Desktop·25,124 discussions

Custom PC build randomly rebooting

Locked

Hello!

I’m experiencing an issue where my custom desktop PC is rebooting randomly when doing intensive tasks: sometimes it can go on and on for hours in a row of video encoding without rebooting, other times it takes 10 minutes of GTA 5 to reboot.

The symptoms go like this: while I’m doing my thing playing or encoding video, I lose the feed to the monitor, while other PC functionalities (like audio and peripherals) are fully functional. In 2-5 seconds after this, the whole PC reboots, with fans at max speed until Windows loads.

I used to have an overclock on this machine: the CPU was running at 3.5 GHz, 1.3V and memory at 1400 MHz. I removed the overclock once the problem started appearing, but it doesn’t seem to have solved much. At first, it seemed to stabilize the PC a little, but it might just be placebo.

While overclocking, temps were at 80-87 degrees C on the CPU cores when running AIDA64 FPU+CPU.

The motherboard doesn’t seem to want to boot if I’m using the normal 1.5V RAM voltage, but rather forces it to 1.56V.

I am also seldomly using one of the frontal USB ports to power a 5V 0.85A HDD, while the motherboard only supports USB 2.0 with 0.5A. The HDD is working properly and reboots don’t seem to appear while the HDD is plugged in and no CPU or GPU intensive tasks are running.

Specs:
Mainboard: Gigabyte GA-EX58-UD3R rev. 1.6
CPU: Intel Xeon X5650
RAM: 2 DIMMs of Kingston 99U5584-007.A00LF, 2 DIMMs of Micron 8JTF51264AZ-1G6E1
GPU: Gigabyte GTX 980 G1
Storage: 1 WD1003FZEX, 1 WD5000HHTZ, 1 KINGSTON SA400S3 240 GB
Power Supply: Super Flower Leadex 2 Gold 650 W

I’d like to know if any of these symptoms point to some specific part. I’m currently thinking it’s the PSU, but I’d like to know other opinions before taking it out of the system and shipping it to warranty.

Any ideas? Thanks!

Topic

Small world.

In reply to Custom PC build randomly rebooting

My son had a GTX 970 that caused this. He wasn’t going to believe me so I let him borrow a new (at the time) GTX 1060. Instant cure.

But as you know you can’t always have your kid agree with you that was the fix so he bought a new PC with the RTX 2080 and more.

Later I had time to dive into the cranky 970. It seemed a shame to not at least try to fix it. I put the 970 into another good PC and sure enough problems would sporadically happen. The one thing that would always cause it to bluescreen, reboot or hang (this was not consistent) was to run FURMARK (see google.)

OK, now for the cheapest fix possible. I got out the canned air and gave it a good clean. Now I pulled the heatsink, removed the then 5 year old compound off and put on new heatsink compound.

Lucky me, that fixed and the 970 lives on for a few more years.

Worth a try is little said

In reply to Small world.

I just used FurMark and GPU-Z to look at the GPU temperatures. Even when idle, the Hot spot sensor is at 63 degrees C and the GPU temp is 49. I think, after 3 years, it’s finally time to change that thermal paste inside the card.

I really hope it’s this easy. I’ll try cleaning it up and I’ll update the topic.

I have hope.

In reply to Worth a try is little said

At least you are going to try this. About half the time the OP or client comes back with “but the temps are not that high.”

We’re talking less than one dollar of DIY money to save what would be a many hundred card today.

If you are new to this work, be sure to watch a few YouTubes as there are some showing the work for the GTX 980.

-> To be complete, yes it could be something else like an aging power supply. We know to swap those past 2 years old since power supplies don’t age well. Why this order of work? Long ago another tech said “try the cheapest part first.”

Some news

In reply to I have hope.

When I powered on the PC today, after loading Windows, it wouldn’t display anything. I rebooted it using the reset button as it seemed like it wouldn’t react to anything. After a reboot, it ran for 2 minutes and it shut off again.

Decided to pull the graphics card out and replace it with an older one. Surely enough, it’s running fine now for two hours.

I believe the issue on the GTX 980 is the old thermal pads on the memory modules. They are dry and dusty. I bought some from China, as where I live they are overpriced, and I hope they’ll solve the issue.
The thermal paste shouldn’t be the issue, but I’ll replace it anyway and update the discussion again.

I think it’s stable for now

In reply to Some news

So, I opened up the graphics card. The pads were fine, actually in a better condition than I remembered them. I cancelled the order from China, but I will replace them once I have the time. The paste on the GPU was quite scarce, so that might have been the issue. I reapplied it, checked it twice to see it spread well. Reinserted the card in the computer.

Now, once Windows reached the login screen, the PC stopped responding. So I decided to reboot Windows into safe mode and uninstall the driver. In safe mode, no issues occured with the PC. As soon as the driver was out (March 2021 update), the issues were gone. Installed the March 2020 one and I saw that GPU-Z reported PCI-E x8 running speed for the GPU. I went ahead and started FurMark once again anyway. The PCI-E speed stayed at x8. The Hot spot temp reached a maximum of 94°C and the GPU – 80°C. Idle temps were 62 for the Hot spot and 49 for the GPU.

Since I wanted to see the GPU run at x16, I started troubleshooting it. I cleaned the PCI-E slot and the PCI-E plug on the GPU and, on that occasion, I also cleaned the PSU fan dust filter, which was full of dust (had no cleaning done to it in 2 years).

Rebooted again, started GPU-Z. PCI-E x16. Idle temps down to 42°C and 51°C, respectively. Ran FurMark, max temps remained unchanged. Left it run for 10 minutes and no crash.

Seems stable for now. If anything occurs, I’ll update the discussion.

Clean and tidy …

In reply to I think it’s stable for now

Is always a good thing. Good to read that it appears to be one of the cheap fixes.

No more reboots

In reply to Clean and tidy …

The PC is stable now. Haven’t had any reboots in 4 days, but I have been gaming and rendering video just as before.

It wouldn’t be a bad idea to order some thermal pads anyway, since 94°C is not at all little and more airflow in the case I’m using (Deepcool Dukase) can’t be done. I have one fan in the front and one on the top. Maybe moving the top fan to the front might help? I’m not sure, since the front is mostly blocked by the door for the optical drive.

Thoughts about….

In reply to No more reboots

1. Given the age, the work done it may be best to do nothing.
2. If I was to do anything I’d see if the existing fans “work like new” and change any that fail this test.
3. If there is a spare vent where a fan could be mounted I might install one but this machine appears to have been fine for many years so my advice given only the story here and not seeing the PC firsthand is to give another nod to items 1 and 2.

-> I think you fixed it. Might want to enjoy it rather than possibly unfix it?

The end of story is worth sharing

In reply to Thoughts about….

The Chinese thermal pads arrived and, after installing them and applying thermal paste on the GPU dye’s edges, the idle GPU temps dropped to 35-36°C, way lower than the previous 49.

I repositioned the two fans I have: one is in the back of the case, pulling air out, the other in the front, also pulling air out.
Why both out?
Because, when the system needs more air, it gets it from the top, as it is open. At some point, I’ll also buy a fan to position it on the top, pushing air inside the case.
But, for now, after setting the fans up like this, I notice less GPU fan RPM for the same temperature.

I also documented myself a bit more regarding the Hotspot temperature and it seems like the Hotspot temp reported for Graphics Cards older than Pascal is simply a fixed 14°C above GPU temp.

[Author]

Replies