Rebooting during games (no BSoD or dumps)

[Dark_Oppressor]

Hello. I am using a computer I built last May (in 2014), and the specs are on my profile. My OS is Windows 7 Professional 64-bit (up to date as of 5 days ago), and my GeForce drivers are 353.30. I have not overclocked or underclocked anything.

The computer ran fine with quite a bit of daily gaming use until last month (June 2015). Since then, whenever I play a fairly graphics-intensive game, my computer will eventually reboot. When it reboots, it is just instantly off and then starting back up again. There is no BSoD, and there are no dumps/minidumps left behind. I verified that there are no BSoDs by setting Windows to not restart automatically.

The crash does not occur if the computer is idling or doing simple things like browsing. I left it running for two days straight at one point doing some downloads, and there was no crash.
It also doesn't crash when playing videos. Especially since it started crashing, I am mainly using it as a video player, as I have it hooked up to my TV via HDMI. I can play videos on it for 8 hours straight and it does not crash.
It DOES crash if I play a game for a little while (about two hours is the most I've ever been able to play before a crash). The game I have been playing the most since the crashing began is Path of Exile. I can play for about two hours, and then it will crash. After that, if I keep trying to play after it reboots, it will begin to crash more and more frequently.
With this information, I believed my problem was probably graphics card related, and probably heat related.

Before performing any tests, I double-checked my build. My PSU is well-rated, and has more than enough wattage for my computer.
To be thorough, I checked my RAM before anything else, since it's an easy check. I ran Memtest86+ for 5 passes (which took about 7.5 hours) and no errors were found.
I then ran both GPU-Z and HWiNFO with logging enabled (one entry logged per second) and fired up Path of Exile, playing until it crashed. It crashed after a little over an hour and a half of play.
Here are the logs: http://cheeseandbacon.org/temp/log-gpuz.txt, http://cheeseandbacon.org/temp/log-hwinfo.txt (I've just got them hosted on my own website for the moment; I was unable to get the forum to let me upload them here)

I cannot find anything in the logs that would indicate the system should have crashed. The GPU temperature was 46 C just before crashing, which to my understanding is fine. The clock speed, voltage, and temperature of the GPU do increase in the last several seconds before the crash, but as far as I can tell they are still perfectly normal values. In fact, all three of these values reach higher numbers at earlier points in the test.

According to my research, the type of crash indicates either a power or a heat problem. As I understand it, the most likely culprits, in decreasing order of likelihood, are the GPU, the PSU, the CPU, and the motherboard itself. My current best (semi) educated guess is that something is just wonky with the graphics card.
Does anyone have any other ideas for me? Or does my conclusion seem valid? Thanks!

 

[RCoon]

I'd say hardware issues with the GPU, or the PSU is tripping balls during heavy load. Shame you can't measure the quality of the power you're getting. Any large deviations from PSU spec would cause a reboot.

Just on a side note, have you tried running a clean installation of some new drivers? That'd be my first port of call.

 

[Dark_Oppressor]

I'd say hardware issues with the GPU, or the PSU is tripping balls during heavy load. Shame you can't measure the quality of the power you're getting. Any large deviations from PSU spec would cause a reboot.

Just on a side note, have you tried running a clean installation of some new drivers? That'd be my first port of call.

How exactly would I measure the PSU situation? I confess I've never done anything like that. I'd love to be able to rule the PSU out (or discover that it is the problem).
I read somewhere that the problem, if it is power related, could also be the actual power in my home just sucks. I'm in an apartment, and don't know how I would look into that. All I know to say is that my computer is plugged into a surge protector, and some other things that are also plugged into that surge protector seem unaffected by the reboots.

As for your sidenote, when the problem first occurred, I was running much older graphics drivers (so I hadn't changed them recently). I did a clean install of the newest ones (353.30) as one of my first steps.

 

[FreedomEclipse]

Agree with Rcoon i do. Personal experience tells me that it could be the PSU going. I had a corsair Hx1050 do the same thing to me after 3-4 years use and i never even pushed that unit beyond 70% of its power rating​

 

[RCoon]

How exactly would I measure the PSU situation? I confess I've never done anything like that. I'd love to be able to rule the PSU out (or discover that it is the problem).
I read somewhere that the problem, if it is power related, could also be the actual power in my home just sucks. I'm in an apartment, and don't know how I would look into that. All I know to say is that my computer is plugged into a surge protector, and some other things that are also plugged into that surge protector seem unaffected by the reboots.

