Question PC freaks out and turns off when GPU is used

[agent031693]

System specs:
CPU: i7-4790k
MoBo: Asus Q87M-E
GPU: Vega 64
Ram: (mixed) 16Gb
PSU: EVGA G5 80+ Gold


More info:

I've tried several things to solve this issue. Everything I try seems to work... but only for a couple of days before the problem creeps up again. Last three things I've tried:
1st, bought a new PSU (before I had a Raidmax vampire)
2nd, changed PCI-E slots
3rd, took GPU out, uninstalled all drivers, reset bios to default with the exception of the automatic GPU being set to iGPU, then restarting the PC with the GPU in, reinstalled drivers, and restarted the pc once again...

All three of these things seemed to have temporarily fixed the issue. For the first day or two, after trying these fixes, I would have no problem gaming or video editing, or running benchmarks for even hours on end. After that, the issues slowly evolve - meaning, first it seems random (after about 30 minutes of GPU use), then eventually I cannot get even a 10-20 minutes of GPU use until eventually, I cannot even get past the main menu in games.

Specifically what happens. Everything is running fine. The fans spin at a moderate 50% or so, and the temps look 100% normal. Then, without warning, all fans ramp up to 100%, about 2 seconds before the monitor says, "No signal, entering powersaving mode". Then the screen goes black, and the fans stay on full-tilt until I literally unplug or flip the switch on the PSU. The power button acts unresponsive, reset button acts unresponsive. After I get the PC turned off, it won't turn back on for an hour or two. If I let it run, it will run at full tilt until my UPS System gives. I've ran it with and without UPS Systems. Different UPS Systems. Plugged into in-wall surge protectors... etc... Always the same thing. When this happens, the internal clock always goes offline. So I've let my computer sit for a couple of days, unplugged it, then plugged it back in. It has the exact time and date that this incidence last happened, still on its (current time) for the internal clock. When I get this issue "temporarily fixed" everything is normal. I can power off my PC when I am done using it, without a worry in the world. When this issue comes back, even if I were never using the GPU, if I simply turned off the PC (and left it plugged into the wall), it would turn itself back on, ramp up the fans, and surge itself off, without warning or reason.

The issue seems to be non-existent when I use my PC with no GPU installed. You know, except for I can't game or anything else graphic intensive on the integrated graphics of this nearly 7-year-old CPU

The clips on the PCI-E slots are broken. So I cannot "lock" the GPU into the slots, however, I figure if I get the GPU sitting right, and never move the PC, this should not be an issue

The next thing I want to try is buying a new MoBo... Possibly its the GPU, however, I feel as if this happened (just not to the same extreme extent) with the GPU I owned before (an RX 580, 8Gb), and even when I ran two RX 580's in x-fire, things like this happened occasionally, but not too often. Those RX 580's, especially in x-fire ran hot, and I assumed that's what caused the random shut-downs. This new GPU has an aftermarket cooler on it, and seldomly goes over 70c. My memory is fuzzy if the powering off issue was the exact same with my old GPU, that was nearly a year ago. Assuming it was the same issue, it was never nearly this bad

Any suggestions, outside of buying a new MoBo or GPU is more than welcome. If you are convinced the MoBo and/or GPU needs to be replaced, kindly explain why you think so. I am willing to try nearly anything. This i7 still performs very well in games, even 6+ years after its release. So I would like to hold off as long as possible on upgrading my MoBo/CPU/Ram

Thank you!

 

[Lutfij]

before I had a Raidmax vampire
That might've caused the rest of your system innards to take a hit(electrical damage).

What is the wattage of the EVGA PSU? Is the PSU as old as the processor? If so then the culprit is there, see if changing the PSU out for a donor unit with at least 750W of power for the entire system changes the issue.

 

[agent031693]

before I had a Raidmax vampire
That might've caused the rest of your system innards to take a hit(electrical damage).

What is the wattage of the EVGA PSU? Is the PSU as old as the processor? If so then the culprit is there, see if changing the PSU out for a donor unit with at least 750W of power for the entire system changes the issue.

