PC reboots while playing games, reason uncertain

[yaapelsinko]

Hi.

I have a long time issue with my PC, still can't diagnose though. It reboots through power off/power on when I play games, but there are many details making exact reason uncertain.

My PC consists of:

Core i5 750 2,66@3,33Ghz
Gigabyte P55A-UD3R rev.1, with bios upgraded to F13 version
Radeon 6970 videocard
Thermaltake Toughpower 750W PSU, some of old series

Those are main suspects.

Picture is following. When I play some (not just any) games, at some moment system shutting down for a couple of seconds, and then turns off again.

I have noticed that it happens often (and sometimes pretty fast after game is started) when some massive overdraw is taking place, like when game draw some layered interface (namely Skyrim, videocard's fan goes crazy almost instantly when in-game menu is open), overlapping trees and so on.

I've noticed also that when environment temperature is higher, it happens more often.

Another thing is when I'm downclocking the videocard (for example, set 800MHz and -20% power consuming in graphics OverDrive settings), problem is usually gone (at least it happens rarely enough to remember no last time). Or I could also set the fan speed to be fixed high - it usually helps too.

Okay, looks like videocard just overheats, right? Not that easy.

I've played Hitman: Absolution some time ago, almost highest settings, considering fan noise the temperature must have been about 90 Celsius and higher - no reboots.
And I'm playing WarThunder now, graphics settings almost lowest (except texture resolution and anisotropy) so it makes 120 fps all the time, and it reboots often. But on the second screen I see GPU usage and temperature, and in most cases it is around 50% and 80 degrees, very rarely about 70% and close to, but not higher than 90 degrees. Last reboot it was 77 degrees, then several seconds passed and boom - enough games, go rest.

Just after last reboot I've used Furmark. Videocard heats up to 92-94 degrees, then fan goes faster and temperature stabilizes. 900MHz GPU clock, default power balance, no reboots.

So the issue is not just the GPU overheating?

Maybe overclocked CPU somehow affects? Well I've used Furmark on it too - no problems.
Maybe overclocking through bus clock affects on memory? Memory clock is divided by some factor to be the same 1333 as usual, but possibly there are some latency issues? I've filled it up, all 12 GBytes - no problems.
Maybe PSU wants retirement? 750 watts seems enough, but maybe it just getting old? I've overloaded CPU and GPU together with Furmark, while watching voltage monitor - no deviations, no reboots.



It still looks like GPU problem, not in general overheating, but maybe some local block?

The question is how to diagnose source of problem for sure? It would be funny if I'll replace videocard and find the problem is still there.
Maybe someone had the same issues?

 

[yaapelsinko]

Okay, now it happens while GPU is downclocked to 800MHz, -10% power balance and fan speed fixed at 60%, which results in GPU temperature below 70 degrees (around 67). Default clock settings for CPU and the whole system also was tested - not helped.

It seems that thing getting worse.

That power off/power on thing looks just like when motherboard reboots the system after BIOS tweaks. Perhaps this issue happens because MB decides it need to reboot for safety reason. So how I'm going to find the cause to?

My PC is pretty tight inside, but it is pretty well ventilated, with GPU and PSU fans throwing hot air outside and two big fans pushing fresh air into. 'System' temperature is about 40 degrees, not looks like too hot either. CPU, when idle, shows 29 degrees (Grand Kama Cross, indeed), if I could believe EasyTune6 utility. Gives the idea about air temperature inside.

 

[Automata]

I wasn't expecting Furmark to work for you. If the system powers off, that usually indicates the power supply can't handle the load and is shutting down. Memory and CPU issues generally lock the system or throw a stop error.

If you have another power supply handy, that is the first thing I would try. At very least, measure voltages at idle and under load to make sure voltage isn't dropping too much.

 

[yaapelsinko]

Triyng longer burn-in run with furmark right now. CPU burner assigned to cores 0-2 with 3 threads, Furmark GPU burner assigned to core 3, which results to +99% load on both CPU and GPU.

According to EasyTune6, CPU temp is 70, 'System' (whatever it means) is 51. GPU is at 90 degrees. CPU voltage slightly higher under load, I believe it is because of those SpeedStep and another stuff lowers it when idle. 3.33, 5.05v, 12.3v - not looks like voltages are dropping down, and deviations are just about 1-2% between idle and under load. However, there is four 12v lines in the PSU, so I don't know what exactly voltage for 12v HW monitor shows.

CPU is overclocked again to 3.33Ghz, GPU is at reference 880MHz, but power balance adjusted to +20% (thus instead of 27fps I'm getting 33fps in Furmark, so I assume it really rises the consumption), the system is really heavy loaded now. Anyway, comparing to this, WarThunder's is nowhere with its 50% GPU usage at lowest settings.

35 minutes of test now - still working fine. I think, power is not the issue.

By the way, when I just bought 6970, I have had one issue with it - system froze often when playing online video (and that wasn't happening with previous 8800GTX). It was resolved after BIOS upgrade. Don't know, maybe current problem is another feature between P55A-UD3R and 6970...

