GPU Caught on Fire - How do I know no damage was caused to the rest of the system?

Daboa

Distinguished
Feb 17, 2013
66
0
18,540
I was browsing the web this morning when my desktop turned off. It tried rebooting a couple of times while making unusual sounds. By the third boot attempt, I had smelled the burning and unplugged the PSU. Not being sure which component was failing, I tried turning on the PC again and immediately saw flames erupt from my GPU. I unplugged again and pulled out the GPU. My PC seems to be working fine with integrated graphics for now and I have some photos of the damage below. I do not see any visible damage to the MOBO or PCI slot and it looks like the part that failed wasn't right against the MOBO. Still, my questions are:

How can I be confident that the rest of my system is "okay"?
How can I know that installing a new GPU won't instantly fry the new GPU?

oA5Mlny.jpg


An album with a few photos:
https://imgur.com/a/5qEAB1W


Btw, the card is 4 years old, outside of warranty.
 
That GPU is toast. Literally. What a shame. :/

As for the rest of your rig, the only way to know for sure is to drop in a new card and turn it on. There's just no other way of finding out. I'm interested to know the make and model of the card (if you don't mind disclosing)?
 

Daboa

Distinguished
Feb 17, 2013
66
0
18,540


Yup, the GPU is definitely done for. As for my specs:

GPU - Gigabyte Radeon R9 280X 3 GB (rest in peace)
MOBO - MSI Z87-G45
PSU - EVGA SuperNOVA 650W
CPU - i5-4670k
RAM - G. Skill ripjaw X series 16GB
Case - Corsair 200R
 


I had a friend who had a several years old GPU go "Pop" and left burn marks and a little smoke smell. Everything else works fine last I spoke with him.
 

Daboa

Distinguished
Feb 17, 2013
66
0
18,540
I think I will order and try a new GPU. Any tips for monitoring temps and such to make sure the new one is operating safely? I have HWMonitor, but I'm not sure what temps/voltages are safe.
 


Different GPUs tolerate different voltages. First, I'd recommend using HWInfo instead of HWMonitor. HWInfo is a much more detailed and reliabel monitoring program.
 

Daboa

Distinguished
Feb 17, 2013
66
0
18,540


What kind of things would I be looking for with HWInfo? What values are safe or not? Sorry, but I'm not sure how to properly use the tool. Thanks.
 


I'm not sure you can actually forsee what happened to your card using software. I use HWInfo to monitor temps and utilization of CPU, GPU, and memory and voltage.

HWMonitor has been known to give false readings.
 
Based on the other chips next to it, it looks like a MOSFET power switch. It's likely used to control the load/power on other components on the GPU. The good news is that it looks this is the device that failed versus a failure of of another device (that may not be visible) that cause the MOSFET switch to fail. It's very likely that the PCIe connector and bus on the mobo is just fine. Inspect the PCIe connector for any discoloration around the pins.

One thing you could do is get a cheap, used GPU to test your PC before buying a new, expensive GPU.
 

Daboa

Distinguished
Feb 17, 2013
66
0
18,540


Someone on another thread said the same thing and I don't see any discoloration around the PCIe connector. I'm considering buying a cheap GPU on Ebay to see if it also explodes (example for $10). But it would likely have different power requirements and could have its own issues, so I feel like the test might not be meaningful. Otherwise, my friend can give me a decent card for $150 and I might just plug it in and cross my fingers.

I might also try calling Gigabyte. Even though I'm outside of warranty, this could have burned down my house if I weren't home.