As for your sidenote, when the problem first occurred, I was running much older graphics drivers (so I hadn't changed them recently). I did a clean install of the newest ones (353.30) as one of my first steps.

What's your Surge protector rated at? What sort of fuse (ampage) does it have in it, and what else have you got plugged into it?

What's the total wattage of all the other items plugged in?

It could be that you have a low amp fuse in the surge protector (assuming its a multi-socket one), and all the other things plugged in are taking up most of the wattage its able to provide. After gaming for some time, you might be pulling more power than it can provide with everything else plugged in and turned on, and your PSU is cutting the power to prevent damage.

 

[Pill Monster]

Here's your logfile if it helps....

Hmm I took a couple more pics but HDD they're saved on just decided to reallocate a few sectors, so they're gonne.

Btw you're using about 7GB of RAM, fwiw.

 

[SASBehrooz]

if you have any other pc or friends and family that can help you with , do it. change the gpu and psu and see what happens. this way is the best and easy way to ensure the safety about your hardware.

 

[Caring1]

Those log files show the card running stable and rock solid for around 40 minutes in the middle of that run.

 

[Pill Monster]

Yeah but there's also a big spike right before the system crashed. CPU/PSU also, not just the GPU.


I lost the screenshots.


Btw OP if turbo core is disabled it can help performance, and maybe stability.... The CPU P-States graph looked like a Backgammon board

 

[Bill_Bright]

Sadly (unless my old eyes just missed it) your logs don't show your CPU temps. Rebooting could easily be due to the CPU getting too hot and it shuts down out of self protection, OR because the threshold set in the BIOS Setup Menu is set too low and causes the system to reboot. Since a CPU can cool down about as fast as it can over heat, it boots again.

I do agree to swap out the PSU because stable operation of everything else depends on good, clean, stable power. So eliminating a possible failing PSU is the first step. And while in there, be sure the system is clean of heat trapping dust. Also while in there, inspect the capacitors surrounding your CPU socket to make sure none are leaking or bulging. They look like tall soda cans and if leaking, they will have white to brown dried foam looking stuff oozing out the tops or bottoms. If bulging, that means they about to start leaking. While leaking caps are not as common as they were several years ago, it still happens with some boards.

I also recommend you use a real time temperature monitor such as CoreTemp or Speccy.

 

[Pill Monster]

Sadly (unless my old eyes just missed it) your logs don't show your CPU temps. Rebooting could easily be due to the CPU getting too hot and it shuts down out of self protection, OR because the threshold set in the BIOS Setup Menu is set too low and causes the system to reboot. Since a CPU can cool down about as fast as it can over heat, it boots again.

I do agree to swap out the PSU because stable operation of everything else depends on good, clean, stable power. So eliminating a possible failing PSU is the first step. And while in there, be sure the system is clean of heat trapping dust. Also while in there, inspect the capacitors surrounding your CPU socket to make sure none are leaking or bulging. They look like tall soda cans and if leaking, they will have white to brown dried foam looking stuff oozing out the tops or bottoms. If bulging, that means they about to start leaking. While leaking caps are not as common as they were several years ago, it still happens with some boards.

I also recommend you use a real time temperature monitor such as CoreTemp or Speccy.

CPU never got over 50c....

 

[tabascosauz]

CPU never got over 50c....

Vishera is also notoriously horrible at monitoring temps.

That shouldn't be the problem though, and CoreTemp and Speccy won't do a better job of reading temps. I'm leaning towards the PSU.

"P-States looked like a backgammon board" You can bet that I'll be using that line in the future

 

[Pill Monster]

Vishera is also notoriously horrible at monitoring temps.

That shouldn't be the problem though, and CoreTemp and Speccy won't do a better job of reading temps. I'm leaning towards the PSU.

"P-States looked like a backgammon board" You can bet that I'll be using that line in the future

Lol be my guest.

Vishera has a 90c HTC trip and won't shut down before then.
They rarely get over 55c anyway with Turbo enabled because Turbo is controlled by a package TDP limit programmed into the registers.

