My way of cooling the memory temps of the 3080 Ti

This has been my experience with the 3080 Ti with X Plane (GPU works very hard) and Davinci Resolve Studio.

My memory temps on this GPU would hit 110 when looking within the cockpit of the aircraft and maybe drop to 100-105 when looking out of the windscreen. Throttling happened at 110 but I always took it as "that's the way the card runs". Until a member here asked if I tried to under volt the card. That got me to experimenting with settings for the card within MSI Afterburner. I dropped the voltage and got maybe a 5 degree temp drop when looking within the cockpit in the sim which was better but I wasn't satisfied. But I found my solution.....running XP now at 85-88 C memory temps no matter where I was looking in the sim.

My power limit is set to 60% on the card....that's it. Well, I've laid a case fan on the case of the video card pulling air from the vents also. Everything else is stock settings. I have lost 1-2 FPS in XP and a render of a ten minute video in Davinci only takes 5 secs longer. Numbers I will gladly accept to keep my card going for a while.

I know this has not much to do with Davinci Resolve Studio but maybe it will help some of those out there that really push their 3080 Ti's and do not like high memory temps.

Depending on how old that card is and which manufacturer you'd be better off repasting it with something like PTM7950 or better thermal pads. If its at the limit and throttling its probably dried up paste.

Undervolting to solve a very high temperature is like rubbing alcohol on a wooden leg and it might be working only temporary.

Depending on how old that card is and which manufacturer you'd be better off repasting it with something like PTM7950 or better thermal pads. If its at the limit and throttling its probably dried up paste.

Undervolting to solve a very high temperature is like rubbing alcohol on a wooden leg and it might be working only temporary.

But I didn't under volt it. I reduced the power limit for a card that is notoriously bad for memory temperatures. I'm now running a cool 3080 Ti

Undervolting and lowering power limit go hand in hand. Setting a power limit to 60% is likely making your card underperform quite a lot. But if you're happy that's all it counts I guess.

Good luck.

bro!

bro!

Gross.

Air cooling for life!

Depending on how old that card is and which manufacturer you'd be better off repasting it with something like PTM7950 or better thermal pads. If its at the limit and throttling its probably dried up paste.

Undervolting to solve a very high temperature is like rubbing alcohol on a wooden leg and it might be working only temporary.

Undervolting is often attempted and is a good method to use but I think your suggestion is good, also - the PTM7950 or 'improved' thermal pads is often a way to 'fix' temp problems - ironically, the PTM7950 is often recommended for 7900 XTX cards.

'Anybody reading this thread, know if the 3090 has similar issues? I believe it can - but, some 3080s/3090s had more temp problems than others - for e.g. the Gigabyte 3080 series often had complaints from owners with the memory pads. The 3090 series *can* as sell depending on the brand/model. Undervolting is one method to deal with it but many ppl replaced the pads.

I was wondering if use in DR would result in these higher temps. It was just one concern I had to buying a 3090 for DR.

Undervolting and lowering power limit go hand in hand. Setting a power limit to 60% is likely making your card underperform quite a lot. But if you're happy that's all it counts I guess.

Good luck.

The only way I can tell the performance difference is looking at the FPS counter in X Plane (down about 1-2 FPS and by looking at the amount of time it took to render a 10 minute video in DV. (about 5 secs longer at reduced power limit settings). I wouldn't call that a lot of under performance hehehehe. The only thing that has dropped significantly is the memory temps. At stock settings, my GPU temps never even got close to maxing out or throttling....but my memory temps were always high, most of the time within the "allowed" range for the memory specs but I would notice a tad bit throttling if the memory temps hit 110. So now I have the performance I have without the odd throttling hit (stutters in X Plane)..simply a smoother running computer especially in XP.

By the way, I did try under volting but saw very little change in the memory temps on my card.

Edited to add:

Another thing I noticed is this. I use Intel Extreme Tuning Utility and use the OPTIMIZE button to overclock my CPU. Before reducing the power limit on my GPU, it would usually run at 4.9hz...now is running at 5.1hz. I really can't see why or if the GPU settings has any affect on the CPU (my guess temps are involved) but that may explain the very little loss of performance when reducing my power limit.

'Anybody reading this thread, know if the 3090 has similar issues? I believe it can - but, some 3080s/3090s had more temp problems than others - for e.g. the Gigabyte 3080 series often had complaints from owners with the memory pads. The 3090 series *can* as sell depending on the brand/model. Undervolting is one method to deal with it but many ppl replaced the pads.

I was wondering if use in DR would result in these higher temps. It was just one concern I had to buying a 3090 for DR.

Resolve can use high memory bandwidth because I recall either a 3070 works much better in Resolve than a 3060ti, OR a 3070ti works much better than 3070, and reviewers decided it was due to increased memory bandwidth between the models.

