A new GPU has been activated and is running smoothly.
A new GPU has been activated and is running smoothly.
I've been using integrated graphics for some time, but recently swapped in an ASRock 7900 XTX Taichi and installed it with the Liani li vertical mount. It ran smoothly for about 18 hours without any problems. During a 30-minute OCCT stress test, it completed successfully within the first few hours. The next day I noticed a spark, a faint smoke, and the system shut down. I quickly switched off the power supply and unplugged the cable to cut the power. After removing the card, everything reconnected properly—fans, RGB, and all components booted up fine. Windows loaded perfectly.
Key points I considered:
- I’m very confident I didn’t use an EPS instead of a PCIe cable, as that would have required forcing it in.
- The issue occurred when the card was using 1–3% load at temperatures of 18–20°C—new Windows install, no other changes.
- The PSU seems to be under its minimum capacity, yet it struggled significantly.
- It happened after a recent OCCT, which didn’t trigger any issues.
Questions for next steps:
- What might have led to this without damaging the rest of the system?
- How can I verify the PCIe slot still supports a GPU safely?
- Is there a way to test the slot without a GPU?
- Do you have any advice to prevent this in the future?
Thanks for the help!
It also wouldn't last for 18 hours if you did, you would've noticed issues immediately. My theory would be the GPU had a faulty power stage that functioned enough to get past QC but didn't function enough to last more than a day. It happens. Products tend to follow something called the bathtub curve for reliability, where there's a bunch of failures initially (generally referred to as infant mortality) or manufacturing defects that occasionally get past QC, then are reliable for a while, then die off as they age. You're likely just one of the unlucky few that got a card in that infant mortality portion of the bathtub curve. You really can't. Try using the card and hope for the best, odds are it'll just work and be fine. Technically you could check for shorts/higher than expected voltage with a multimeter, though that is admittedly very tedious and if you don't see any physical damage on the motherboard that's highly unlikely.
Thank you for your thoughts! For tomorrow’s updates, consider a 1000W unit just in case. It’s also possible the PSU needs attention, but current functionality suggests everything is okay.
It wouldn't be the PSU since a crash would cause the GPU to smoke and spark. The issue lies with the GPU, as mentioned by @RONOTHAN.
It really isn't necessary—your 850W should be sufficient. If the system is functioning, I don't see a problem with the PSU. Yes, if the PSU were the issue, it could cause the system to overheat and damage the GPU. It would also likely consume more power than the GPU alone, which wouldn't explain why it's still working now.
If there were any issues, it would likely be with the capacitors or inductors right next to the PCIe port. They seem fine and not damaged. This isn't related to the motherboard; the graphics card probably wasn't working when you purchased it.