PC reboots when playing games

[shu ra]

I have similar issue with my GTX 670, it's already 4+ year old GPU.

 

Anyway the problem can avoided if you play game on lower settings. Try lowering the setting until you can keep playing without crashing the GPU.

I think problem could be one of two, either PSU not giving out enough power to GPU under massive load, then making it crash. You can check this with PSU output of 800 or more.

or GPU chipset is screwed! as it can't handle load over certain point. Might related with processor or memory or anything other.

[Mindovermaster Global Moderator]

4 minutes ago, shu ra said:

I have similar issue with my GTX 670, it's already 4+ year old GPU.

 

Anyway the problem can avoided if you play game on lower settings. Try lowering the setting until you can keep playing without crashing the GPU.

I think problem could be one of two, either PSU not giving out enough power to GPU under massive load, then making it crash. You can check this with PSU output of 800 or more.

or GPU chipset is screwed! as it can't handle load over certain point. Might related with processor or memory or anything other.

I doubt his PSU is at fault here. He is well within PSU requirements.

[DaCypher]

@shu ra -- I did look up the specs for power requirements for the GPU and power output from the PSU and I think I should be good.  I believe the Rx470 wants around 40A of power and my PSU outputs 54A. (at 12V that multiplies out to 650W).

 

So I tried tinkering with the GPU clock speed using ASUS GPU TweakII (though I think this can be done within the Radeon software itself?).  I ended up underclocking the GPU clock to 1150 (from default 1250).  So far things have been working OK, though just the hour or so worth of gaming is far from conclusive...

[shu ra]

I don't know how PSU works. 

 

If it output total overall of 650W, then it might be at fault. As your CPU and other component do take require a massive power under load.

But 

If it's output 650W individually from each output, then it's GPU fault for sure. 

 

Anyway that's just perspective could be wrong,

Also If you have utility, try under-volting  the card and see if it works. Just be careful with it.

Edited February 22, 2018 by shu ra

[Mindovermaster Global Moderator]

14 minutes ago, shu ra said:

I don't know how PSU works. 

 

If it output total overall of 650W, then it might be at fault. As your CPU and other component do take require a massive power under load.

But 

If it's output 650W individually from each output, then it's GPU fault for sure. 

Totally not true. If he was running a 400-500W, yes, I would be worried. But, he is not driving a very beefy system, by far.

 

All components use it, yes. But everything, other than your CPU and GPU, doesn't use up that much.

 

If you don't know how it works, how can you give support on it?

[shu ra]

48 minutes ago, Mindovermaster said:

Totally not true. If he was running a 400-500W, yes, I would be worried. But, he is not driving a very beefy system, by far.

 

All components use it, yes. But everything, other than your CPU and GPU, doesn't use up that much.

 

If you don't know how it works, how can you give support on it?

Calm down!

 

I'm just presenting my view, As I mention in the post. You can take that how you want! Also considering your view I'm not giving support on PSU.

 

Secondly, I already presented the workaround (temp solution), in my first post. Which I know for fact works from my personal experience.

[domboy]

13 hours ago, DaCypher said:

 

@domboy  -- interesting suggestion; I suppose I could test by temporarily disconnecting the power button from the motherboard and see if the problem occurs?  Maybe the additional fans running during gaming might cause vibration resulting in the faulting power/reset buttons triggering?

 

Right, that's what I did - disconnect the power button cable and used a small screw driver to make a connection between the two pins to turn it the PC on. It's sounding like this probably isn't your issue, but it's something I won't forget as it took me awhile to figure out. I'm assuming the internals of the button had worn down and maybe like you said, a vibration would somehow cause it to short out and tell the PC to shutdown or reset. It was very strange and very random.

[DevTech]

On 2/22/2018 at 12:14 AM, DaCypher said:

@shu ra -- I did look up the specs for power requirements for the GPU and power output from the PSU and I think I should be good.  I believe the Rx470 wants around 40A of power and my PSU outputs 54A. (at 12V that multiplies out to 650W).

 

So I tried tinkering with the GPU clock speed using ASUS GPU TweakII (though I think this can be done within the Radeon software itself?).  I ended up underclocking the GPU clock to 1150 (from default 1250).  So far things have been working OK, though just the hour or so worth of gaming is far from conclusive...

