GPU overheats (possible bad factory thermal paste application?)

[Iggy Wolf]

So of course I recently RMA'd my Gigabyte RTX 2060 Super because it was running hot at 83C and sometimes higher, even at 100% fan speed. Gigabyte fixed it about as well as you can expect (read: NOTHING AT ALL WAS CHANGED) They might as well have not bothered and simply returned me the card as it was, and I wouldn't be surprised if they did. I know the GPU is thermal throttling because the boost clock jumps between 1800 and 1900, causing stuttering in the game.

Decreasing the temp limit as well as the power usage stabilizes it somewhat, but only because I'm artificially throttling it myself at that point. I'm not the only person in the reviews for the card to have an overheating issue, but if they're not gonna bother fixing it, I don't see the point in RMAing it a 2nd time. Could the problem possibly be bad factory thermal paste application, and would reapplying it fix the problem? Their idea of a "fix" was to simply put it in and test the BIOS (if I'm complaining about overheating, WTF does that have to do with the BIOS?)

[Bad 💀 Motha]

That should alot way more then how it is now yes.

Side fan as an exhaust often can mean it's disrupting or fighting with the cpu and gpu, which are grabbing cool air to push at those heatsinks.

[SeriousCCIE]

I'd experiment with the fan placement and try gaming while a temperature monitoring program was running. This way you can determine the optimal fan placement.

On my old desktop case I ended up attaching fans where there was no mount, but where ventilation holes existed, because drawing air out (or forcing air in--depended on the location) helped significantly.

I ended up having a 140mm and a 92mm on the panel/side; the 140mm blew air over the GPU/PCIe slot area. The other fan was higher up and pulled air out. Based on temperature readings, the overal averages were improved with those additions.

Mind you those were not the only fans nor cooling solutions in place, but these proved to be excellent use cases for spare fans I had and had noticable in-game impacts due to keeping the system cooler for longer before any GPU throttling took place.

Also, I would suggest that if you can handle the stress levels involved with the GPU thermal inteface material replacement, have at it. I replaced mine with some of the liquid metal cooling material, and also replaced all of the vram and component thermal pads. It was not exactly inexpensive, but helped a great deal in my quest to overclock and keeping temperatures stable.

There are a great many types of thermal materials out there; and might prove to be a value if they help you accomplish your goals. That said, many people can't be bothered, but I guess they pursue different means of accomplishment than an enthusiast might do.

[Zireth]

I agree with this^

I'd try putting your 120 to intake and have 3 exhaust (one back, 2 top)

[Iggy Wolf]

One thing I noticed is that while playing, despite me putting a temp limit of 75C in MSI Afterburner and 88% power limit, the card still managed to exceed the temp limit by 4 C, putting it at a max 79C. This is why I believed it was overheating when put at its initial default settings. It's max temp limit at default is 83 C at 100% power usage, but the card manages to exceed that and rise up to 87C. That's why I'm assuming it's either my poor airflow or bad factory thermal paste application. Or that these Gigabyte cards simply have a ♥♥♥♥♥♥ heatsink. It's noticeable that even at 100% fan usage, it's still quieter than my old 970, which had a max temp limit of 79 C and NEVER exceeded that at 100% fan usage and power limit, but was louder.

[Iggy Wolf]

Escrito originalmente por Rachel, The Promised End:

I agree with this^

I'd try putting your 120 to intake and have 3 exhaust (one back, 2 top)

I don't have a gaming case as some people are assuming here. There's no grille on top to put a fan there. The only places for fans are the side panel, the back, and the front. It only allows for 3 fans. That's why I'm assuming that the way I should set it up, is to place an intake in the front, and make the 120 mm fan intake, so it'll pull cool air unto the CPU which is right below it, and the back will be exhaust, to pull the hot air out of the case.

[SeriousCCIE]

Oh I have punched a bunch of holes in my cases in the past. I've also drilled them and put gaskets in them; the hardest part can be getting gaskets of the right size that fit...

Your limit of 3 fans is what the case allows; what I had done was add fans via duct tape and balancing via gravity. I had to press my 140mm fan into the seams of the side panel, where this allowed the fan to be situated on an angle--instead of 'flat' like a square, it was sideways like a diamond.

The bottom point was what was held into place by the panel seam. The top 92mm fan I ended up screwing into place by utilizing one of the post holes for the power supply that was at the top of the case--there was room for larger supply (or a longer one anyway) and a free hole, so, I got a long screw to keep it in place.

I also learned that fan filters help a lot when you add a lot of fans that aren't necessarily part of the original engineering standard of design. However, they also clog... so the filters had to be cleaned regularly. Being outside of the case, it was easy to see just how dirty things get, but my elderly mother told me that is because I never vacuum. She might be right about that.