 

[Automata]

Software measuring of voltages is useless because it is wildly inaccurate. You will need a multimeter to test from the plugs.

Use RealTemp or CoreTemp to get actual processor temps (they read directly from the sensor inside the processor). It may match EasyTune, I'm not sure. Best to be sure by using a trusted program.

Since the system is randomly crashing, even under heavy load, I would move testing to stability testing. Create a bootable (not the Windows one) Memtest disk, restart, and boot to it. Run that for a pass or two. If that is ok, grab LinX, hit the "All" button, set it to 20 times, and start the test. This won't finish in a reasonable time, so stop it after a few hours. Make sure to have RealTemp or CoreTemp open to monitor temps. This will stress it more than anything else.

 

[Mr.Scott]

Multi rail PSU involved.
Load balance it.

 

[Silver_Pharaoh]

You said" of old series"... how old is that PSU?

Even good ones won't last forever...
Do as thideras said, check the voltages: guide to testing PSU
sounds like a PSU issue, especially if the PSU is old.

 

[yaapelsinko]

Multi rail PSU involved.
Load balance it.

It may be the issue, but not explains why system reboots when loaded at about 50% both CPU and GPU, but works well under Furmark.

 

[yaapelsinko]

You said" of old series"... how old is that PSU?

Even good ones won't last forever...
Do as thideras said, check the voltages: guide to testing PSU
sounds like a PSU issue, especially if the PSU is old.

I don't remember exactly, maybe 5+ years. This model, I believe. Well at least it haven't spent its MTBF yet. 5 years (even 6-7) still less than a half of 120 000 hours.

Thank you for guide, I'll try it, just need to obtain a multimeter.

update: I've found, it was released in 2006, and when I bought it, it still was one of the latest, likely in 2007. So it is about 7 years old now, yay. Needs to be tested, of course, but would be nice if at least some of PC parts could work 10+ years without upgrades.

 

[yaapelsinko]

Software measuring of voltages is useless because it is wildly inaccurate. You will need a multimeter to test from the plugs.

Use RealTemp or CoreTemp to get actual processor temps (they read directly from the sensor inside the processor). It may match EasyTune, I'm not sure. Best to be sure by using a trusted program.

Since the system is randomly crashing, even under heavy load, I would move testing to stability testing. Create a bootable (not the Windows one) Memtest disk, restart, and boot to it. Run that for a pass or two. If that is ok, grab LinX, hit the "All" button, set it to 20 times, and start the test. This won't finish in a reasonable time, so stop it after a few hours. Make sure to have RealTemp or CoreTemp open to monitor temps. This will stress it more than anything else.

Thank you, I'll try those too.

 

[Automata]

To test for a PSU issue, run LinX by itself for longer than it takes before it shut down. Fire up Furmark and see if it crashes immediately. If it does that a few times, then the power supply is certainly the issue.

 

[Mr.Scott]

It may be the issue, but not explains why system reboots when loaded at about 50% both CPU and GPU, but works well under Furmark.

Furmark only loads the GPU.
Run Furmark and Prime 95 at the same time and I can almost guaranty a re-boot.

EDIT- Basically, what Thideras said. lol

 

[yaapelsinko]

Furmark only loads the GPU.
Run Furmark and Prime 95 at the same time and I can almost guaranty a re-boot.

EDIT- Basically, what Thideras said. lol

It actually has 'CPU burner' button, which heats CPU up pretty well. Probably higher than Skyrim, the best observed rebooter for my system yet.

 

[yaapelsinko]

Hmmm I've cleaned PC from dust, and after that it haven't rebooted yet. Although I have no chance to test it with a long run today.

There was a lot of dust, but not enough to make any hot parts to overheat. Again, it always worked well under Furmark and rebooted mostly when running some games that are really not too heavy on the performance.

Some very specific power surge or shortcut caused by dust and consistently reproduced as the dust accumulates? Miracle.

 

[EarthDog]

Furmark only tests the GPU though. When you game, you also use the CPU a lot more heavily.

Oops...We said that already 3 times now...

 

[yaapelsinko]

Furmark only tests the GPU though. When you game, you also use the CPU a lot more heavily.

Oops...We said that already 3 times now...

It actually was able to heat CPU to 70 degrees. Now that I've cleaned the cooler and renewed thermal grease, it is only 60. WarThunder heats CPU up to only 48-50 degrees.

And yeah, you are third time wrong now.

 

[Silver_Pharaoh]

It actually was able to heat CPU to 70 degrees. Now that I've cleaned the cooler and renewed thermal grease, it is only 60. WarThunder heats CPU up to only 48-50 degrees.

And yeah, you are third time wrong now.

Well, it's a new feature it seems, and since we don't use furmark, we never knew about the new "feature"


bottom line is still don't use it!

 

[yaapelsinko]

Well, it's a new feature it seems, and since we don't use furmark, we never knew about the new "feature"

bottom line is still don't use it!

Then you must probably thought 'he must be talking about some new feature we don't know' for three times now, instead of talking of absence or presence of features in software you don't even use.

 

[EarthDog]

LOL.. haha.. we did. And since we dont use it... we did know.