System crashes when GPU under-load

[bthewolf]

Computer crashing with power surge report
--------------------------------------------

Summary :

+ Dependant on GPU
+ Crashes within 5 minutes of running, often quicker based on load
+ Seems consistent regardless of operating system running
+ GPU fans don't seem to be running
+ Hardware seems fine
+ Problem only recently occurring

My 6-month-old build has started crashing over the past week, but only when I am running graphics intensive programmes. I have done some tests and I have realized that the fans aren't running, though they are capable of doing so as I noticed them start up and further tested the assertion using MSI gaming app, which boosted the fans to rather a high rpm count. I noticed when running the Uniengine Heaven benchmark, that it seems to quickly rise to about 60 degrees celsius and then crash. The GPU runs with excellent performance until the crash suddenly occurs.

I have tried running Ubuntu off of a USB device without installing, and running the Heaven benchmark, which also had the same issue.

This issue only started the other day, the fans are able to run, yet the problem is consistent. What has gone wrong? Is there something wrong with the firmware? The hardware seems fine. 60 C should be within operating limits, so that seems unusual too.

-------------------------------------------
build specs:

PSU - Corsair Builder Series CXM 600W Modular 80 PLUS Bronze Certified ATX/EPS PSU

Motherboard - Asus Z97-A/USB 3.1 Motherboard (Socket 1150, Z97, DDR3, S-ATA 600, ATX, USB 3.0, PCI Express 3.0, M.2 Socket)

CPU - Intel Core i7-4790K CPU (Quad Core 4GHz, Socket H3 LGA-1150)

GPU - MSI NVIDIA GTX 970 Gaming Twin Frozr HDMI DVI-I DP Graphics Card (4GB, PCI-Express, DDR5, 256 Bit)

CPU Cooler - Cooler Master Hyper 212 EVO (120mm)

Memory - Corsair CMY16GX3M2A1866C9R Vengeance Pro Series 16GB (2x8GB) DDR3 1866Mhz CL9 XMP Performance Desktop Memory Kit Red

OS: Windows 10 pro
------------------------------------------

If it is the power supply, why is it only crashing when the GPU is running, regardless of the CPU or memory usage. If it's the GPU, why is it capable of running as good as new for 1-10 mins before it crashes. If it is overheating, why would it crash when I kept cooling it using MSI gaming app, boosting the fans regularly and keeping the temps around 40-50 C.

I have tried testing multiple games, programmes, benchmarks, running Linux from a flash drive and running benchmarks there, opened the computer checked everything seemed okay. I unplugged the GPU and ran a game and a benchmark without crashing, so it's definitely dependent on the GPU. I have, with the GPU installed and while streaming to my TV through it, ran the machine for 2 days before starting a game and seeing the system crash within minutes. I have updated the drivers and the operating system, etc.

When the system crashes, it says American Megatrends and warns me of a power surge.

Please advise, I desperately need your help!

 

[plasmastorm]

Sounds like the PSU is maxing out and spiking.
Nvidia recommend a 500w Minimum for the PSU, so the 600w you have should be fine but they do loose capacity as they age.

the CX range is fairly well known for using cheap parts as Corsair do not make it themselves: http://www.overclock.net/t/1415677/whats-exactly-the-problem-with-corsairs-cx-psus

 

[bthewolf]

Sounds like the PSU is maxing out and spiking.
Nvidia recommend a 500w Minimum for the PSU, so the 600w you have should be fine but they do loose capacity as they age.

the CX range is fairly well known for using cheap parts as Corsair do not make it themselves: http://www.overclock.net/t/1415677/whats-exactly-the-problem-with-corsairs-cx-psus

It's only 6 months old! It was literally fine until just the other day. My CPU is set to automatically overclock, but I got this PSU because it was supposed to be more than adequate for my system.

 

[13thmonkey]

you are right, the GPU is not at all stressed temp wise That GPU would be happy up to 95C so no trouble there. . This screams PSU fault at me, especially when I see that it is a CX. The GPU is the biggest load on a system by a long way. Given your penultimate sentence it looks like the PSU is going out of regulation and a voltage is going out of spec and something is shutting down safely either the mobo or the PSU itself.

That GPU will not spin up the fans until 60C, so that's of no concern, and it's natural that it would rapidly reach 60C under uniengine.

The only thing that makes me concerned it's not the PSU is that it crashes at 60C, that seems to therefore crash when the fans spin up, that tiny extra load should be irrelevant (2-3Watts compared to a fluctuating 175watts for the GPU)/

 

[bthewolf]

you are right, the GPU is not at all stressed temp wise That GPU would be happy up to 95C so no trouble there. . This screams PSU fault at me, especially when I see that it is a CX. The GPU is the biggest load on a system by a long way. Given your penultimate sentence it looks like the PSU is going out of regulation and a voltage is going out of spec and something is shutting down safely either the mobo or the PSU itself.

That GPU will not spin up the fans until 60C, so that's of no concern, and it's natural that it would rapidly reach 60C under uniengine.

The only thing that makes me concerned it's not the PSU is that it crashes at 60C, that seems to therefore crash when the fans spin up, that tiny extra load should be irrelevant (2-3Watts compared to a fluctuating 175watts for the GPU)/

I didn't realize the fans shouldn't start until then, but that's interesting to know. I suppose if something has to be replaced, better that it's the cheaper component. But how can I make sure it's the PSU, and how can I make sure it need to be replaced? What could be going wrong with it that could be fixed? I can't really afford to be replacing parts so soon after getting it, it wasn't the cheapest of builds as is.

 

[13thmonkey]

IF it is the PSU then replacement is the option, not repair. You could RMA the PSU that might cost you shipping and time, but you'll get another CX and you'll hit the same problem or worse.

The only way to prove it is to replace it, find a friend with one, find a store with really good returns policy and keep it for a day or two (as the fault is readily replicable), if it does solve it then keep it, if it doesn't then come back. But everything is point at a PSU issue and a coincidence on the temp front.

 

[bthewolf]

IF it is the PSU then replacement is the option, not repair. You could RMA the PSU that might cost you shipping and time, but you'll get another CX and you'll hit the same problem or worse.

The only way to prove it is to replace it, find a friend with one, find a store with really good returns policy and keep it for a day or two (as the fault is readily replicable), if it does solve it then keep it, if it doesn't then come back. But everything is point at a PSU issue and a coincidence on the temp front.

I tried to put all my settings to eco and start the game on low settings with nothing else open, yet it still crashed. In fact, it crashed much quicker... is it still probable that its the power supply?

 

[13thmonkey]

Yes, still looks like PSU.

 

[bthewolf]

Yes, still looks like PSU.

I've sent an email to corsair, hopefully, they reply soon. Thanks for your help so far!