That said, the most important thing I can say is: You should do as you wrote and see how it works and don't be afraid to experiment. If you opt to poke holes in the door and are OK with that--I endorse that too, for what its worth.

The only risk is it being ugly when you are done. But if you leave the fans in place after the creative installation of them, no one will see the holes and only you will know they were there.

(I have specifically kept old cases and cut the sheet metal/put holes in them etc to help excessively abuse older hardware to perform past their expected returns on investment--so an ugly computer here and there doesn't bother me, provided it can deliver the goods!)

[Iggy Wolf]

Escrito originalmente por SeriousCCIE:

Oh I have punched a bunch of holes in my cases in the past. I've also drilled them and put gaskets in them; the hardest part can be getting gaskets of the right size that fit...

Your limit of 3 fans is what the case allows; what I had done was add fans via duct tape and balancing via gravity. I had to press my 140mm fan into the seams of the side panel, where this allowed the fan to be situated on an angle--instead of 'flat' like a square, it was sideways like a diamond.

The bottom point was what was held into place by the panel seam. The top 92mm fan I ended up screwing into place by utilizing one of the post holes for the power supply that was at the top of the case--there was room for larger supply (or a longer one anyway) and a free hole, so, I got a long screw to keep it in place.

I also learned that fan filters help a lot when you add a lot of fans that aren't necessarily part of the original engineering standard of design. However, they also clog... so the filters had to be cleaned regularly. Being outside of the case, it was easy to see just how dirty things get, but my elderly mother told me that is because I never vacuum. She might be right about that.

That said, the most important thing I can say is: You should do as you wrote and see how it works and don't be afraid to experiment. If you opt to poke holes in the door and are OK with that--I endorse that too, for what its worth.

The only risk is it being ugly when you are done. But if you leave the fans in place after the creative installation of them, no one will see the holes and only you will know they were there.

(I have specifically kept old cases and cut the sheet metal/put holes in them etc to help excessively abuse older hardware to perform past their expected returns on investment--so an ugly computer here and there doesn't bother me, provided it can deliver the goods!)

Like I said, I have no intention or interest in MacGyvering my rig just to make the GPU cooler. It's the only thing really that's hot. The CPU and the rest of the system stay cool. I just need that heat that gathers near the GPU AWAY from it and out the case. if the fans still don't improve its temps, I might just have to settle for replacing the thermal paste.

[SeriousCCIE]

Well.. if you want to Macguyver the GPU... as long as you are considering cracking open the GPU, consider getting a CPU AIO cooler bracket for it. (I say "if you want" because you mention the gpu repasting--so that was kind of tongue-in-cheek and kind of serious as well).

I have two of those NXZT "kraken" brackets; this allowed me to add a CPU all-in-one cooler. I have them in use in two different systems.

I used the AIO radiator in each situation (and case as it happens) as the exhaust fan mount; the 120mm radiator that came with each AIO cooler was secured in the exhaust fan location in the respective PC case.

Then I attached a push/pull on each side. I had to use an external fan control module to properly control the fan speeds; I used the gpu card power to handle the vrm cooling and the pump itself, then originally used a chassis fan header on the motherboard for the radiator fans, but soon learned that just because the card was busy doesnt mean the motherboard had any idea. I had to control that directly. That may be over the top for you; it is not uncommon though. I just had to remember to "turn the dial" when I started to play. It wasn't a burden to do so.

The NZXT/Kraken bracket by design exposes the vram and forces you to replace the thermal interface material on the GPU die itself.

I was able to overclock my r9-290s by 20% in this way; the only real problem I've had is gravity; the copper vram heatsinks I added sometimes fell off because the way the case is situated, the video card's vram faces down. If it was the other way around, it wouldn't be an issue.

The stickers on the copper vram heatsinks did not hold up over time, so I ended up having to use "arctic silver 5" thermal permanent adhesive. I had to wait a day for it to dry, as well, so it's not for the impatient.

Depending on how your case is causing the card to be oriented, that may not be an issue. Also, aluminum heatsinks may shed heat faster and are more light weight; I prefer copper but they were heavy enough to require additional efforts to secure them.

You do not want the heatsinks to fall off and hit a fan and get ejected at high speed, or worse, richochet back at the motherboard and short something.

Lastly, I recoginize that is far beyond adding a couple fans. I am not above double sided tape and haphazardly sticking a fan where it counts and pretending it doesnt look stupid, but I mention the other options because sometimes... after doing the easy path, I start to wonder how much better it can be if I put more effort into it. Maybe you will want to later? Either way, you're good to go with your approach. I expect it'll help!

[Iggy Wolf]

Escrito originalmente por SeriousCCIE:

Well.. if you want to Macguyver the GPU... as long as you are considering cracking open the GPU, consider getting a CPU AIO cooler bracket for it. (I say "if you want" because you mention the gpu repasting--so that was kind of tongue-in-cheek and kind of serious as well).

