine suggested I lookup the offending games' executables in the windows Event Viewer application, since they all crash "silently" without errors and lo and behold... All games had one single common reason to crash! Please see the provided screnshots.

That reason is the system had run out of virtual memory and autos stops the most resource hungry application! Now as embarrassing to admit as it is, I am very pedantic about my SSD usage, so I had disabled windows' page file and virtual memory altogether and that turns out was causing those crashes for all those months!

I just played problem free 3 rounds of Apex Legends till I was killed (it always used to crash right before the characters land) and I also played Watch Dogs 2 problem free for about 40 mins.

I genuinely thought 16GB of RAM would be enough for the system to run games and background apps, but obviously I thought wrong. MSI afterburner was set to report RAM usage as well and it never did come close to 16GB, however the reality is that it obviously did.

 

So in a nutshell, I learnt to not disable the page file and virtual memory, just to leave it alone as it was originally.

[rsgametech]

This little tool can work wonders.Free version worked for me on two occasions. https://www.tweaking.com/

FYI I think Apex Legends i crashing for everyone, I don't think they have enough servers for the amount of people playing.

[andydsm]

1 hour ago, rsgametech said:

This little tool can work wonders.Free version worked for me on two occasions. https://www.tweaking.com/

FYI I think Apex Legends i crashing for everyone, I don't think they have enough servers for the amount of people playing.

Thanks for sharing this tool, however the fix for my issue was a simple setting within windows  

Unfortunately obscure tools built in the windows os, don't get enough credit honestly … After more than a month of troubleshooting my hardware and exhausted all ways to do that, a simple pointer from someone to take a look at Event Viewer, shined the light exactly on the spot, as to why I had the games crash silently. Hopefully, people with turned off virtual memory and page file can find this thread helpful

 

As for Apex Legends, on my side the game hasn't crashed once, after applying the fix. 

[rsgametech]

38 minutes ago, andydsm said:

Thanks for sharing this tool, however the fix for my issue was a simple setting within windows  

Unfortunately obscure tools built in the windows os, don't get enough credit honestly … After more than a month of troubleshooting my hardware and exhausted all ways to do that, a simple pointer from someone to take a look at Event Viewer, shined the light exactly on the spot, as to why I had the games crash silently. Hopefully, people with turned off virtual memory and page file can find this thread helpful

 

As for Apex Legends, on my side the game hasn't crashed once, after applying the fix. 

Ah, well mine is more server disconnects the game itself was running fine. I just got a new laptop with 32gb so i be ballin it!

 

[X_X]

9 hours ago, andydsm said:

I genuinely thought 16GB of RAM would be enough for the system to run games and background apps, but obviously I thought wrong. MSI afterburner was set to report RAM usage as well and it never did come close to 16GB, however the reality is that it obviously did.

What a coincidence, I just posted in this very same section about pagefile maybe less than an hour before you posted your find. Although this guy was capping to 2GB rather than disabling the pagefile altogether.

https://linustechtips.com/main/topic/1033354-memory-leak-help/

 

Good you got it sorted, must have been really annoying having those programs disappear on you. FWIW it's not that you run out of physical memory (DRAM) but the addressing needed to use it. An explanation from about 5 years ago.

[andydsm]

15 hours ago, X_X said:

What a coincidence, I just posted in this very same section about pagefile maybe less than an hour before you posted your find. Although this guy was capping to 2GB rather than disabling the pagefile altogether.

https://linustechtips.com/main/topic/1033354-memory-leak-help/

 

Good you got it sorted, must have been really annoying having those programs disappear on you. FWIW it's not that you run out of physical memory (DRAM) but the addressing needed to use it. An explanation from about 5 years ago.

You have no idea how many hours I've wasted conducting all the stress tests described in my OP and everything in between 

 

I am a hardware and specs guy, and am always blown away how unbelievably complicated software workings are …

My logic dictates: RAM usage below 16GB > no virtual mem and page file is needed. > Done.

But oh how wrong was I, lol … There are a gazillion other things to take into consideration when it comes to software.

[0x1e]

46 minutes ago, andydsm said:

You have no idea how many hours I've wasted conducting all the stress tests described in my OP and everything in between 

 

I am a hardware and specs guy, and am always blown away how unbelievably complicated software workings are …

My logic dictates: RAM usage below 16GB > no virtual mem and page file is needed. > Done.

But oh how wrong was I, lol … There are a gazillion other things to take into consideration when it comes to software.

Thank you for posting the fix. I'll keep this in mind.

I personally have never run into this problem. I have 16GB ram and have not used a page file since 2006ish.
If you ever do further work on it, it would be good to know what triggers the issue, as games shouldn't be doing it.
Games are the window that's in focus and so should never be put into the page file.