• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Gigabyte 3070 Gaming OC Temp issue?

Tony359

Tony359

Tony359

Associate
Joined
22 Nov 2022
Posts
42
Location
United Kingdom
Hi all,

I've got a Gigabyte 3070 Gaming OC today - second hand. I read on reviews that the card runs at 68-70C under gaming load. Well mine doesn't! It skyrockets to 79-80, fans run at 100% and it throttle itself because of temperature. This is with the "OC" BIOS but I read that the difference is just a couple of degrees with the Quiet bios.

After undervolting, the situation is much better, I get 75C maximum. I'll open it up and replace thermal paste but I'd be surprised the paste degraded so much as this is a 3YO card. Can someone confirm that this should be a relatively cold/quiet card?

I hope it's the paste, when fans were running at full speed (they were loud!) the air coming out of the heatsink didn't feel particularly hot, hopefully showing a lack of heat transfer from the GPU IC.

Thanks!

www.techpowerup.com

Gigabyte GeForce RTX 3070 Gaming OC Review

Gigabyte's GeForce RTX 3070 Gaming OC is a factory-overclocked custom-design variant of the GeForce RTX 3070. It comes with a large power limit increase, and temperatures are still much better than with the NVIDIA RTX 3070 Founders Edition.
www.techpowerup.com www.techpowerup.com
 
Last edited:
There are some variables which you might need to take into account. Your temps seem a bit extreme, but could be explained.

1. is it getting enough airflow in your case? (Is it horizontal or vertical mounted? How far away from a panel is it? Etc)
2. What were the ambient temps in the room the PC was today?
3. Was it in direct sunlight? (Today was fairly sunny)
4. What CPU cooler do you have? (Air coolers tend to heat the area on the back of the GPU more than water coolers with the rad somewhere else in the case)
 
Last edited:
Thanks for your reply.

It's a very large case, I have 4 120mm fans in it. The case was open when I tried which I appreciate it could make things actually a bit worse. No direct sunlight. Cooler is BeQuiet but blows in front of a 120mm fan which is expelling air out. Airflow is front to back - 2x120mm at the front and 2x120mm at the back.

I also see up to 22C difference between GPU temp and hotspot. What puzzles me is that the card seems to enter a "thermal emergency" mode where fans are spinning at "more than full speed" despite the software reporting 80-90%. Temp limit is set to 83C but I see I am in thermal throttling at around 79C (GPU temp which is about 100C on hotspot).
 
I had one for around six months and never experienced any temperatures like that. I did have mine undervolted and was using a custom fan curve, but the curve was to quieten it down rather than deal with high temperatures, as I found the stock one a bit too aggressive. It ran in the low 60s under full load in that state, so I'd suggest something's definitely wrong if you're getting 75 even undervolted.
 
I made a video of running 3DMark "time spy". This is with stock settings, "OC" BIOS. The card gets into thermal throttling seconds after starting the benchmark so I hope the thermal paste has gone dry. You'll also notice the fans are spinning up "higher" than what the software says, I think that's a reaction to the thermal throttle, some kind of HW protection which is invisible to the software. (Hint: it goes in thermal throttle when the "hot spot" hits 104C)


This is with an open case but it's 21C right now. Being a 3-slot card, I do have another card in front of the GPU but I feel the gap is enough to pull all the required air in.

I went ahead and disassembled it. Thermal paste was bone dry. Good.

I replaced it and also replaced the thermal pads. I measured them with a caliper, I measured 1mm so I replaced them with 1mm.

Upon putting it back together I noticed that the heatsink screws were giving me a spongy feeling as they were not properly bottoming out and I could see the plate holding on the GPU a bit bent? I tested it and... it went straight into thermal throttle when I tested. Panic!

I removed the heatsink again and saw this:
IMG_6312.JPG
IMG_6313.JPG
IMG_6314.JPG


clearly the GPU is not contacting with the heatsink. I thought maybe my thermal pads were a bit stiff so I removed them and tried without - just to see the paste footprint. See below: much better though still not great. BTW I didn't notice an issue with the pads getting stuck where the screw posts are.

IMG_6315.JPG
IMG_6316.JPG


I put the old pads back, tested again and it's much better. I get 78C with unigine and the "hotspot" is now a much more acceptable +12C delta.

So the conclusions are

1. It looks that this card is not the best in terms of heatsink alignment. Any suggestions on how to improve that? I can carefully sand down the posts which coincide with the area of the GPU which doesn't seem to make good contact.
1. I need new thermal pads. Can anybody recommend something for this card? I don't see an intermediate size between 1mm and 0.5mm. Mine are definitely a bit stiffer than the original, maybe that is the issue.

Thermal paste is MX-4.

With undervolting the card is totally fine now don't get me wrong. But I still feel I can tweak things a bit more to get it to a better point.

Thank you!
 
Update: I've played with the mechanical side of things, I cannot polish the heatisink for a number of reasons number one being the fact that the pipes - where the die rests - seem to be flat but everything else is not so I can't flatten it unless I have a CNC machine! :)

Still, I tried a few things and I think I have improved it - but the paste still does not distribute as I'd like when the (original) pads are in. I think I gained an extra 2-3 degrees though - still 75C at stock settings though. Fans are quiet and there is no thermal throttle anymore which is good. But not as it should be apparently.