I have two of those NXZT "kraken" brackets; this allowed me to add a CPU all-in-one cooler. I have them in use in two different systems.

I used the AIO radiator in each situation (and case as it happens) as the exhaust fan mount; the 120mm radiator that came with each AIO cooler was secured in the exhaust fan location in the respective PC case.

Then I attached a push/pull on each side. I had to use an external fan control module to properly control the fan speeds; I used the gpu card power to handle the vrm cooling and the pump itself, then originally used a chassis fan header on the motherboard for the radiator fans, but soon learned that just because the card was busy doesnt mean the motherboard had any idea. I had to control that directly. That may be over the top for you; it is not uncommon though. I just had to remember to "turn the dial" when I started to play. It wasn't a burden to do so.

The NZXT/Kraken bracket by design exposes the vram and forces you to replace the thermal interface material on the GPU die itself.

I was able to overclock my r9-290s by 20% in this way; the only real problem I've had is gravity; the copper vram heatsinks I added sometimes fell off because the way the case is situated, the video card's vram faces down. If it was the other way around, it wouldn't be an issue.

The stickers on the copper vram heatsinks did not hold up over time, so I ended up having to use "arctic silver 5" thermal permanent adhesive. I had to wait a day for it to dry, as well, so it's not for the impatient.

Depending on how your case is causing the card to be oriented, that may not be an issue. Also, aluminum heatsinks may shed heat faster and are more light weight; I prefer copper but they were heavy enough to require additional efforts to secure them.

You do not want the heatsinks to fall off and hit a fan and get ejected at high speed, or worse, richochet back at the motherboard and short something.

Lastly, I recoginize that is far beyond adding a couple fans. I am not above double sided tape and haphazardly sticking a fan where it counts and pretending it doesnt look stupid, but I mention the other options because sometimes... after doing the easy path, I start to wonder how much better it can be if I put more effort into it. Maybe you will want to later? Either way, you're good to go with your approach. I expect it'll help!

Yeah, that's the interesting thing. The card on its own simply has a poor design. It's heatsink pipes are copper, and there's only two of them sticking out of the top. My previous ASUS STRIX GTX 970 had 3 chrome pipes. Two leading into the die, and a 3rd one that curved under the card outward on the right side and would pull the heat to the edge of the card. This one only has two pipes, and the aluminum block looks cheaply made. Overall, it's a poor card design that I'm not surprised it has a heat issue.

[Bad 💀 Motha]

Try...
Power limit: 80%
Temp limit: 75*C

When you see what works well for you, save this to a profile# and enable "oc on os startup"

However if you go to mess with an nvidi driver install, make sure msi AB is never running anymore, and not in os startup, until after driver is installed and rebooted

[Iggy Wolf]

Escrito originalmente por Bad 💀 Motha:

Try...
Power limit: 80%
Temp limit: 75*C

When you see what works well for you, save this to a profile# and enable "oc on os startup"

However if you go to mess with an nvidi driver install, make sure msi AB is never running anymore, and not in os startup, until after driver is installed and rebooted

Thanks for mentioning that. I didn't know. And currently, 84% power limit at 74 C works (74 because the card still goes 5 C higher than the temp limit, capping off at 79 C).

[SeriousCCIE]

What happens, Bad Motha, if you install the Nvidia driver and have MSI AB running?

I have done that, but not recently. I didn't notice any specific issue but that doesn't mean I hadn't rolled my figurative dice and just got lucky.

[[☥] - CJ -]

AB/Rivatuner cant attach itself to the GPU because it cant read the info
So installing drivers with AB open would likely cause the system to have to be restarted after the drivers are installed.

Software like AB should be closed when installing a graphic driver

[Iggy Wolf]

In the end, I've just installed a new intake and exhaust fan, putting it at 3 fans overall (and the max I can put in). Unsurprisingly, it gave me all of 1 C improvement, putting temps at max 78 C as opposed to 79 C. No amount of fans or thermal paste replacement can salvage a card's ♥♥♥♥♥♥ heatsink and cooling fans (the card's heat gets pushed out the bottom as opposed to the side, therefore pulling back in some of the hot air and ultimately keeping the GPU hot). The only positive perhaps is that it'll keep the rest of the PC cooler, moving that hot air away from the motherboard and out the back.

[SeriousCCIE]

Hey, but think of how much richer you are inexperience!

(That's what I tell *myself* after outcomes like that, anyway...)

CJ as to the comment about the rebooting, I guess I never thought to install drivers and not reboot anyway. No matter how well stuff has become in regards to handling that, I seem to be the type that finds problems when loading new stuff and putting off the restart until later.