I remember people saying don't buy a 3080ti or 3090 2nd hand with limited to no warranty due to the hot ram problem, and expected shorter lifespan of card. I'm sure you know this by now but it's normally the ram on the rear of the card that overheats due to lack of airflow and some people do various physical mods to cool the ram.

i have both 3080Ti and 3090 in various machines
rarely see them above 70, normaly around 52-54
unbranded Nvidia server cards, blower /2 slot inside either a z820 / z840 / z8 workstation
cool clean air flow front and rear
really noisy... like really, really noisy
suspect the temps are from a gamer card with mutiple fans blowing heat around, but not removeing it well enough

for professional work blower cards inside a good workstation is worth considering, but if you cannot isolate the noise, gamer card and underclock to control heat seems like a decent option

i have both 3080Ti and 3090 in various machines
rarely see them above 70, normaly around 52-54
unbranded Nvidia server cards, blower /2 slot inside either a z820 / z840 / z8 workstation
cool clean air flow front and rear
really noisy... like really, really noisy
suspect the temps are from a gamer card with mutiple fans blowing heat around, but not removeing it well enough

for professional work blower cards inside a good workstation is worth considering, but if you cannot isolate the noise, gamer card and underclock to control heat seems like a decent option

Are you talking about GPU temps or MEMORY temps? My card's GPU temp always stays well below 70C for anything I do. My concern with the card was the memory temps which hit 110C at times which I've since corrected by reducing the power limit. I rarely hit 90C MEMORY temps.

GetTechPowerUp GPU-Z and look at the memory temps.

Are you talking about GPU temps or MEMORY temps? My card's GPU temp always stays well below 70C for anything I do. My concern with the card was the memory temps which hit 110C at times which I've since corrected by reducing the power limit. I rarely hit 90C MEMORY temps.

GetTechPowerUp GPU-Z and look at the memory temps.

I think you're better off on a GPU or gaming or overclocking forum, but I don't think reducing power limit has anything to do with power or voltage on memory, so either your fixing your hot ram problem by reducing the memory bandwidth alone(a bad thing), or by reducing the GPU temperature you help to improve your memory temperatures indirectly.

That doesn't sound like a real fix such as replacing thermal pads, doing the various backplate modifications such as adding more thermal pads or thermal paste to get more heat to back plate then use fan with heatsink to extract the heat. If a user frequently used Video noise reduction, enhanced 2x upscaling, speed warp, film grain etc they could see a slowdown where you don't.

Are you talking about GPU temps or MEMORY temps? My card's GPU temp always stays well below 70C for anything I do. My concern with the card was the memory temps which hit 110C at times which I've since corrected by reducing the power limit. I rarely hit 90C MEMORY temps.

GetTechPowerUp GPU-Z and look at the memory temps.

I think you're better off on a GPU or gaming or overclocking forum, but I don't think reducing power limit has anything to do with power or voltage on memory, so either your fixing your hot ram problem by reducing the memory bandwidth alone(a bad thing), or by reducing the GPU temperature you help to improve your memory temperatures indirectly.

That doesn't sound like a real fix such as replacing thermal pads, doing the various backplate modifications such as adding more thermal pads or thermal paste to get more heat to back plate then use fan with heatsink to extract the heat. If a user frequently used Video noise reduction, enhanced 2x upscaling, speed warp, film grain etc they could see a slowdown where you don't.

I started this thread because of a question on using the notoriously "hot" 30 series of video cards and if they are useable for DV . I'm just putting my two sense worth in. And no where have I found that lowering power limit on video cards reduces bandwidth. Lowering power limit only affects the hz in both gpu and memory....and in my case, barely noticeable. Also, these cards run more efficiently than running at full advertised or factory set power as throttle limits are never reached, therefore no throttling down.

And no where have I found that lowering power limit on video cards reduces bandwidth. Lowering power limit only affects the hz in both gpu and memory....and in my case, barely noticeable.

You are seeing performance degradation, this is happening because you're restricting the power to GPU by a large amount. Nvidia use a technique called power-gating where cuda cores are turned off to get better performance from the ones left on, That will lead to less memory throughput, but also your 3080ti is no longer a 3080ti, it has less effective cores and frequency won't boost as high (although higher than if there was no power-gating)

When I try a benchmark at 100% power the voltage is stable and frequency remains at it's highest clock frequency, but at 60% power there are declining voltage fluctuations which causes lower clocks. So even the cuda cores allowed to keep operating see a performance decrease. I don't think that' s ideal.

Undervolting and power limiting are two different things.

Undervolting is used to reduce voltage running through card (which in turn reduces temperatures) while still letting card accomplish maximum frequency it is capable of (which, in case of non-OEMs, might be higher than AMD/Nvidia/Intel designed them for). Because card is running at lower voltage than designed for owner needs to assure card is still running stable.

Power limiting is way to prevent card from exceeding power limit it was designed for / limit how much can it exceed it. AMD/Nvidia spec is 100%. Asus/ASRock/etc might feel their components and design can handle more and will put in place their own power limit (say 110%) and voltage curve but there is no guarantee 100% of them will be able to hit that 100% of the time, especially after years have passed. In turn some of them end up unstable and owner has to put in lower limit.

