rtx 3080 Ethereum Mining and RTX 3080 VRAM Temperatures

[TheRedManX]

Some of you may already be aware of the issue with the VRAM Temperatures on the RTX 3080 and 3090 GPUs (Especially the FEs). HWINFO64 now shows the GPU Memory Junction temp. While gaming, the temperature of the VRAM on these cards loves to hover around 90c-103c! According to Nvidia, this is within spec and "should be fine". 

 

Like almost everyone else and their dog, I have also been following the current crypto craze and began mining with my 3080 while not gaming to experiment and earn a bit of side cash. However, these high VRAM temps seem to be exacerbated when performing any VRAM intensive tasks (e.g. Rendering or Mining). I have observed temps of 106-108c while maintaining a MH/s of 99 and 75% PL on 80% Fan Speed. If the fans are turned down lower, the VRAM will redline at 110c and throttling begins.

 

I engaged an Nvidia Customer Service rep via their live chat and asked if this behavior was normal and if operating these cards at this kind of temperature for 24/7 workloads would cause excessive harm (obviously using your GPU will shorten it's lifespan, but there's a difference between getting 3 years out of a card and getting 3 months). The customer service rep said that as long as the core clock temp was below 70, the card was working as expected the VRAM is within spec to go up to 110c. If it were to fail under these conditions within three years of purchase, it would qualify for an RMA. Now, weather or not Nvidia Customer Service is qualified to make statements like that, I'm not sure.

 

Would any of you see an issue with allowing GDDR6x to sit at around 104-108c for 24/7. You would assume that operating your GPU underneath the throttle point would be fine, but it's hard to tell how hot these chips are actually supposed to get.

Whats the community's take on this issue?2

[Mick Naughty]

First I’ve heard of a temp issue. Aside from crappy cards like msi and such. 
 

Better airflow always helps. Wouldn’t worry about it too much. An overbuilt card like the nvidia cards will be more than fine. I’d just run the fans faster if it’s an actual issue. They will last the lifespan of the card. 

[TheRedManX]

38 minutes ago, Mick Naughty said:

First I’ve heard of a temp issue. Aside from crappy cards like msi and such. 
 

Better airflow always helps. Wouldn’t worry about it too much. An overbuilt card like the nvidia cards will be more than fine. I’d just run the fans faster if it’s an actual issue. They will last the lifespan of the card. 

Oh yes, it’s actually an issue for all of the cards but it’s especially bad for the FEs. Igors Lab did a good write up about it: https://www.igorslab.de/en/gddr6x-am-limit-ueber-100-grad-bei-der-geforce-rtx-3080-fe-im-chip-gemessen-2/

[Atomic Explosion]

Remember to under volt your card, smart miners know this is beneficial.

[e22big]

On 2/25/2021 at 3:47 AM, Atomic Explosion said:

Remember to under volt your card, smart miners know this is beneficial.

undervolt seem to only help with core temp though, I am trying this out with 65 percent power limit and max fan speed on Asus TUF and memory temp never go lower than 100c

 

the only time I can get it down to 96 is with -300 on underclock on memory speed (but it hurt mining performance signficantly in doing so)

[dividido]

On 3/1/2021 at 1:56 AM, e22big said:

undervolt seem to only help with core temp though, I am trying this out with 65 percent power limit and max fan speed on Asus TUF and memory temp never go lower than 100c

 

the only time I can get it down to 96 is with -300 on underclock on memory speed (but it hurt mining performance signficantly in doing so)

I also have a 3080 tuf - with core @ -502 memory @ +750 and power limit @ 67 I was mining around 92 mh/s with ~415 efficiency at vrm 104c

 

I just tried using the optimize setting for nicehash quickminer which brought me down to vram @ 94c

 

optimize set:

gpu @1080

memory @1269

watt limit @217w

 

Still mining around 90-95 mh/s with efficiency ~420-430

 

fans definitely go higher (almost 100% all the time) but vram is solid at 94

 

gpu temp is ~44c

 

I had to close afterburner to properly apply the optimize

[e22big]

21 hours ago, dividido said:

I also have a 3080 tuf - with core @ -502 memory @ +750 and power limit @ 67 I was mining around 92 mh/s with ~415 efficiency at vrm 104c

 

I just tried using the optimize setting for nicehash quickminer which brought me down to vram @ 94c

 

optimize set:

gpu @1080

memory @1269

watt limit @217w

 

Still mining around 90-95 mh/s with efficiency ~420-430

 

fans definitely go higher (almost 100% all the time) but vram is solid at 94

 

gpu temp is ~44c

 

I had to close afterburner to properly apply the optimize

yeah - just figure this out but apparently it's just my ambient temperature, it's pretty hot from where I live and as soon as I turn on my aircon the temp got to around a proper 96 (with fan spinning nearly 100 percent all the time) 

 

I also use nicehash setting, they actually did a pretty good job, much better than my manual optimsation setting

[Knopfjimthereal]

Hi, I have an ambient temp of 22c. My 3080tuf is set to(MSI Afterburner): core-300; memory +~1000; fan 95; power limit 71%
 

95-97MH/s with 40c core and 92c memory.

Hope it helps 

[Ronin1759]

Have evga rtx 3080 ftw3 

 

T rex shows 101mhs 212w

Curve locked to 737mv

Core clock set to 1365mhz

Vram +1400

Latestest nvidia drivers.

Fan on auto for now testing and sitting at 72% in afterburner. 

Pw limit set to 100, still going to slowly lower and see but using curve and lock this effects things slightly different.

 

Gpu temp at 44c in afterburner

Hwinfo

Gpu hot spot 58c

Mem junction sitting at 80c

 

This card seems to be performing very well 

 

[copytco]

Hi, I am new to this, so got a question. How to put actual Hz in NHOS config? In docs they say do give numbers for change of Hz, so + or - value to add to nominal frequency. Currently my card is running on default with 57 C on core, but not sure how to check RAM remotely, as SSH is not set (yet).

[Consaka]

I have 6 different 3080 and one 3090. The 3080's are EVGA and Gigabyte I believe and honestly every single one has different heat outputs mining the exact same thing on the exact same settings. They only get close when I take them apart and modify the heat sink pads. Even then they are different. I bought one used card and it literally throttled itself immediately. The main GPU temps were low but the VRAM was super high.  I could get it to run without throttling if I just used it for windows. ANY game and it throttled. I could get it to not throttle while mining if I used MSI Afterburner and turned the fan on high and all the other settings as low as I could set them. Of course it mined like crap. 30 to 40 mh/s if I remember correctly. Basically a stock 3060 speed.
After modifying the heat pads its running 100 MH/s 229 watts GPU temp is 47c VRAM is 100c Hotspot is 64c. Thats daggerhashimoto on Excavator v1.7.3b

[Boying]

After applying Gelid Thermal pad with 15 conductivity, my Gigabyte RTX 3080 Vram reduced from 110c down to 74c.  The GPU is at 49c running at 100% power and mining at 100 m/h.  I do have an extra fan blowing to my rig helping at least 2 degrees lower.  I have 4 Gigabyte RTX 3080 and some card comes with some good thermal pad and 2 other card have cheap thermal pad and some are missing thermal pad on the back place and some do.  It is not acceptable for Nvidia to say that its normal at high temp.  I get 45-47c and  72-82c on GPU and Vram respectively with an ambient temperature of 80f.  These card manufacturers should have the card with the right thermal pad and hot have some missing on some area.