Bottleneck woes? or How my PC keeps crashing in a very inconsistent way

Edit: This issue has evolved somewhat so please peep my comments lower down for more info.

Hello guys,

Sorry to use the forums as tech support but I'm at my wits end. I moved my PC via car and now it is set up but with a major issue. In modern games it seems to be running with only the CPU which sits at 100% load and ends up rebooting due to heat (I think?). The GPU never uses more than 30-40% so this sounds like a bottleneck from what I've googling. It wasn't like this before so I have no idea how to get it back to how it was before. It seems like it isnt using my GPU at all and my CPU is killing itself to compensate. How can I fix/troubleshoot it? I've check in the case and everything seems connected. What could it be?

GPU 980Ti

CPU i5 4690k @ 3.50 ghz

Edit: It also seems a bit inconsistent. Hearts of Iron IV which is a strategy game only crashed after about 40 minutes of play where as Battlefield 1 sometimes crashed immediately and sometimes lets me play a little of the game while it lags badly.

I'm very unsure as to this issue overall. Seems to handle net browsing and most other things fine but intense videogames kill it hard.

There might be a problem with the heatsink + thermal paste + cpu.

This has happened to me before, a few times actually. The last time I thought my 6 yr old PC had finally bitten the dust and it was time to Upgradde, turns out it was just poor ventilation and dry up thermal paste on the heatsink wasn't creating a bond between it and the CPU.

Remove and reseat the heatsink on the CPU and reapply with a fresh thin coat of thermal paste around the whole chip, but not too much that it leaks over the sides. Just remember to carefully scrape off the old dry remnants of thermal paste before reapply. Don't listen to those kids who say "just use a pin prick of paste" that don't work. Also clean off caked on dust and crud on a heatsink fins, it can block up a lot air flow resulting in poor ventilation and help trap heat, its like a wrapping a fur coat around a heatsink.

Hopefully that's your problem.

Thanks. I'll definitely give that a go when I next get my hands on my thermal paste but would that account for the low GPU usage? It's hard for me check as when I alt tab to see my monitoring software it all drops as the game isnt the focus but the CPU has been getting real hot. One time I got BF1 working in windowed mode it was using around 30-40% GPU and running poorly.

Edit: just tested with Hearts of Iron and it seemed to run well with CPU heat being like 60ish. So I tried zooming right in on 3 armies (represented by a character) and it was fine. Panned over to more shit and pop. Straight to reboot.

I also noticed that before bios there is a brief black screen that says A2 in the bottom corner. Not sure if thats always been like that but it SEEMS to boot normally after that. Anyone know what that is and if it could be related?

The CPU is the bottleneck, and it just can't keep with GPU because of the heating problem. In my situation the things you described about how some games would run better than others before the computer restarts itself was exactly what was happening to me. It would run idle fine, web stuff, video etc but once I would load a game it would reset after a few minutes. The idle temp of the CPU was anywhere high 50-70 degrees celsius, the system reset itself once the CPU hit 90 degrees because thats the default auto reset if the CPU temperature gets too hot. After I reseated my heatsink and cleanout all the dust, the CPU was at 40 idle and 60 when gaming.

The specs on that PC

AMD Phenom II X6 1090T with stock cooler

GTX 970

other examples.

http://www.pcgamer.com/cpu-temperature-overheat/

http://www.tomshardware.com/answers/id-2208758/bottleneck.html

So it took me until now to get around to swapping out the thermal paste and put in my new cooler. Good news is the CPU works fine and is staying at nice temps under many stress tests. The con is that the issue still happens. So I assumed it might be a GPU issue so I downloaded an old GPU benchmark (Heaven Unigine) and about 2 minutes in the PC rebooted. Thing is the GPU never overheated or even really worked out. It just turned off at some point. A friend suggested it could be PSU but we have no idea how to check that as I dont have a spare PSU or a spare GPU to swap out. If anyone here has any recommendations I'd love to hear it, otherwise at least this update will inform anyone with similar issues that sees this post.