In other words, undervolting moves down the voltage curve across frequency range but does not change maximum frequency. Power limit does not change voltage curve down, it just limits maximum frequency. In turn they aren't mutually exclusive and one can do both at the same time.

If one has to lower power limit to less than 100% in order to prevent overheating and/or crashes then something is wrong with the card/system, every card should be able to run at 100%.

If one doesn't feel significant impact when running at just 60% then something (software used, CPU, etc.) is a bottleneck and not taking full advantage of that card to start with.

Y'all do realize that I'm talking about memory temps only, right? I have yet to get my GPU temp over 65C with anything I do with this rig. Even my GPU hotspot struggles to get over 80C. It's the memory junction temperature I was concerned about. That would hit 110C while running extreme software and I'm working to keep that down..I don't need throttling stutters in X Plane etc. Also , by keeping them down, I'll prolong the life of the card.

So, I'm doing an experiment right now...having run the OC Scanner in MSI WITH the power limit set at 70. Currently I'm hitting 1972 mhz ( set on a curve) on my GPU core clock, 9830 mhz on my memory clock (that's over clocked 200) and voltage is showing around .890 v. Gpu temp is 62C and memory junction temps are basically locked at 92C which is very ok for me.

If these mem temps stay where they are for the remainder of the flight , I'll be happy.

Please realize I'm not trying to argue....this seems to be the only way to knock down those memory temps

Are you talking about GPU temps or MEMORY temps? My card's GPU temp always stays well below 70C for anything I do. My concern with the card was the memory temps which hit 110C at times which I've since corrected by reducing the power limit. I rarely hit 90C MEMORY temps.

GetTechPowerUp GPU-Z and look at the memory temps.

i have GPU-Z already thanks... rarely bother with it as the GPU's rarely give me any reason to
right now memory temp = 62C playing a DCi4k timeline, Arri35 RAW, DVO and Neat in the node tree
mem temp = 44C (ish) at idle

put new paste on the cards a few months ago, kept the pads as they are doing ok

there's something to be said for a well designed workstation in a clean cool room.. you could fry an egg on the GPU vents, but inside the machine it's pretty reasonable temps

@ Harry,
just powered down the machine, will double check in the morning

Are you talking about GPU temps or MEMORY temps? My card's GPU temp always stays well below 70C for anything I do. My concern with the card was the memory temps which hit 110C at times which I've since corrected by reducing the power limit. I rarely hit 90C MEMORY temps.

GetTechPowerUp GPU-Z and look at the memory temps.

i have GPU-Z already thanks... rarely bother with it as the GPU's rarely give me any reason to
right now memory temp = 62C playing a DCi4k timeline, Arri35 RAW, DVO and Neat in the node tree
mem temp = 44C (ish) at idle

put new paste on the cards a few months ago, kept the pads as they are doing ok

there's something to be said for a well designed workstation in a clean cool room.. you could fry an egg on the GPU vents, but inside the machine it's pretty reasonable temps

44C on memory temps? Wow! What kind of card again? Oh, I just realized, that was at idle. What's the highest you get with your memory temps under load?

44C on memory temps? Wow! What kind of card again? Oh, I just realized, that was at idle. What's the highest you get with your memory temps under load?

mid 60's under load, exporting masters for a feature the machine can run for days and days at 100% and the GPU mem shows in 62-65C area, (this is the one time i do look at GPU-Z)

the 3080Ti and 3090's are unbranded cards built by Nvidia for servers, no name on them, hard to find, not imossiable tho

there's a new 4070Ti Super blower card out, about 25% faster than the 3080i/3090,

a 2 slot blower card is LOUD, and really meant to be in a machine with heat shrouds and designed to move air through front -> back and needs 24" clear clear air front and rear to make that work, these are reliable cards if those conditions are met, zero idea if they could work in a standard gamer box, prolly better off with a matching gamer card

44C on memory temps? Wow! What kind of card again? Oh, I just realized, that was at idle. What's the highest you get with your memory temps under load?

mid 60's under load, exporting masters for a feature the machine can run for days and days at 100% and the GPU mem shows in 62-65C area, (this is the one time i do look at GPU-Z)

the 3080Ti and 3090's are unbranded cards built by Nvidia for servers, no name on them, hard to find, not imossiable tho

there's a new 4070Ti Super blower card out, about 25% faster than the 3080i/3090,

a 2 slot blower card is LOUD, and really meant to be in a machine with heat shrouds and designed to move air through front -> back and needs 24" clear clear air front and rear to make that work, these are reliable cards if those conditions are met, zero idea if they could work in a standard gamer box, prolly better off with a matching gamer card

Well, those memory temps you get are what I get with my GPU temps lol. Good on ya for getting to those memory temps! By the way, hi from Vancouver Island