Sorry, Thought I had it in the original post...

Both PSU's listed were 850w.

The Raidmax was bought new, less than 3 years ago. I figured Raidmax isn't a very reputable brand, and I got it for a very cheap price, so my first hunch was to go with a brand new, very highly regarded EVGA PSU. The EVGA PSU is about 6 months removed from its original packaging/seal. I believe it to be a 2019 or 2018 model, however.

 

[spentshells]

Well, here is a suggestion drop the mixed ram and try again. Go with the single set of matching dimms. Same issue, try the other set. Same issue? The issue is not the ram.

 

[agent031693]

Well, here is a suggestion drop the mixed ram and try again. Go with the single set of matching dimms. Same issue, try the other set. Same issue? The issue is not the ram.

Trying that next... Not holding breath. I had been using these 4 sticks of ram for years. I forgot what ram it was. Having pulled out all four sticks and observing their specs, they are basically identical. All four sticks are: 4Gb, G-Skill, 1866mHz, CL9-10-9-28, PC3-14900, 1.50v... They all have heat spreaders. The only difference is that two are branded as G-Skill Ripsaw, and the other two are branded as G-Skill Ares...

However, it's 100% worth a try. So I pulled out the two Ripsaw sticks, and if I have issues, I'll try the Ripsaw. I will also try all four slots, to see if one of the ram slots is defective

Thank you for the advice. Will let you know what comes of it

 

[spentshells]

Trying that next... Not holding breath. I had been using these 4 sticks of ram for years. I forgot what ram it was. Having pulled out all four sticks and observing their specs, they are basically identical. All four sticks are: 4Gb, G-Skill, 1866mHz, CL9-10-9-28, PC3-14900, 1.50v... They all have heat spreaders. The only difference is that two are branded as G-Skill Ripsaw, and the other two are branded as G-Skill Ares...

However, it's 100% worth a try. So I pulled out the two Ripsaw sticks, and if I have issues, I'll try the Ripsaw. I will also try all four slots, to see if one of the ram slots is defective

Thank you for the advice. Will let you know what comes of it

hey it's worth a shot

 

[agent031693]

Trying that next... Not holding breath. I had been using these 4 sticks of ram for years. I forgot what ram it was. Having pulled out all four sticks and observing their specs, they are basically identical. All four sticks are: 4Gb, G-Skill, 1866mHz, CL9-10-9-28, PC3-14900, 1.50v... They all have heat spreaders. The only difference is that two are branded as G-Skill Ripsaw, and the other two are branded as G-Skill Ares...

However, it's 100% worth a try. So I pulled out the two Ripsaw sticks, and if I have issues, I'll try the Ripsaw. I will also try all four slots, to see if one of the ram slots is defective

Thank you for the advice. Will let you know what comes of it

Gave it a go. Same issue. Used only 2 of the 4 sticks of ram (made sure they were matching)... Ran Unigine Superposition at 4k for quick stress... Didn't last even half-way through the stress test.

Was worth a shot. Sincerely wish it were that easy

 

[agent031693]

System specs:
CPU: i7-4790k
MoBo: Gigabyte GA-Z97M-DS3H
GPU: Vega 64
Ram: (mixed) 16Gb
PSU: EVGA G5 80+ Gold


More info:

I've tried several things to solve this issue. Everything I try seems to work... but only for a couple of days before the problem creeps up again. Last three things I've tried:
1st, bought a new PSU (before I had a Raidmax vampire)
2nd, changed PCI-E slots
3rd, took GPU out, uninstalled all drivers, reset bios to default with the exception of the automatic GPU being set to iGPU, then restarting the PC with the GPU in, reinstalled drivers, and restarted the pc once again...

All three of these things seemed to have temporarily fixed the issue. For the first day or two, after trying these fixes, I would have no problem gaming or video editing, or running benchmarks for even hours on end. After that, the issues slowly evolve - meaning, first it seems random (after about 30 minutes of GPU use), then eventually I cannot get even a 10-20 minutes of GPU use until eventually, I cannot even get past the main menu in games.