Check every connection, make sure nothings bent, remove your gpu and test, reseat your gpu and test, and make sure all your drivers are up to date and/or roll back to older stable drivers.

Not sure their is a way to test if its the gpu or psu without replacing them

It doesn't sound like a bottleneck or GPU issue to me. I'd say PSU or bios settings got messed up, did you ever do an overclock on your CPU? Could try clearing CMOS.

@oursin_360: @chatmonchies:

For reasons I cannot determine it seems to be working mostly now. I was trying to see what caused crashes and at some point they just stopped happening. I changed the nvidia settings from Quality to Performance and that stopped games from crashing it but when I changed it back to Quality it didnt crash.

I didnt OC knowingly but I was in the bios to look at voltages but it all _seemed_ normal. The only OC is the Intel Boost that is built into i5 4690k cpus. I'd like to OC more but not until its stable.

I'll check the connections again but I did do so intially. When you say check for bent things which bits should I be looking for? The GPU sags a bit but it always has because its massive and heavy.

I've never cleared CMOS before but I've seen it done. Any downside to doing it?

I changed out motherboards recently and didnt need to reinstall Windows fully. Win10 detected the hardware change and did some stuff then it booted as normal. After that I played about 50 hours of Wolfenstein without incident.

Thanks for the comments.

Wolfenstein 2 and Prey both run great but BF1 ran like total dogshit which it didn't used to. I'm doing a total reinstall of that and praying for the best.

Clearing CMOS just resets BIOS to factory settings. If you haven't done any tinkering, then this has no downsides.

But before doing that, I'd do what @oursin_360 suggested and isolate the issue.

1. Remove GPU
2. Run a CPU stress test (e.g. prime95 for 30 min. should do); monitor temps while this is running (abort if it gets too hot)
3. If it passes, great, issue is not your CPU
   1. In this case, the root cause is likely the GPU or PSU
4. If it fails (or temps get too high), try clearing CMOS and retrying the stress test
   1. If it fails due to temperature issues - cooling issue
   2. If it fails otherwise - CPU issue

I guess faulty memory shouldn't be ruled out either. Try running a memtest to rule that out.

Boy this is turning out to be quite a saga. The issue is/was changing that I feel like I need to explain problems it had before vs. problems it has right now. I'm also going to be very specific because if you look at it broadly you don't get the shape of it.

Initial problem/symptoms:

After driving PC from my old place to a friends place (30 min drive) PC would reboot (straight to black and into a reboot. No BSOD or "Windows didnt shut down correctly) when changing settings on the GOG version of SWAT 4 (playing it on default settings yielded no issues).

Hearts of Iron would run fine for 20-30 minute blocks but occasionally reboot. I was able to force a reboot once by zooming in and panning across armies (represented in game by small soldier character models).

Battlefield 1 would run badly and sometimes crash and sometimes not. Ran poorly regardless of crash.

Zero issues running Keep Talking and Nobody Explodes.

In between then and now:

I drove my computer to my new place (3 hour drive but not off road or rough). Turned it on to check internet speed and some games. Seemed to work fine. As recommended above I changed thermal paste but while doing so installed a CPU cooler I'd bought ages ago. Initial boot had some display issues (monitor not staying on) but after two reboots it booted normally without error. BF1 ran badly but at this point I am willing to classify that as a separate issue. With the new cooler I set to attempting an overclock with the help of a friend. Before changing anything in bios I ran IntelBurnTest and RealBench stress tests and the CPU and cooler did well. Once we checked them out bios options seemed to be missing features familiar to my friend so after looking at it for a few hours didn't end up doing anything with it at this point. Attempting to diagnose if my BF1 woes were GPU usage related I went to muck around in Wolfenstein 2 with HWMonitor up so I could see temperatures and GPU load and it did the crash shortly into gameplay. I ran some benchmarks as I described in the above post they crashed the PC until after changing the GPU nvidia perf thing to a setting and back again they didnt anymore. I have no idea why it stopped crashing at that point.

