Extremity

kevinc313

Extremity
My 3090 FTW3 with the hybrid cooler is getting 108-110c junction temps. It's killing me inside. I've opened it up twice to make sure everything is making proper contact and that none of the thermal pads have slid off or something, everything is fine. The performance remains surprisingly good, but seeing that junction at 110c makes me lose my mind. While the junction is at 110c, the main temp remains around 48c while mining. I generally run it at 74% power and 85% fan speed (which I'm now increasing, I just hate running over 90% fan) and maintain 118-126mh on average.
 
Would replacing/upgrading the thermal pads make any sort of difference? I ordered two packs of 3mm, and one each of 2mm and 1mm for another card (MSI -- def needs them) and to have some extra. Can anybody give me any insight as to whether this would help, and which I should go with if so? I've read the entire thread but can't tell if the consensus is that thinner pads would be better to reduce resistance before the copper, or if thicker would be better to ensure they're properly sandwiched.

You have a bad mount or swapped the factory pads to some (fake) aftermarket ones from a dodgy vendor.  Or you don't have proper/sufficient paste between the AIO and vram plate.  Or you probably want to get a better fan config going on the rad.

Haven't swapped any pads. As i said, I've meticulously checked and rechecked that everything is making proper contact. Rad is using the two fans that came with it as pull, with more on the other side as push. Core temps are great, it's just the junction. I've ordered some gelid pro pads because i don't know what else to try. Any idea what sizes i should be using?
 
Also, tried with side panel off, big fan blowing in from it, fans blowing the backplate, everything. Only thing I haven't tried id adding heatsinks to the backplate, but that always seemed weird to me.

Extremity


Both the VRAM and rad fans are now running at 90% at all times. I hate having them this high, before they were at 72%. I'm still at 110c at the moment, although it was at 108c for about half an hour. Case has good airflow, I've got 6 140mm fans -- two intake, three exhasting through the CPU's rad, one exhausting out the back) and two giant fans in the front also as intake, I believe they're 220mm. That's not counting the fans on the GPU's rad. Ambient temps in my house are 76F, although the office gets to around 78-79F during the day as I have a lot of things running in here; and that's with my office having a dedicated midea wall AC unit, in addition to my central air.
 
My laser thermometer thing shows the air exhausted from my case at 33.6c. The GPU sensor temp right now, with the junction at 110c, is 47c -- that's at 100% load for the past ~4 hours.
 
The card is mounted vertically, not horizontally, which is just how my case is. It has some airflow across the backplate, but not great. I was actually worried the issue was that the intake fan on the GPU was too close to the glass side panel (it's got around three inches) and/or the airflow along the back of the card was not sufficient, so I completely removed the side panel to remove any possible obstructions and mounted a *very* powerful fan blowing air directly behind the card along the backplate. After 45 minutes of this it did quite literally nothing -- I had let the card cool down beforehand and it did take longer for it to reach these temps, maybe 50% longer, but it still hit 108-110c.
 
I'm totally at a loss. The thermal pads won't be here until the 16th, I guess a lot of people are buying them from Amazon as that was the Prime shipping date (and I ordered them like four days ago). As far as I can tell they are my only hope. I have also been looking into getting some various thickness of copper, most of it at 0.9mm. I have read some articles and watched some videos of people using it precisely cut to the size of the thermal pads. Using thermal paste on either side, they replace the pads with the copper and have *significant* reductions in cooling. With how much more efficient copper is at transferring heat, this really interests me now.

Extremity

kevinc313

Extremity


Both the VRAM and rad fans are now running at 90% at all times. I hate having them this high, before they were at 72%. I'm still at 110c at the moment, although it was at 108c for about half an hour. Case has good airflow, I've got 6 140mm fans -- two intake, three exhasting through the CPU's rad, one exhausting out the back) and two giant fans in the front also as intake, I believe they're 220mm. That's not counting the fans on the GPU's rad. Ambient temps in my house are 76F, although the office gets to around 78-79F during the day as I have a lot of things running in here; and that's with my office having a dedicated midea wall AC unit, in addition to my central air.
 