- make sure Load Line Calibration (or similar setting) is on in the BIOS (it is an overclocking setting that maintains steady power to the CPU, so good for stressed non-overclock)

 

- GPU RAM speed is a;ways close to limit for some reason and usually there is a separate setting to underclock the GPU RAM

 

The curious thing here is what exactly might have deteriorated over time. I always like to find out any long term answers in case it helps anyone else...

 

[DaCypher]

@DevTech-- thanks, I'll check both of those.  So far things seem to be running OK with GPU clock at 1150 instead of 1250.  I may try to slowly bring it back up towards 1250 and if that starts showing issues again I'll try dropping the memory clock speed (default is 6600 so looks like there is some room to play here).

As for what has caused this, I'm not convinced that this is new.  I've had this video card for about 18 months now and I remember seeing similar issues previously that I just shrugged off as buggy games.  But recently I've gotten back into gaming a bit and the issue has become more frequent as a result.  I suspect the GPU itself was probably faulty when it left the factory but only just enough to be annoyingly problematic and not completely broken.  Googling has shown I'm not the only one with this issue.  Needless to say, I don't think I'll be getting an AMD or ASUS video card again in the future...

[sc302 Veteran]

16 hours ago, DaCypher said:

  Needless to say, I don't think I'll be getting an AMD or ASUS video card again in the future...

I don't think that it has anything to do with the brand or the manufacturer of the chipset as a whole.  I think it is very specific to that specific model/video card/chipset.  ASUS does not have a lot of bad press regarding their hardware as a whole, and AMD is one of the 2 main providers of video card chipsets.  I think it is a very isolated incident to a specific product/line, if you look at asus's flagship nvidia cards they have stellar reviews and almost no complaints.  

[DaCypher]

@sc302 -- Fair point, I probably shouldn't be so quick to point the finger.  That said, this thread on AMD's forums suggests there are some issues with the rx470 as many people seem to be having the same problem across many different versions of the Radeon drivers.  It doesn't seem to be specific to a particular manufacturer of graphics cards, but does seem to be specific to the rx470.

[+Mirumir Subscriber¹]

On 2/27/2018 at 4:11 AM, DaCypher said:

 So far things seem to be running OK with GPU clock at 1150 instead of 1250. 

If down-clocking has alleviated the issue, it does sound like a problem with the heatsink.

 

Most OEM's heatsinks and fans are garbage that should be replaced immediately after the purchase (if you are a gamer).

[sc302 Veteran]

But temps are well within normal operation.  Which doesn’t point to failing heat sink, fan, or thermal paste. 

[+Mirumir Subscriber¹]

7 minutes ago, sc302 said:

But temps are well within normal operation.  Which doesn’t point to failing heat sink, fan, or thermal paste. 

It's possible that, on a GPU board, different sections could have different temperatures.

 

Some memory module can be overheating while the core GPU's temp would be normal.

[Mindovermaster Global Moderator]

25 minutes ago, Mirumir said:

It's possible that, on a GPU board, different sections could have different temperatures.

 

Some memory module can be overheating while the core GPU's temp would be normal.

Darn.. Ninja'd. Was just going to say that...

[sc302 Veteran]

44 minutes ago, Mirumir said:

It's possible that, on a GPU board, different sections could have different temperatures.

 

Some memory module can be overheating while the core GPU's temp would be normal.

I can agree that it is possible esp being that the heat sink doesn't touch all areas of the board and if it is like my 1080, the fans don't move until the heat levels are reached at the core.  My fans remain at 0 rpm at full settings on overwatch and sc2 @1080p.

[LaP]

On 2018-02-20 at 9:53 PM, DaCypher said:

OK I been gaming for several hours tonight before the issue occurred again.  This time I was running OpenHardwareMonitor with logging enabled.  The CPU temp stays steady at around 65-ish.  While this is hotter than I'd prefer I think this is well below the cutoff limit (someone correct me here).  The GPU core temp seems to be hovering right at 69.  While it does get up to 70 a couple times through out, I notice that 70 shows up a bit more frequently near the end of the log file.  I wonder if the GPU could have a cutoff at 70/71?