I think this is the last Gigabyte item I'm purchasing. My older 2060 is also Gigabyte and the cooling was also substandard.

So the next and - I guess - final step would be to replace the worn out pads, which I have moved around too many times. I see there are some very expensive pads around, what I am looking for is something which can replace the original ones because they're worn out, not because they're inefficient. I've learnt that there are some "super-soft" pads available, if someone could point me to the right direction, that would be great!

The below is without pads installed so not bad. But as soon as the pads go back... I hope that over a number of days they flatten a bit more and the distribution of paste improves.

IMG_6322.JPG



Thank you!
 
Yes, a bit too much but if the GPU made contact with the heatsink, it'd be pushed out. Last picture particularly that was one of the many tests, I didn't bother cleaning the previous attempts around the die.

From what I could see the previous owner did not replace the paste. It looked factory to me. The pads look factory to me too. But cannot be sure.

Thanks for the link! 0.75mm? 1.25mm? Those are impossible to find LOL! But would explain why my 1mm almost killed the card.

Any ideas of what I can do? add some thermal paste maybe? Obviously I'll have to go with the thinner type. Any recommendation of a reasonably good type without breaking the bank?

Thanks for now!
 
Last edited:
You need to use some soft pads. 1.5mm and 1mm. I like Gelid Extreme pads and Alphacool do some nice soft pads also.
Dont cheap out on the pads as you will regret it later.
 
The ones currently on the board are super-soft. It feels weird that they're responsible for this. I've found some 0.75 and 1.25 ones, I'll see how that goes.

Would I be able to run a test without pads on RAM? I don't think it would hurt the ICs for say 15 minutes?
 
Tony359 said:
Would I be able to run a test without pads on RAM? I don't think it would hurt the ICs for say 15 minutes?
Click to expand...
Yeah, they'll be fine. It's become pretty standard to cool memory with the main heatsink these days, but historically a lot of cards have left it without direct cooling. You also have the 3090 where the chips on the back of the card get essentially zero cooling on most models. Probably not great over a period of years, but they're not just going to burn up instantly.
 
Interesting, thanks. What is the MM?

And what kind of issues are being reported?

This is from some footage I did this morning when I disassembled the card. The paste looks factory to me (but yes, I know I'm speculating) and the pads looks pretty thin. I measured a small corner that tore off the larger one and I have to say it looks like 0.75mm. I assumed it was 1mm as I wasn't aware there were intermediate thickness.

3070factory_1.1.1.png
 
Tony359 said:
This is from some footage I did this morning when I disassembled the card. The paste looks factory to me (but yes, I know I'm speculating) and the pads looks pretty thin. I measured a small corner that tore off the larger one and I have to say it looks like 0.75mm. I assumed it was 1mm as I wasn't aware there were intermediate thickness.
Click to expand...

There definitely are intermediate sizes. I watch northwestrepair on YouTube and he makes his own with what he calls a "pasta maker" *, but I don't know what it actually is :D

* 5:44
https://youtu.be/vCzzUPBEVt8?t=345
 
Last edited:
Oh you mean actual pasta maker! LOL! ahahah I love ingenious people! (I do have that machine BTW...)

I found 0.75 and 1.25 and placed and order. Also got 0.5mm as there are quite a few threads online suggesting that 0.5mm is the right size for VRAM.

My plan is to test in an open rig to record current performance at fixed fan speed. Then remove the VRAM pads and test again. That should give me a reference.

Then I can try 0.75mm and 0.5mm and see which is performing better.

Slightly OT: I bought two second hand GPUs over the years. BOTH had dried out paste and were thermal throttling. Both went back to normal operations after paste was replaced. Maybe something big reviewers should look at? This 3070 is only 3 years old, if the paste is dry it means it was bad quality to start with.
 
Thank you!

TechPowerUp tested your card and measured 68C, the 3070 measured 70C. Do you know what you see with Furmark on your card?
www.techpowerup.com

Gigabyte GeForce RTX 3070 Ti Gaming OC Review

The Gigabyte GeForce RTX 3070 Ti Gaming OC is a factory-overclocked custom design sporting a new cooling solution. The triple-slot, triple-fan cooler lets air flow through the card and achieves excellent memory temperatures thanks to a new baseplate design. As soon as you activate the "quiet"...
www.techpowerup.com www.techpowerup.com

Now to be fair I do 75C now - I suspect that in an open rig with 20C I might do a bit less so maybe I am not far away from the target. I'll swap the pads anyways as they are tatty and will experiment with 0.75mm and 0.5mm. I like the method shown in the video @Tetras sent me: place pads, put the heatsink, press, remove, check whether paste has spread and also confirm there is a little indentation on the pads.

But I'll test without VRAM pads as well, that should tell me how far I can go with temps.
 
Tony359 said:
Oh you mean actual pasta maker! LOL! ahahah I love ingenious people! (I do have that machine BTW...)

I found 0.75 and 1.25 and placed and order. Also got 0.5mm as there are quite a few threads online suggesting that 0.5mm is the right size for VRAM.
Click to expand...

You can make your own too now then :D
 
You must log in or register to reply here.
Back
Top Bottom