[SOLVED] Help with GPU fan noise under load

[Alpha]

Hi all, I've just build the following system with mostly new parts apart from a 2nd hand GPU.

Ryzen 5700x
16gb g.Skill Trident Z
Gigabyte RTX 2060 Super Gaming 3X OC 8GB Links: Gigabyte | TechPowerUp
1tb WD SN770 SSD

It was going well until I booted up Forza Horizon 5 and then the GPU fans ramped to 100% and stayed there, it sound like a mini jet turbine all the time when the game is running. Temps are a constant 89C with a hotspot temp of 107C. Even at 100% fan speed this does not change. I haven't even got the case closed up yet as I'm still tinkering with it. All other current games I've tried have the same end result, a hot card and me going deaf.

Ive tried following a video on Youtube from Optimum Tech and Tech Illiterate about undervolting but this has not made any difference that I can tell, I think I must be doing it wrong.

I also think I'm being a bit dim, I can't get the numbers in the spec linked above to tally with the figures I'm seeing in MSI Afterburner or in HWiNFO. Specifically the clock speeds, on the Gigabyte website it says this card should have a boost clock of 1710MHz and on the TPU site it says the base clock is 1470MHz.

In Afterburner curve editor after resetting everything the curve looks like this:

Am I getting totally mixed up? The clock speed it's showing at the top of the curve is 1995 MHz, how does that relate to the GPU clock speed?

After running the Time Spy benchmark in 3DMark HWiNFO reports this:



The trace from 3DMark looks like this and as you can see the as soon as it starts the temps shoot up and to me it looks like the pink line on the bottom chart shows that the card immediately thermal throttles? It goes from 1470MHz (Base) to 1200 MHZ and wobbles around from there until returning to 1470 MHz at the end of the test. I think throttling is also demonstrated between the demo and start of test one, as the temp briefly drops the clock speed increases for a short period before dropping again



Is this normal behaviour for the 2060 Super's? The seller I got the card from (another site) said he had cleaned the card and replaced the thermal paste and thermal pads, using Arctic pads and paste, but I'm wondering if he didn't quite get it right for it to immediately throttle.

Any help or suggestions greatly appreciated, thank you.

 

[Angilion]

It's not normal behavior for any card and certainly not that one. It's supposed to cool efficiently, more so than average. I wouldn't expect to see GPU temps over 70C or fan speeds anywhere near 100%.

That image shows HWiNFO reporting a max graphics card fan speed of 4359 rpm! That's remarkably high. Really, really high. Higher than I would have thought possible. About double what I would have expected as 100% fan speed. Maybe it's misreporting the fan speed because that's remarkably high.

Boost clock for that card is 1815MHz, according to Gigabyte's website. I think you were looking at a different Gigabyte 2060 Super.

You're interpreting the pink line correctly. Your card is thermally throttling. A lot. With the fans at max speed. Something is quite badly wrong with the cooling on your card.

I couldn't find a review with detailed noise level monitoring, but I did find several reviews covering that specific card with performance, fan speed, temps and power draw. It's supposed to run up to very slightly hotter than I was expecting, peaking at 72C. At clock speeds well over 1800MHz and quite low fan speeds. The version you have to clocked to the max out the box, so it pulls more power and generates more heat.

Here's a review that covers most of what you should be getting:

Gigabyte GeForce RTX 2060 Super Windforce OC, Gaming OC Review: Super Performance, Good Cooling

Three fans or two?

www.tomshardware.com

Averaging 1899MHz clock speed during testing. Quieter than the reference design 2060 Super despite pulling significantly more power to reach those higher clocks. Max fan speed 2100 rpm.

Something's quite badly wrong with your card's cooling. Given that it's been disassembled and the paste and pads replaced, I'm also wondering if that wasn't done quite right. Wrong thickness of pad, maybe?

 

[Alpha]

Thank you for having a look at this and confirming what I thought. I've double checked the model number on the card and it is definitely the GV-N206SGAMING OC-8GD with the a boost clock of 1710MHz. I think the one with the boost of 1815MHz has the same code but ends with a C.