Now:

That leads us to now. As of now the only crash I can force is on a certain level of Wolfenstein if I engage in combat it eventually crashes. It might happen on other levels but this is where it's repeatable. After reseating the GPU and changing which sockets on my PSU the GPU were plugged into (PSU is a modular Corsair) I attempted that same level and it crashed as soon as I fired my gun. The fact it is so inconsistent makes it very hard to diagnose or test. Prey doenst crash, Dishonored 2 doesnt crash, BF1 runs badly but doesnt crash.

Things I have done:

• Reseat GPU and check cables for all parts of the PC.
• Run Memtest, no errors
• Run IntelBurnTest and RealBench both WITH the GPU in and with the GPU removed. Both pass.
• Run Unigine Heaven and Superposition Benchmarks. GPU reaches 100% usage, gets warm. Doesn't crash.
• Run Avast scan. No viruses just "broken registry items" and asking me to buy premium to solve those.
• Reinstall Battlefield 1. No change.

Things I am considering doing

• Swapping the power cables (ie. the PCIe ones that run from my PSU to GPU) out for totally different ones as the modular PSU came with more than I use.
• Clearing CMOS
• Reinstalling Windows 10 entirely
• Taking it to a PC shop (although with it being this hard for me to reproduce I'm not sure it'll work for them)
• Becoming a console gamer

Thank you for reading my novel "Why don't it work?" it should be out in hardback soon. If you have any suggestions I am all ears.

I believe Wolfenstein has crash issues so im not sure thats a good game to test stability

@oursin_360: I played all of it previously without too much drama. It's the same sort of crash I got on all the other instances but its the only thing I can reliably have it happen with now. It's not a crash to desktop it's a full system reboot. Previously it didn't do this so I'm trying to get back to there.

What I mean is. This isnt a Wolfenstein issue, this is an issue with my PC. Wolfenstein just has a level where it is pretty reproducible

The first two are seperate examples of the crash and the third one is the view of the speccy stats as it did so. Also the longest it has lasted without crash on that bit.

Have you checked if your mobo has a BIOS version update with fixes related to your issues?

@chatmonchies: Not as yet. I'll look into that now however it worked for ages without issue with the new motherboard. It was only setting it up again in my new place did any issues emerge.

Reseat everything!

At first post to me it sounds like you moved it and knocked the CPU cooler away from the CPU slightly, I saw the first suggestion was to reseat and repaste the CPU, have you done this? I would then go through and literally disconnect/clean/reconnect everything even if it looks fine. Somethings probably just been knocked/dust has moved into somewhere. Everything is very inconsistent sounding.

I would also reinstall win10 however if this issue is from you moving it I dont see how it can be any kind of software issue.

@nuttyjawa: Yeah since the move I installed the cooler and reapplied paste. I'm fed up for today but tomorrow I'll take out everything but the motherboard/CPU and basically rebuild the pc. Fingers crossed.

Are the games running off your hdd or ssd? I had an issue with dark souls 3 where the data on my hdd got corrupted. You could try moving the games to your ssd or hdd (whichever you haven't tried) and also checking your drives for errors.

Also maybe you have but check if you crash on your stock intel graphics and bypass your gpu entirely.

Your temps all seem fine so my guess is the psu if the HDD and ssd check out.

I'll still suggest you clear CMOS (and update bios version if there is a new update). Just write whatever OC/custom settings you have down and re-apply once you're stable.

@chatmonchies: @oursin_360:I did think this actually but I've already tried it. Games seem to run better on the SSD but I can still reproduce the reboot when Wolfenstein is on the SSD. I did benchmark the CPU without the GPU in but I suppose I could try and see if it reboots without a GPU at all. Assuming it even lets me launch on integrated graphics.

