Computer Freezing and Restarting on Its Own? Check the Graphics Card
A desktop that freezes and restarts, mostly during video, was the graphics card, not the power supply or memory. The crash log said so: THREAD_STUCK_IN_DEVICE_DRIVER (0xEA). Reproduce the crash, read the log, then replace the right part.
A customer had a desktop that would freeze solid and then hard-restart on its own, roughly once a day. It happened most often while watching video. Nothing he did made it predictable, and nothing he tried made it stop.
Most people assume this is the power supply, or memory, or a virus. In this case it was none of those, and there was a way to prove it rather than guess.
Reproduce it before you replace anything
The first useful thing was catching it live. I played a full-screen video and the screen froze instantly. That is worth far more than a description after the fact, because a fault you can trigger on demand is a fault you can verify you fixed.
If your machine does this, pay attention to what you are doing when it happens. Video, games, and anything full-screen all lean hard on the graphics card. A crash that clusters around those is telling you something.
The crash log named it
Windows records why it went down. This machine's bugcheck was THREAD_STUCK_IN_DEVICE_DRIVER, code 0xEA.
That one is specific. It means a driver told the hardware to do something and the hardware never came back, so Windows waited, gave up, and restarted. In practice it points at the graphics card or its driver nearly every time. It is not a generic "something went wrong" code.
What that rules out, and why
This is the part worth understanding, because it saves you from replacing the wrong part:
- Not the power supply. A failing power supply cuts power. The screen goes black and the machine is simply off. This screen froze while still lit and still powered. That is a graphics hang, not a power loss. Different symptom, different part.
- Not malware. Startup entries, browser extensions, and network settings all came back clean. Malware does not usually produce a repeatable hardware bugcheck the instant a video plays.
- Not memory. Memory faults tend to throw different codes and crash unpredictably rather than locking to one specific trigger.
The card in the machine was an AMD Radeon RX 560, running a graphics driver from 2020. An old driver on a card that is starting to fail is a common pairing, and one makes the other worse.
Why the old driver has to be scrubbed, not just updated
When you swap graphics cards, installing the new driver over the old one is how you get a machine that is unstable in a brand new way. Fragments of the previous driver stay behind and fight the new one.
The tool for this is DDU, Display Driver Uninstaller, which strips the old driver out completely before the new hardware goes in. Skipping that step is a common cause of "I replaced the card and it is still crashing."
Matching the part to the machine, not to the catalog
Here is where it would have been easy to sell something bigger. This PC runs a Ryzen 5 1500X on an entry-level board with an entry-level power supply.
I put in a fanless, low-power GT 710. It is not an impressive card and that is the point. A high-end gaming card in this machine would be held back by the processor anyway, and it would ask more of the power supply than that supply can safely give. You would be paying more for no gain and buying a new instability in the process.
The right card is the one that matches the rest of the system and the actual job, which here is browsing and video. Fanless also means silent, and one less moving part to fail later.
Then prove it
After the new card and a current driver went in, I re-ran the exact full-screen video test that used to trigger the crash every time. Clean. No freeze, no restart.
That is the step that turns "it should be fixed" into "it is fixed." If somebody replaces a part for you and never reproduces the original failure afterward, nobody actually knows whether it worked.
If your PC is doing this
Note what you are doing when it locks up, and note whether the screen goes black or goes frozen. Those two point at different parts, and that one detail narrows it enormously before anyone opens the case.
I diagnose and repair this in your home, and I would rather tell you the cheap card is the right card than sell you the expensive one. That is what desktop and custom PC work should look like.
Questions people ask
What does bugcheck 0xEA mean?
THREAD_STUCK_IN_DEVICE_DRIVER: a driver told hardware to do something and it never came back. It points at the graphics card or its driver nearly every time.
Could it be the power supply instead?
A power supply failure usually cuts power outright; it does not produce a graphics-driver stop code. The log rules it out.
How do I check why my computer restarted?
Windows keeps the reason in the Event Viewer and minidump files. I read those before recommending any part.