Very highly unlikely. At worst the cutoff is 80 but probably 90 or 100.

 

70 is a little high but it's not a problem. Those temp while playing games are fine. Higher than i would run my computer at but it wont affect the stability just the durability.

[LaP]

6 hours ago, Mirumir said:

If down-clocking has alleviated the issue, it does sound like a problem with the heatsink.

 

Most OEM's heatsinks and fans are garbage that should be replaced immediately after the purchase (if you are a gamer).

If it's a stock cooler it could effectively be the problem. Stock coolers are often awful and yeah while the temp of the core might be fine the rest of the card might not be properly cooled. Many years ago i flashed my 9800 pro to a 9800 xt and replaced the cooler by a 3rd party cooler. The cooler was not properly cooling the ram and while the temp were fine the computer kept crashing while playing games. I had to buy some ram heat sink to cool the ram of the gpu.

 

 

[JustGeorge]

Bad/overheating VRM on GPU perhaps? Under-volting alleviating the stress on the faulty VRM???

[DaCypher]

I'm still tinkering with the clock settings... For the past couple of days been I've running at default values with no issues at all!  So clearly more testing (gaming, hehe) is needed to see if I can find any rhyme or reason to this behavior.

[jnelsoninjax]

9 minutes ago, DaCypher said:

I'm still tinkering with the clock settings... For the past couple of days been I've running at default values with no issues at all!  So clearly more testing (gaming, hehe) is needed to see if I can find any rhyme or reason to this behavior.

Did you say that this happens when you are playing Assassins Creed Origins? If you press F1 in game, it will bring up the performance meter and you can see the GPU and CPU as well as RAM loads, that might give you some idea if it is hardware related. I know that that game is very graphically rich and at the highest settings I can see some minor lagging on my system (AMD FX 8370, 16 GB's RAM and a GTX 1080)

[DevTech]

1 hour ago, DaCypher said:

I'm still tinkering with the clock settings... For the past couple of days been I've running at default values with no issues at all!  So clearly more testing (gaming, hehe) is needed to see if I can find any rhyme or reason to this behavior.

Nothing you have provided so far is any indication of overheating.

 

 

By underclocking, you changed RAM timing and also power requirements.

 

If you have a new GPU card that takes 100% of its power from the PCIe bus in an older motherboard, than can also be an issue.

 

Also, some cards had a glitch where they grabbed too much power from the PCIe bus and not enough from the external power connectors which can be fixed via BIOS update for GPU card.

 

[DaCypher]

4 hours ago, jnelsoninjax said:

Did you say that this happens when you are playing Assassins Creed Origins? If you press F1 in game, it will bring up the performance meter and you can see the GPU and CPU as well as RAM loads, that might give you some idea if it is hardware related. I know that that game is very graphically rich and at the highest settings I can see some minor lagging on my system (AMD FX 8370, 16 GB's RAM and a GTX 1080)

Oh cool, didn't know about the F1 thing, I'll have to check this out.  I was just playing for several hours now without problem and I also tried upping the resolution to 2k.  My system clearly can't handle that (very laggy/jittery) but I wanted to push it for a bit to see if anything bad would happen.  No issues at all (other than terrible framerate as expected).  I suspect the info displayed by the F1 thing would probably show my system as pretty much maxed out though.

 

3 hours ago, DevTech said:

Nothing you have provided so far is any indication of overheating.

 

 

By underclocking, you changed RAM timing and also power requirements.

 

If you have a new GPU card that takes 100% of its power from the PCIe bus in an older motherboard, than can also be an issue.

 

Also, some cards had a glitch where they grabbed too much power from the PCIe bus and not enough from the external power connectors which can be fixed via BIOS update for GPU card.

 

Definitely good info, thanks.  A couple days ago I looked into BIOS update for my GPU.  It seemed to install OK but after a reboot, Windows booted up and then after about 5-10 seconds of just moving my mouse around on desktop the system would crash (hard lockup; reset/power buttons on my case did nothing, had to toggle the power supply switch).  This happened consistently so clearly the BIOS update caused problems.  Fortunately I took a backup of my GPU BIOS prior to the update and I was able to revert back.  What an adventure that was...  But I wish it had worked correctly as I suspect there is something up with the GPU itself that is causing all of this.