Specifically what happens. Everything is running fine. The fans spin at a moderate 50% or so, and the temps look 100% normal. Then, without warning, all fans ramp up to 100%, about 2 seconds before the monitor says, "No signal, entering powersaving mode". Then the screen goes black, and the fans stay on full-tilt until I literally unplug or flip the switch on the PSU. The power button acts unresponsive, reset button acts unresponsive. After I get the PC turned off, it won't turn back on for an hour or two. If I let it run, it will run at full tilt until my UPS System gives. I've ran it with and without UPS Systems. Different UPS Systems. Plugged into in-wall surge protectors... etc... Always the same thing. When this happens, the internal clock always goes offline. So I've let my computer sit for a couple of days, unplugged it, then plugged it back in. It has the exact time and date that this incidence last happened, still on its (current time) for the internal clock. When I get this issue "temporarily fixed" everything is normal. I can power off my PC when I am done using it, without a worry in the world. When this issue comes back, even if I were never using the GPU, if I simply turned off the PC (and left it plugged into the wall), it would turn itself back on, ramp up the fans, and surge itself off, without warning or reason.

The issue seems to be non-existent when I use my PC with no GPU installed. You know, except for I can't game or anything else graphic intensive on the integrated graphics of this nearly 7-year-old CPU

The clips on the PCI-E slots are broken. So I cannot "lock" the GPU into the slots, however, I figure if I get the GPU sitting right, and never move the PC, this should not be an issue

The next thing I want to try is buying a new MoBo... Possibly its the GPU, however, I feel as if this happened (just not to the same extreme extent) with the GPU I owned before (an RX 580, 8Gb), and even when I ran two RX 580's in x-fire, things like this happened occasionally, but not too often. Those RX 580's, especially in x-fire ran hot, and I assumed that's what caused the random shut-downs. This new GPU has an aftermarket cooler on it, and seldomly goes over 70c. My memory is fuzzy if the powering off issue was the exact same with my old GPU, that was nearly a year ago. Assuming it was the same issue, it was never nearly this bad

Any suggestions, outside of buying a new MoBo or GPU is more than welcome. If you are convinced the MoBo and/or GPU needs to be replaced, kindly explain why you think so. I am willing to try nearly anything. This i7 still performs very well in games, even 6+ years after its release. So I would like to hold off as long as possible on upgrading my MoBo/CPU/Ram

Thank you!

Does anybody know, is there a way to disable the motherboard temperature sensors all together?

Next tried undervolting my GPU... That caused the computer to freeze and restart even without doing anything that would use the GPU... Got the undervolt off, and it's back to "normal"

I noticed something. I reopened Open Hardware Monitor, for the first time in a while. The core temp GPU at idle are both 36°C... However, the GPU states that the rest of the temperatures (Memory VRM... etc) are all 0°C. Then the CPU seems completely normal. Without doing anything the package and all four core temps range from 32-37°C. The other abnormality would be the Motherboard reading 7 different temperatures. The first 4 seem completely normal. The last 3 all read in the mid-high 90's. So my new theory is that the motherboard has broken temperature sensors... So... is there a way to disable the motherboard temperature sensors all together? Furthermore, (probably less important) is there something I can or should do about the GPU temperature readings at 0°C?

 

[agent031693]

Sort of found a solution... I think the drivers helped, however not completely. Since all of this (I believe since I reset the drivers), I can now use the GPU without the PC crashing. However, it's very specific. If I run at 3440×1440 resolution or higher, it's going to crash. At 2560×1080, there are almost never any issues. If I run in Vulkan or OpenGL, then it crashes. If I only run at 1080p, in DirectX, then everything is smooth as silk. Another guess is something to do with the wattage... It's a bit weird. Running Unigine stress test at 1080p will max my GPU to 100% with a high temp of 56c... If I run at (let's say) 4k, the GPU usage (obviously) is still at 100%, and temps still at about the same. However, I noticed, at 1080p, the GPU doesn't crack 200 watts of usage. At 4k, the GPU hits nearly 220 watts.