I have two 3080 RTX Founders Edition Cards.
When completely stock; one of them just about handled things on a reasonably performant overclocking setup.
The other one went up to 108°C so quickly it was frightening.
Quick enough that I thought it best to disable mining on the card while I fumbled around for something to make a screenshot with:
The cards are *so* inconsistent. You can put two Founder’s Edition cards next to each other in your rig and they’ll come out at entirely different levels of hashrate and temperature performance. The screenshot above shows stock cards at stock settings, no overclocking yet.
Case in point – my two 3080’s couldn’t have been more different.
It’s *always* the thermal pads
I’ve learned the hard way over the past few days but I feel like I understand this issue well enough to share my learnings. Basically, you can’t rely on the manufacturer *at all* to do a consistent job with the thermal consumables on a GPU. It’s a terrible shame but it does make sense if you think about what the production lines must look like.
Imagine you’re the guy who stood on a factory floor watching hundreds of GPU cases slide past on a conveyor. It’s your job to stick a few thermal pads in prescribed locations. Nasty, sticky things – and your job depends on your keeping up with a high-volume production line.
There’s no chance each pad is going to be placed consistently. And, to add to the problem, we know that the pad material we find inside NVIDIA’s 30 series GPUs are very poor quality.
The sudden overheating issue on my 3080 FE was this:
I’m reasonably confident this was the cause of the runaway temperatures on that one GPU. But this is the lottery that seems to be production quality assurance with the founder’s edition cards!
For mining, I think you’re pretty lucky if a card is even close to being able to handle a decent hashrate without a pad modification. Thankfully, there are lots of good guides to the procedure:
Guide to Nvidia 3080 FE Thermal Pad Mod (Will Norris)
3080 FE Thermal Pads (FreakoFreako – Reddit)
The consensus is to use thermal pads as follows:
Backside: Thermalright 1.5mm 12.8w/mk
Frontside: Gelid 2mm 12w/mk
Here are some of the things I’ve learned that is important to pulling off a successful pad modification.
Something I’ve learned, the hard way, is don’t choose pads that are too hard or crumbly. On my first attempt, I rather naively used Gelid Ultimate pads. Don’t do this! I was absolutely devasted to find my time had not been rewarded by temperature performance!
A nice squishy pad seems to surround the chip more nicely – you can test by pushing your GPU mainboard into the cooling plate and then removing it. You should see little imprints of the chips in the pads. Gelid Ultimates show almost no evidence a chip touched it at all!
Choose your pads carefully! The GP-Extreme pads are nice, as are Thermalright and Iceberg Thermal Drift Ice are all very nice to work with.
Honestly, your pad choice and thickness decision will make so much difference!
This is something that really caught me out. You can’t rush the job, the pad placement has to be right and neatness is everything. I appreciate the hypocrisy of this photo:
Note the coverage of the pads. I’ve covered the chips, no more and no less. I think allowing a gap so the chips are covered but nothing else is make a difference because a) you’re allowing what little airflow there is to move between components, b) you’re not conducting heat from something else and c) you’re not trapping heat unnecessarily.
This approach, leaving gaps and cutting things neatly is probably the biggest contributor to the thermal performance of my card. Note the image above – poor pad placement was letting temperatures shoot up.
Also note, top left – an earlier experiment with Ultimate pads. You can see how crumbly those edges are!
Use the right pad thickness. 1.5mm on the GPU die side, 2.0mm on the rear.
The reason why, as I’ve learned after a few different trials is that anything over 2mm on the GPU side adversely affects GPU cooling. I’ve used 2mm pads on the GPU side, and despite pushing them down and squishing them into the chips a bit, I gained almost 20c on the GPU temperature. That gap between the cooler and the GPU matters – and tiny distances have proportionally massive impacts.
1.5mm pads stop this from occurring.
This helpful tip from Igor helps:
The back of these Founder’s cards relies on a simple backplate for cooling. Igor at Igorslab.de used an infrared camera to locate the hotspots on his GPU and added a further 3mm pad in this area.
My card’s rear is a bit of an interpretation of a theme, let’s say:
A lot of what I’ve done above turned out to be unnecessary, and I’ve since removed all of those little bits that don’t do anything. Aim for something like this:
VRAM down to 82c at 96mh/s!
Finally, then, the 3080 that had those terrible runaway temperatures (bottom) is now outperforming the stock 3080 (top). Happy days!
Thermal pad modification does work provided that you choose the right pads and focus on doing a nice, neat and tidy job. I’m delighted with the outcome of this, only 9 more cards to do 🙂