 

Also, I have played with the ASUS GPU Tweak tool.  If I overclock the GPU in the smallest increment possible (to 1260 Mhz from default 1250), then I almost immediately get the crash just playing a youtube video in full screen.  So it seems the GPU's default settings might be right at the very limit of what it's capable of and even then sometimes under stress it might go beyond which results in the crashing with default settings?

[DevTech]

2 minutes ago, DaCypher said:

Oh cool, didn't know about the F1 thing, I'll have to check this out.  I was just playing for several hours now without problem and I also tried upping the resolution to 2k.  My system clearly can't handle that (very laggy/jittery) but I wanted to push it for a bit to see if anything bad would happen.  No issues at all (other than terrible framerate as expected).  I suspect the info displayed by the F1 thing would probably show my system as pretty much maxed out though.

 

Definitely good info, thanks.  A couple days ago I looked into BIOS update for my GPU.  It seemed to install OK but after a reboot, Windows booted up and then after about 5-10 seconds of just moving my mouse around on desktop the system would crash (hard lockup; reset/power buttons on my case did nothing, had to toggle the power supply switch).  This happened consistently so clearly the BIOS update caused problems.  Fortunately I took a backup of my GPU BIOS prior to the update and I was able to revert back.  What an adventure that was...  But I wish it had worked correctly as I suspect there is something up with the GPU itself that is causing all of this.

 

Also, I have played with the ASUS GPU Tweak tool.  If I overclock the GPU in the smallest increment possible (to 1260 Mhz from default 1250), then I almost immediately get the crash just playing a youtube video in full screen.  So it seems the GPU's default settings might be right at the very limit of what it's capable of and even then sometimes under stress it might go beyond which results in the crashing with default settings?

1. Does your GPU card have an external power connector? If it does then starvation from the PCIe bus (which provides about 1/2 the power in that case) is not very common.

 

2. A huge amount of GPU cards are very near the RAM spec limit for reasons I don't know. If anyone does, please inform. So your experiment could be exposing either a GPU or RAM or less likely power limit. If it is RAM, then underclock that a bit and forget you ever had the problem.

 

If it is GPU, also underclock a good 5% and hope you can forget about it. You will never notice the performance difference and hopefully can get long term stability. But my concern in the case of GPU being close to the limit is that unlike RAM, most GPUs tend to overclock a bit in general without any issue, so you either have a sub-standard statistical anomaly off the production line (AKA Lemon) or something is getting worse over time.

 

[DaCypher]

8 hours ago, DevTech said:

1. Does your GPU card have an external power connector? If it does then starvation from the PCIe bus (which provides about 1/2 the power in that case) is not very common.

 

2. A huge amount of GPU cards are very near the RAM spec limit for reasons I don't know. If anyone does, please inform. So your experiment could be exposing either a GPU or RAM or less likely power limit. If it is RAM, then underclock that a bit and forget you ever had the problem.

 

If it is GPU, also underclock a good 5% and hope you can forget about it. You will never notice the performance difference and hopefully can get long term stability. But my concern in the case of GPU being close to the limit is that unlike RAM, most GPUs tend to overclock a bit in general without any issue, so you either have a sub-standard statistical anomaly off the production line (AKA Lemon) or something is getting worse over time.

 

1. Yes, it has a direct cable connected from the PSU.

 

2. I think you're right.  I'm experimenting a bit with the clock and memory settings.  For the past couple of days I've been running everything right at their defaults and haven't noticed an issue.  But the next time I see a problem I think you're small underclock suggestion is ultimately the way to go.  I don't think I'll notice a performance hit as I can run most games at 1080p with high graphical settings without problem.  1440p isn't doable even if I could overclock the GPU significantly so there's no point in pushing it.

 

And I agree, I'm thinking it's a lemon.  But given how many people have posted in that AMD forums thread I linked earlier, I don't think this was quite the statistical anomaly that it should be.  That is, it seems like there are a lot of rx470 GPUs out there with this same problem...