Question Fried GPU, now unsure what caused it

Cozzolino

Prominent
Jun 28, 2019
11
2
515
0
First, my specs:

OS: Windows 10 Pro 64bit (fully updated)
Mobo: AsRock G41c-GS R2 [link]
CPU: Intel Core2 Quad Processor Q9400 [link]
RAM: Crucial (2x4gb) DDR3L-1600 UDIMM CT51264BD160B [link]
Fried GPU: AMD Radeon Sapphire R7 265 2GB GDDR5
PSU: Corsair VS Series 650W [link]

Some details on my build: A P5Q-pro mobo fried when a lightning struck close to my home, it fried my original GPU and PSU aswell. CPU and RAM survived, so I bought the best mobo I could find for my CPU, a new PSU, and later decided to upgrade RAMs because why not and it wasn't easy. I finally found the best compatible ram I could get and had to use it at 1000mhz instead of the original speed and has been holding up for a while, until I forced a USB in the back panel which bent the pins and caused an immediate shutdown. I figured the problem was the bit of metal separating two USB next to each other which was slightly moving and touching the bent pins, so I bent the pin back to their original position, put a piece of folded paper in between the pins and the metal sides of the usb sockets and it has hold up just fine for months now. I know, I know. This is pretty much the whole build history.

Onto the problem: I was playing Civ V Vox Populi and the screen just froze with vertical violet lines. Upon rebooting I was met with a black screen right after Win10 screen so I started the usual tests that took two days and involved a friend air compressor catching fire and finding out about the sad state of my HDD integrity. Here's what I tried and found out:

  • GPU is undetected in Device Manager
  • GPU fans still spin
  • A 100% reproducible black screen while installing 4 different versions of AMD drivers (including latest) on a fresh win10
  • DDU was used for each uninstall
  • Used another working PSU to no avail
  • Used another HDD to no avail
  • Put my GPU in another throwaway build and the black screen appeared again upon installing AMD drivers
  • Said screw it, and unscrewed my GPU to see what's inside, it looked mint with no evident sign of wear or blown capacitor - there was some dust nevertheless - I cleaned the PCI pins and replaced thermal paste for good measure, to no avail
The GPU is dead beyond reasonable doubt.

But I have no idea what caused it. I have performed so far:
  • chkdsk to repair my HDD
  • sfc /scannow for good measure
  • Windows Memory Diagnostic Tool twice
  • Half a hour of Memtest64 (9 loops), and plan to run a 4 hour test tonight
Another friend lend me his GPU, which is sitting on my desk at this moment, is pretty good, and I'd be very sad (and broke) if my system fried this one too. I am unsure what exactly caused the GPU to fry in the first place so here we are, I just want to know what's the chance some component fried the GPU without frying itself in the first place (I'm currently using the same PC, just without the GPU and working with the integrated graphics). Would you suggest buying a second hand throwaway card and testing the system before putting a good GPU inside? That is probably the best way to know, but is there something else I could be doing before going through the painstaking and long process of finding and buying an affordable graphic card while my entire country is still on Corona lockdown?

On another note (very anecdotal evidence follows) yet another friend had the very same GPU which he bought about 6 months earlier. It fried too, the two GPUs had pretty much the same lifespan (and at least I'm sure I cleaned my case from time to time). Is planned obsolescence a thing now (I'm out of the loop)? I would rather give up PC gaming than make some bright fatass rich with his mastermind plan to slowly and systematically turn every component of my PC into very expensive paperweight while sucking the earth dry of all rare metal.

On yet another note, let's say I get a second hand GPU and it gets fried again, I'd be looking for a new Mobo and CPU and would want to keep my brand new RAM. What's a decent budget LGA1150 mobo fitting both my current RAM and a - let's say - i5-3570 or whatever close? Feel free to ignore this question, as it is very premature and pretty much unrelated.

Thanks for reading, I'll try to provide all info you might need upon request.


EDIT: I was unsure where to put this thread in the first place as it is only partially related to GPUs, but my broken english was probably at it again, so can a mod move this thread or close it so I can repost in the correct section? Thanks
 
Last edited:

Cozzolino

Prominent
Jun 28, 2019
11
2
515
0
Just a little bump with what I dug up on the internet.

It turns out that I own the infamous Corsair VS650 2012, the one with the orange VS print. It's rated E tier in the PSU tier list and has a pretty worrying design defect that sounds threatening although I don't understand the complete ramification of the issue (quoted below).

From tomshardware, Corsair VS650 Power Supply Review: Affordable Yet Capable:
This is an outdated design using a double-forward topology in the primary side, whereas in the secondary side a number of Schottky barrier diodes (SBRs) regulate the rails. The major downside is the lack of independent circuits for the regulation of the +12V and 5V rails.

This means that the performance under highly unbalanced, among the rails, loads will be bad. For example if we apply a light load at +12V and a high load at 5V (and vice versa) the corresponding voltages will go crazy. This is because those two rails are tied together so when the 5V rail has to cope with a heavy load, the feedback signal increases its voltage level affecting as well the 12V rail's voltage, although the latter might only have a light load to face. Whenever this is the case the 5V rail will struggle to keep its voltage in control while the 12V rail's voltage will go sky high (see the CL1 test in the 10-110% load tests).
You can find the Cross Load tests on the second page under "10-110% Load Tests" (I can't tell what's wrong exactly)

Although this pretty much answers my original question (does it?), I am now very concerned and fairly confused. What would be a realistic condition in which such issue occurs? I understand the 5v rail is used to feed USBs, speakers, mics, Hard disks, the CD/DVD reader and similar. Yet this ridiculously overpowered PSU for my build should never come close to such a thing as "heavy load", unless I'm missing something obvious? Either that or the underlying design issue slowly toasted my GPU, which doesn't seem too far fetched either.

While I'm waiting for an informed reply, I've identified a decent PSU off the tier list. it's the Corsair CV 650W, rated B tier which seems fair enough. Now if only the market wouldn't be doing its quarantine shenanigan...
 

ASK THE COMMUNITY