As for CMOS and bios that's on my list too although my mobo doesnt make it easy. It's an H97M-E35 and the manualinstructions for reseting the CMOS just say to short the jumper while the system is off. I've seen people do this with a screwdriver but then it says to afterwards "Open the jumper". Does that mean bridge it then dont leave it bridged when you turn it on?

As for flashing the bios the guide for my mobo honestly makes me think I'll fuck it up. "DON'T FLASH IF YOU DON'T KNOW WHAT YOU ARE DOING!!!!" about sums me up at the moment. On the website all the bios versions are listed as K.X while in my bios the version is listed as E7846IMS V20.10 so I have no idea whether thats up to date or not.

Sounds like you need a new power supply. Either that or the motherboard is shorting out. If you were getting a BSOD that would be hardware issue usually due to drivers or hardware failure, bu since your machine is totally crashing/rebooting as if you hit the reset button that seems to me like a power supply issue, mother board short or CPU/GPU thermals. Unfortunately the only way to figure this kind of stuff out is process of elimination.

• CPU and GPU stress tests are normal? 20 mins long at 100% without a crash?
• Have you used CAM software or CPU-z and GPU-z to record temps while stress testing? What are the temps idle and at load load? Should be under 90c for both, preferably around 60-80ish max load.
• Removed GPU and used onboard graphics? Still crash?
• GPU and chipset drivers up to date?
• Clean install of Windows?
• Have you tried another power supply or tested the connections with a voltage meter to make sure you are getting the correct power? (specs should be on the PSU box) (tutorials on how to do this on youtube and google. look up testing pc power supply)

@stonyman65: I agree with this. PSU getting old or just dying for whatever reason happened to me before.

@hayt: the fact that Wolfenstein is crashing at the same spot really makes me think it is an issue with the game, i think there is a thread on here about a part where it can crash for a lot of people over and over. So that could be a red herring, even if it didnt crash previously.

An update: Took PC to PC repair place. Guy looked it over, checked ram/hdd run stress tests and found it fine but apparently found lots of errors in Windows and viruses and recommended I reinstall Windows. I am suss on how I got viruses as Avast didnt say shit. He said he didn't suspect PSU issues but it sounds like he didn't test it either. He didn't charge be beyond a booking fee so I took it home and reinstalled Windows. Had some weird behaviour with the inbuilt "Reset this PC" in Windows 10 not working so maybe the Windows really was busted. Ended up using a USB boot reinstall. It's worth noting that in the reinstall process the external HDD I had backed up my work onto (a new WD MyPassport) also died. I could be paranoid but maybe PSU killed it via USB? Who knows?

Wolfenstein 2 had some weird behaviour in loading my old save off causing a blank menu but good said thats just a corrupted save file error and a simple fix fixed that. Played about 5 mins of the singleplayer from the start anyway and off goes the PC. Very bummed I load up the specific spot to see if I can force a crash there. This time I cant. I play for about 30 minutes in two 15 minute chunks (interrupted by me plugging in my second monitor and that becoming my primary display for some fucking reason.) With it not crashing on that sequence I load up another and muck around for about 3 minutes and go to the menu to load a new spot and it reboots there again.

I'm gonna take it back to the PC guy and insist he check the PSU and swap it out. I haven't yet been able to force a crash on other games (I haven't sat down and played then for hours at a time) but I am not convinced this is just Wolfenstein. Wolfenstein for whatever reason is easy to cause it. It's the same behaviour I saw initially across all games.

My main worry now is whether a PSU issue like this (which seems to happen suddenly) would even be caught be voltage meters.

@stonyman65: I haven't tried since the Windows reinstall but I have no reason to believe the CPU/GPU stresses would be different. They run both totally fine without issue. GPU has the newest driver thanks to geforce experience. Chipset driver is straight off the mobo website and as above Windows reinstall seems to not really have changed much.

Edit: How could I test for a motherboard short?