Edit: Found this comment on TPU discussing the D version card and it confirms what you said above about max temps expected: Link

Unfortunately I can't return the card so I'll have to have a go at replacing the thermal pads myself. Interestingly in conversation with the seller he did say he's used thicker pads than standard. I've not been able to find much info on the thickness of pads used by Gigabyte, but one video suggests these thicknesses:


Are Arctic pads and okay or are there better alternatives out there that won't break the bank!

Thanks again for your help.

 

[Marine-RX179]

Just curious though...putting the thermal pads for memory etc aside, with the GPU itself hitting 89C and fan spinning at insanely high rpm, doesn't it point toward the GPU cooler heatsink is not making proper contact with the GPU itself (for example may be screws are not tighten down enough)?

 

[ED209]

Interestingly in conversation with the seller he did say he's used thicker pads than standard. I've not been able to find much info on the thickness of pads used by Gigabyte, but one video suggests these thicknesses:

If he has used thicker pads it could be causing the cooler to not sit\make proper contact

If you can I would replace with the correct thickness pads and see if this helps with the issue.

 

[Alpha]

Just curious though...putting the thermal pads for memory etc aside, with the GPU itself hitting 89C and fan spinning at insanely high rpm, doesn't it point toward the GPU cooler heatsink is not making proper contact with the GPU itself (for example may be screws are not tighten down enough)?

Bingo!!

I took the card out on a hunch following the reply from Angilion and all 6 screws were loose! When I opened it up I found that the paste was not covering the full GPU die in bother top corners and judging from the impressions left in the thermal pads they were not providing good coverage on all of the chips.



I applied some Arctic MX-4 paste that I had ensuring the full gpu die was covered and after putting everything back to gether re-ran 3DMark with the following results:



Max temps according to HWiNFO are 77.6C on the GPU and 89C on the hotspot.

The card now performs as expected with only minimal throttling as far as I can see. To say I'm relieved is an understatement and I can now get back to convcentrating on putting the rest of my new machine together.

I might get some new thermal paste to see if that helps even more as the tube of MX-4 is about 8 years old from when I did my last build.

Thank you both for your help and suggestions on this.

 

[Marine-RX179]

Glad you hear that you have found the issue!

On the note of the Thermal Paste, I don't think there would be anything wrong with your existing MX-4 as long as your don't use the first bit that comes out of the tube that might be slightly dried-up; with that said new MX-4 ain't that expensive, so might as well get some new ones even if you are happier with that.

Also for the paste application, there's no need to "submerge" the entire GPU within it- it could actually end-up trapped some of the heat within the paste when there's far too much. The purpose of the paste is just to improve the thermal transfer from the GPU to the heatsink by filling in the microscopic gaps between the two surfaces, so ideally you wouldn't want to put too much to the point to "spill out" from the side when tightening the heatsink; typically a small bob or cross at the centre of the GPU would be sufficient.

 

[Eddie99]

I might get some new thermal paste to see if that helps even more as the tube of MX-4 is about 8 years old from when I did my last build.

Age of the thermal paste shouldn't be an issue, but MX-4 isn't very well suited to direct die cooling like a GPU due to the pump out effect. After a couple of months temperatures could start to increase significantly. From what I've read Thermalright TFX is the best thermal paste for direct die cooling, but Arctic MX-6 would probably be fine too as it has a much higher viscosity than MX-4.

 

[Alpha]

Age of the thermal paste shouldn't be an issue, but MX-4 isn't very well suited to direct die cooling like a GPU due to the pump out effect. After a couple of months temperatures could start to increase significantly. From what I've read Thermalright TFX is the best thermal paste for direct die cooling, but Arctic MX-6 would probably be fine too as it has a much higher viscosity than MX-4.

Brilliant thank you I'll get some ordered and make sure I just put enough on to just cover the die as per Marine-RX179.