 

[Bill_Bright]

and CoreTemp and Speccy won't do a better job of reading temps.

I totally disagree with that. All these software monitoring programs do is read the sensor data and there is nothing magical, proprietary or secret about that - or how to interpret that information either.

So what makes a program good or bad is just a matter of you liking the UI or not. CoreTemp sits nicely in the System Tray and gives you all sorts of display options and information per core or for the hottest core (which is what I use).

Speccy is nice because not only does it now have a System Tray applet, but in addition to CPU temps (per core or average) it also provides motherboard and GPU temps.

The only reason you may see different temps with different programs is due to the sampling time and sampling rate. But if two programs sample the exact same core on the exact same clock cycle (a virtual impossibility), they will report the same temp because they are using the same raw data.

The only thing more accurate than a software program is to point a precision laser thermometer on the device. But since there typically is a big heatsink in the way, you have to rely on software. And looking in the BIOS is really no good either because running the BIOS Setup Menu is one of least demanding tasks we can ask of our systems.

And Vishera is not a monitoring program so not sure your point there. CPUs work with their chipsets to monitor temps and control CPU throttling.

 

[Random Murderer]

Do we know what happens if the CPU is put under load?
The Giga 990FXA boards have a hard time with the 8-core Visheras, I know this first-hand. Enabling HPC and disabling APM in BIOS can help, but your CPU will draw even more power and the VRMs in turn will be warmer as well.
My suggestion: Stress test the CPU with something like OCCT Linpack or Prime95. Watch your temperatures, and make sure your VRM heatsink on your motherboard isn't getting too hot, the board will automatically throttle the CPU back if the VRMs get too hot, which could explain the P-States looking like a backgammon board .
The PSU isn't great, but not terrible either. I agree it could be the cause of your issues, but unless you have another PSU to throw in the machine for testing, it's kind of a crap-shoot diagnosing the PSU.
GPU's temperatures look great according to your log files, I would say the GPU is quite healthy, but that can't be said with certainty until we know exactly what's causing these shutdowns.

Also, as @Bill_Bright mentioned, check the capacitors around the CPU area of the motherboard, and make sure your heatsinks aren't covered and/or coated in dust.

 

[Bill_Bright]

CPU never got over 50c....

Nothing wrong with that.

 

[tabascosauz]

- snip -

I'm sorry if I implied that Vishera was a software. I'm not sure how that would be a possible interpretation of that sentence, but I apologize. Perhaps a little elaboration is necessary. I meant that AM3+ CPUs report wacky temps all the time. There is a separate sensor close to the CPU and the socket that reports inaccurately, but there is reason to believe that this sensor is not the issue. Usually, there are two parts to the problem: somewhat low load temps, and super low idle temps (sub-ambient). There was an article (not just a forum post) somewhere (I think on TH? dunno) that speculated/claimed/said that the problem with the sensor was that the reported temperature number was incorrectly calculated from the distance to Tjunction or whatever AMD calls the temperature limit.

I also must apologize for the other part of my post about Speccy and CoreTemp. Evidently, you thought that it was a criticism of the UI built into Speccy and CoreTemp, and so I must clarify once again. I merely meant that these software-based solutions shouldn't be too far off from each other since they are all software. Unless the temperature fluctuated over the course of two seconds like the P-states, the software temperatures would be about the same.

So much ado over a misunderstanding. And it might be to everyone's relief that I am not so uninformed/mentally challenged as to think that Vishera is the name of a monitoring software. OP owns a 990FX Extreme9. Compared to many other AM3+ boards, this board would have the least trouble handling high power draw through its VRM. @Random Murderer points out something I forgot to mention; while Intel CPUs will throttle (especially in reviews) because the CPU gets too hot, AM3+ CPUs (typically when overclocking) will throttle usually because the VRM has safeties in place not to draw too much power / get too hot (as long as it's not an MSI low-end board with no protections).

