Graphics cards dying from CPU load... Any idea why?

BananaBandana

Reputable
Jan 19, 2015
4
0
4,510
This is gonna be a long post so I suspect it'll be ignored but I guess I'll post anyway.

Basically a couple of months ago I was in the middle of a heavy CPU task when my graphics spontaneously died. I was using Unreal Engine and baking light. To be clear: this is CPU bound. It is a game engine yes, but the static lighting is done on the CPU only. That is important.

Story:
I spent about a day solving the problem, even contacted EVGA to ask if what had happened was even possible, and eventually concluded that it was a freak coincidence and my card just happened to die at that moment.
Especially since no one else has experienced this issue and probably for good reason since everything I've ever been told is that software can't really harm hardware outside of the obvious problems with thermals. Which there weren't... and again, it was the CPU being used.

I got my RMA, and everything was fine, barring the fact that I hadn't attempted what I had last time. I worked in Unreal for months, played games for months, and no issues. But when I finally went back and tried the lighting bake again, what do you know, the same thing happened. The circumstances are exact(even the part where it happened about 15 seconds after playing a video on YouTube). But... what?

More information:
Unfortunately I can't remember if I had done this before without it happening, but I eventually realized two things that happened 6 - 7 months before. One was getting new memory... which I highly doubt has anything to do with it, but it's the only thing that I actually did.
The other is probably more important: my motherboard got hit by lightning. Power was plugged into a UPS and the lightning seemed to hit a phone line anyway, which went through the modem and into my ethernet, killing the port. I was initially relieved that that was the only thing that got damaged, but it seems that might not have been true anyway. Also before anyone yells at me for not having my ethernet plugged through the UPS... I have DSL, okay. It cuts my already low bandwidth. Plus this lightning happened right at the beginning of the storm and I would've unplugged.

So yeah. So far the only suggestions ]are the motherboard, and power supply but basically I'm hoping someone who knows more about how computer hardware works can shed some light on how something like this could happen... Idle GPU + loaded CPU = dead graphics? My power supply has always been great, but I guess it would also make sense that if there were a problem, the CPU drawing power could have freaked it out and sent some pulse through the PCI cables? The lightning and the motherboard seems so likely but on the other hand I don't know if something could go all the way to the PCI slots and kill it.

I wasn't in the market for a new computer so I just want to know what I have to replace before I am comfortable that something isn't going to kill another graphics card. I'm kind of living on the edge right now and don't exactly want to have to keep killing cards until I get rid of the problem.
 
Solution
A more likely scenario is the CPU load is causing voltage droops. The current has to increase to compensate to maintain the same wattage. Or it is possible one of the voltage regulators on the motherboard is not handling the load the CPU is putting on it. That could cause the voltage on the PCI3e port to be out of spec. You need to monitor your voltages to see if any of these scenarios is happening.

They are all more likely if you have a CPU that isn't supported by your specific motherboard or if you have a low quality motherboard or power supply.

kanewolf

Titan
Moderator
A more likely scenario is the CPU load is causing voltage droops. The current has to increase to compensate to maintain the same wattage. Or it is possible one of the voltage regulators on the motherboard is not handling the load the CPU is putting on it. That could cause the voltage on the PCI3e port to be out of spec. You need to monitor your voltages to see if any of these scenarios is happening.

They are all more likely if you have a CPU that isn't supported by your specific motherboard or if you have a low quality motherboard or power supply.
 
Solution

BananaBandana

Reputable
Jan 19, 2015
4
0
4,510


Thank you for the response! This was the kind of thing I was hoping for since I know basically nothing about the electrical aspects. I assume that, if either of those things is true, it's probably not a fixable thing?

My parts are pretty quality; I have a Seasonic Platinum power supply and an Asus ROG Z87 board with a 4770k, so I don't think compatibility is an issue.