My GPU killed itself... kind of

For the last few days I’ve been fighting some extremely weird performance issues. Randomly during a flight my frame rate would tank, and remain that way for the rest of the flight. I’ve never seen this behaviour before, as it never seems to recover even when the scene simplifies, like being at very high altitude.

I tried a few things, like driver updates, even using DDU for the first time. Going through the settings one by one, even trying unmodified. I also tried a BIOS update. Nothing seemed to work.

After one session I looked down at my case, which has a perspex panel on one side, and noticed that my EVGA 1080 fans were not spinning. That’s when the penny dropped.

Ever since MSFS was released, this has been the only bit of software that made my GPU fans “breathe”, where it accelerates a lot for a few seconds when under load. It looks like both fans have finally given up the ghost, and now the GPU itself is probably thermal throttling itself to avoid a melt down. It was at 98C after one session!

So now I am sadly in the hands of the scalpers if I want to do anything other than read email, or accept <20fps while flying which is no fun at all!

On the one hand I am relieved to have finally found what the problem was, on the other hand I am now looking at a £2000+ bill for a new GPU.

why not just get replacement fans? If the GPU is throttling, then its fine. (not fine at 98, meaning it hasn’t roasted itself yet… don’t run it at 98c)

4 Likes

What GPU? I’ll bet you can replace fans. AND… a good cleaning with some “air in a can” might help too.

98C is too hot.

1 Like

Yeah, that’s possible I guess. It’s an EVGA 1080 FTW. I’m not sure how easy it would be to replace them. It’s also a problem as this is my desktop I use for remote working, so if I break this I can’t work. Or at least until I can source a laptop from work.

I’m sure the card is fine. Corsair LINK is telling me it is currently sitting at 60C as I type this.

It’s done very well, so I can’t complain on that front. I bought it back in 2016.

Just did a quick search… found some suggestions on the EVGA website for fan replacement. Looks like it’s very “doable.” I really like the Noctua fans… quiet, very durable, long warranty.

4 Likes

Noctua would be a good choice. I have two of those in case already, and they are pretty much silent, with a 14” one on the front for intake.

2 Likes

Always always always set a fan profile more aggressive than the manufacturer and monitor their performance.

Good luck mate!

3 Likes

I’ve done a little more digging, but too late to do much else tonight.

From a cold start, the fans are spinning. But they don’t stay spinning, and keep coming to a stop.

I installed the EVGA XP1 software, and it allows for more granular control over the fans, including response curves. It seems to ignore them.

At lower temperatures, the fans will spin, but if I set it to say 100% the fans will very briefly spin up to 3500rpm, but for only half a second, then come to a full stop. If I set them to a lower speed it seems to work okay briefly, but eventually the temperature rises till it gets to a point where the fans just power off.

I’ll take the card out tomorrow to give it an inspection for debris or other obstructions. But the fans don’t appear to be dead,can’t perhaps more an issue with the fan controller.

Best of luck with your repair and research. Hopefully you can find the root cause, and not have it be the GPU. As you mentioned, a new GPU right now will cost a fortune…

1 Like

I ran another test this morning, with the EVGA Precision X1 software. I set up an aggressive fan curve, then notched them all up a small amount:

With the GPU cool, it obeys this, and sits at 40% fan speed, and is silent. I then ran MSFS, and fired up EGLL, and moved the drone up to a point overlooking the airport with London visible.

As predicted, the GPU temperature started to increase, and the fan kept pace. However when it got to somewhere between 70C-80C, it stopped obeying the fan curve completely, and started doing that breathing thing again, ramping between 0-3500rpm. Looking at it now the fans are not even spinning, and the GPU just hit 90C.

I don’t think the issue is with the fans themselves. I’ll let it cool, and try setting an even more aggressive fan curve. If I can stop it from getting to that state where it starts ignoring it, it might be okay. When it got down to around 50C it started obeying the fan curve once more, and the fans came on, and stayed on.

But I think I have to face the fact I need a new GPU.

Got a 1080 Ti from eBay for £75. That will tide me over till I can get something better. I wonder if it managed to find any Bitcoins. :wink:

It still could be the fans, if they are providing wrong rpm information the card could end up in some weird state.

I’d still remove the fans and the plastic shroud from the card and connect some other fans (e.g. your noctua case fans) to it to see if the card lets them spin constantly. You can use cable ties to fix at least two of such fans to the cooler.

All at your own risk, but I try some troubleshooting anyway.

Are you sure that it wasn’t 475? :sweat_smile:

You probably bought a defective card that can only be used for spare parts.

It says it is fully working, and if it’s not it’s on my credit card, and I’ll just charge it back via the bank.

I’ve been using non-stock fan for a few years now with a 980 Ti previously and a 1080Ti right now.
It’s very easy to just get mini 4-pin male to 4-pins female.
De-shroud your current GPU and its fans.
Cable tie some 90mm or even 120mm fans and connect to the 4-pins with some splitter.

1 Like

Maybe you are lucky and saved a bunch of money. I’m just not a big fan of used parts because you never know what the previous owner did. He could have overclocked the life out of it.

As long as it matches the description, and the fans work, I’m good. And if it doesn’t, I have eBay, and my Bank behind me so I have no fear of losing any money.

Another solution would be to just replace the fans with a water block kit. There’s a number of companies that make them. It means a bit of extra work and money, but would give plenty of cooling. EVGA might still have some of their own but the 1080s are oldish cards. There may also be folks who have already installed kits that have spare fans. While they probably don’t allow sale/trade posts on the EVGA forums, if you ask over there if anyone knows where you can get replacement fans for your card, someone might offer you some.

If you’re in northern hemisphere, since we’re headed into summer, a water block kit might be a good way to go depending on cash, case, etc. Fans are very important to these systems as you have now found. Folks should periodically clean fans, heat sinks, and radiators. They are great dust traps that decrease efficiency and cooling.

Good luck!

I don’t know how it is with EVGA, but it is normal with my MSI GeForce GTX 1070 ARMOR 8G OC. I also have the fans off in desktop mode. EVGA certainly does the same. Even more so after a cold start. Below 60°C, my fans automatically turn off. This feature is called “ZERO FROZR” at MSI.

1 Like

Just a quick update. I got in touch with EVGA, and they have very generously offered to send me a Thermal Mod Kit completely free of charge.

While I wait, my 3090 FE turned up. While it hasn’t bought me a significant increase in frame rates in busy areas, as I am CPU limited there, it has other benefits. I am no longer bound by 8GB of GPU memory, and I regularly see that over 9GB at EGLL.

It is quieter, and none of that breathing I experienced with the faulty card, so that’s nice.

I am no solidly CPU bound all the time. :slight_smile: The GPU is typically twice as fast, so it has bought me some overhead for whatever improvements come over from the work Asobo are doing on the X-Box version.

The GPU typically sits around 1% to 8% while parked.

3 Likes

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.