 

[Pill Monster]

I'm sorry if I implied that Vishera was a software. I'm not sure how that would be a possible interpretation of that sentence, but I apologize. Perhaps a little elaboration is necessary. I meant that AM3+ CPUs report wacky temps all the time. There is a separate sensor close to the CPU and the socket that reports inaccurately, but there is reason to believe that this sensor is not the issue. Usually, there are two parts to the problem: somewhat low load temps, and super low idle temps (sub-ambient). There was an article (not just a forum post) somewhere (I think on TH? dunno) that speculated/claimed/said that the problem with the sensor was that the reported temperature number was incorrectly calculated from the distance to Tjunction or whatever AMD calls the temperature limit.

Nah it's not a malfunction, AMD designed it that way. There are 2 different sensors, one on the board and one on the package case.
The case sensor has an offset built in to account for the thermal difference inside the die, it's around 25c depending on dissipation.

The only way AMD could get a reasonably accurate die reading at higher temps where it mattered was to sacrifice the lower temp readings. I don't know why but that's how it is.
At ~40c it's supposedly accurate.

Coretemp/AIDA64 etc read the same registers. I've seen Vishera Thermal trip on a couple of occasions with Coretemp running and Coretemp was displaying 89c when it shut down,.
So it's close enough....

 

[Bill_Bright]

We are good. I just want to make sure we and the OP are on the same page.

 

[Dark_Oppressor]

I cleaned dust out when I first noticed the problem, but I just went back in and made doubly sure any heatsinks were clear of dust. I didn't see any bulging or leaking capacitors. I am currently trying to get my hands on a spare PSU to try swapping out.

My surge protector is a Belkin model f9g930-10-gry. The amperage line reads: "MAXIMUM LOAD: 15A 125VAC 60HZ 1875W".
Plugged into it I have: My computer, 2 LCD monitors, a cable modem, a Wii U, a 5.1 speaker system, and a laptop. The laptop has not been plugged in every time the computer has crashed, however. It got plugged in when I started using it to look things up online while my computer was crashing or I had it open to check things.
The wattage of everything plugged into the surge protector is: 850 (computer) + 32 (Wii U) + 55 (LCD #1) + 42 (LCD #2) + 9 (cable modem) + 65 (laptop) + 70 (speakers) = 1123 W

 

[Pill Monster]

Lol. Man I need glasses. I thought the 1hr 35min on that log was 1min 35s.. Only just realised now . Duh..



I'll convert the other logs and post them. What are your bios settings? Default? Idk much about POE but the graph had fluctuations, everywhere,...... vcore, clocks GPU etc.....

 

[Dark_Oppressor]

Lol. Man I need glasses. I thought the 1hr 35min on that log was 1min 35s.. Only just realised now . Duh..



I'll convert the other logs and post them. What are your bios settings? Default? Idk much about POE but the graph had fluctuations, everywhere,...... vcore, clocks GPU etc.....

Yes, my BIOS settings are the defaults. I just got my hands on a COOLMAX CUG-950B (950 W) PSU, and installed it. I'm posting this from the computer in question. I'm going to run the logs and play Path of Exile for a while, and see what happens. I will report back in a bit.
EDIT: Also, after cleaning the computer again earlier, I noticed that the (original) PSU fan was not spinning while the computer was running. Do those not always spin?

 

[tabascosauz]

http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story4&reid=106

Jesus christ

a mark of 0 in performance

"dangerously high <insert an undesirable aspect in any category of performance> above 700W"

that's a first from oklahomawolf

do the testing then throw that thing away bud for the sake of your computer

 

[Dark_Oppressor]

http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story4&reid=106

Jesus christ

a mark of 0 in performance

"dangerously high <insert an undesirable aspect in any category of performance> above 700W"

that's a first from oklahomawolf

do the testing then throw that thing away bud for the sake of your computer

Yikes! Maybe this isn't the best PSU to test with then... I guess I assumed 950 W meant... 950 W lol. Nothing has exploded yet, at least. If this thing crashes again, will it really indicate anything, if this test PSU is so weak anyway?

 

[tabascosauz]

Yikes! Maybe this isn't the best PSU to test with then... I guess I assumed 950 W meant... 950 W lol. Nothing has exploded yet, at least. If this thing crashes again, will it really indicate anything, if this test PSU is so weak anyway?

It appears to be not so bad below 700W. The ripple graphs look OK, dunno about voltage reg.

It's the same platform as the CUG 700W. The 700W was pretty good through the entire test, up until a secondary side capacitor launched upwards like a rocket and hit the fan.