My laser thermometer thing shows the air exhausted from my case at 33.6c. The GPU sensor temp right now, with the junction at 110c, is 47c -- that's at 100% load for the past ~4 hours.
 
The card is mounted vertically, not horizontally, which is just how my case is. It has some airflow across the backplate, but not great. I was actually worried the issue was that the intake fan on the GPU was too close to the glass side panel (it's got around three inches) and/or the airflow along the back of the card was not sufficient, so I completely removed the side panel to remove any possible obstructions and mounted a *very* powerful fan blowing air directly behind the card along the backplate. After 45 minutes of this it did quite literally nothing -- I had let the card cool down beforehand and it did take longer for it to reach these temps, maybe 50% longer, but it still hit 108-110c.
 
I'm totally at a loss. The thermal pads won't be here until the 16th, I guess a lot of people are buying them from Amazon as that was the Prime shipping date (and I ordered them like four days ago). As far as I can tell they are my only hope. I have also been looking into getting some various thickness of copper, most of it at 0.9mm. I have read some articles and watched some videos of people using it precisely cut to the size of the thermal pads. Using thermal paste on either side, they replace the pads with the copper and have *significant* reductions in cooling. With how much more efficient copper is at transferring heat, this really interests me now.

Is the GPU rad intake or exhaust?  47C seems kinda warm for a core temp of a hybrid at 310w.  If I run my 3090 FTW3 Ultra at 310w, the core temp is 50C.  About 20C ambient right now. 

Keep in mind the ambient in the room is typically around 25.5 - 26c. GPU rad is intake, I didn't want already warmed air being put through it.

critofur

vulcZJust curious, although I think I already know the answer to this, but what are you doing that makes your memory junction temps get that high? There's only one thing I can do on my 3080 or 3090 that can get my memory junction temps anywhere near that high, and it isn't gaming.

I don't know for a fact, that our EGVA FTW3 cards are superior to Gigabyte 3090s, but, I would bet money that they are, so that's part of the problem behind what  I'm about to mention next: I've talked to tons of folks at my local MicroCenter and I met one fellow who had TWO Gigabyte 3090s actually CATCH FIRE (while gaming - he's not (at least wasn't yet) a miner but I explained to him how he could use NiceHash to help pay for his new GPU).   Some games get GDDR6X memory TOASTY, and, various apps used for rendering put basically a constant 100% load on the VRAM.
 
My old PC case that I'm using for my "main" PC doesn't have the best ventilation, but, I do leave the side panel off and point a 140mm fan at the GPU and I first had my EVGA FTW3 3090 Ultra in that PC for gaming and I was unhappily surprised at how HOT the memory temp was showing in GPU-Z.
 
Anyway, as I have "only" a 1440p gaming monitor I felt that the 3090 was overkill for games (any that I play thus far) so I pulled it out and stuck my 3080 in for gaming duty and now the 3090 has been sentenced to hard labor in the mines until some new, more demanding games are released at which point it will be freed to return to it's gaming glory... 
 
My new EVGA FTW3 3090 Ultra mines at 123.5 Mh/s at 295 watts and without any added heatsinks, the VRAM stays at a relatively chilly 86 degrees C I'm fairly certain it would happily sit there and run with no errors or problems for YEARS (unless some of the thermal interface material they used is a sort that "drys out").
 
And, for yet another anecdotal bit of "evidence": my old 5700XT ran very well w/no problems, until I ran Cyberpunk 2077 at which point it started crashing due to overheating (I briefly tried mining on the 5700XT, but decided to sell it to someone who was in need of a GPU and couldn't afford a new 3070). 
 
The single most important thing that's been said in this thread is probably about how a THICK "thermal pad" is actually not a great heat conductor and even some of the ratings shown on pads like the Gellid "Extreme" have a disclaimer that the rated performance is only achieved after significant compression of the pad!  I for one, don't want to be trying to use a 1mm thin copper "plate" to try to apply compression!  3mm would be OK though.  Also, yes, it should absolutely cover the RAM chips 100!!!