Skip to main content

GDDR6X in 3080 and 3090 Hits 110C While Mining Ethereum

GeForce RTX 3080 FE
(Image credit: Tom's Hardware)

Popular hardware monitoring tool HWInfo64 has just received a new update, version 6.42 that adds support for "GDDR6X Memory Junction Temperature." This feature will allow you to check your VRAM temperature if you have a GDDR6X equipped Ampere card like the RTX 3080 or RTX 3090. This could be particularly useful if you own a Founders Edition model, as several months ago we reported on a discovery from Igor's lab in which GDDR6X modules on the RTX 3080 Founders Edition model ran dangerously hot.

After some testing with the new HWInfo64 version 6.42, we've validated these same high VRAM temperatures that Igor discovered. Looping Metro Exodus at 4K Ultra settings for several minutes on an RTX 3080 Founders Edition allowed the GDDR6X modules to hit a peak temperature of 102C. The TJmax for GDDR6X is 95C.

However, this changes when enabling DLSS and Ray Tracing. The RTX 3080 with Metro Exodus at the same settings hit a peak temperature of 94C. We also tested an RTX 3090 Founders Edition in Cyberpunk 2077 with DLSS and Ray Tracing enabled. GDDR6X temperatures for that card peaked at 100C.

But when it comes to Ethereum mining, temperatures go to a whole other level: When mining on both the RTX 3080 and RTX 3090, we found that the GDDR6X modules would peak at a much higher 110C, and the GPU would downclock itself severely to compensate for the ridiculously high VRAM temperature. This occurred on multiple different boards, from various vendors. And that's before applying any overclocking settings, which some miners like to do in order to chase every last bit of hashing performance.

There are ways to reduce the GDDR6X temperatures, which appear to cause the 3080 and 3090 to throttle only once the temperature reaches 110C. We used MSI Afterburner to drop the GPU and GDDR6X clocks as far as possible (-450 MHz on the core and -502 MHz on the RAM), which didn't really do much. After a few minutes, the cards would still drop down to much lower GPU clocks (around 900-950 MHz), though mining performance was still decent (95MH/s). Decreasing the power limit to just 60 percent finally allowed the VRAM temperatures to drop down to 90C, but then mining performance also dropped to around 60-65MH/s.

These high junction temperatures do give us some cause for concern, but it's not clear how they'll affect the cards long term. For gaming, with the GDDR6X running at 100C or lower, that appears to be within tolerance from Nvidia's perspective. 24/7 mining on the other hand with memory running at 100-110C seems like the sort of thing that could cause premature card failure.

We don't have precise answers as to why Nvidia is allowing the GDDR6X chips to hit these temperatures. We also don't know exactly what "GDDR6X Memory Junction Temperature" means, as far as the other GDDR6X chips are concerned. Presumably it's the hottest part of any of the GDDR6X chips, measured internally. And again, Nvidia's cards don't appear to start throttling GPU clocks until the temperature hits 110C.

If you find your GDDR6X temperatures run super hot with HWInfo64, it's a good idea to keep your graphics card at stock settings. And if you're mining, either decrease your power limit or hope that these 100-110C memory temperatures don't pose a problem. Don't say we didn't warn you.

HWInfo64 Release Changes:

  • Improved support of some future AMD CPUs and APUs.
  • Fixed a possible hang on some systems with Intel Thunderbolt controller.
  • Enhanced sensor monitoring on ASUS ROG STRIX B550-XE.
  • Added reporting of DIMM module location if BIOS provides correct data.
  • Fixed VRM monitoring on ASRock Z490 Taichi and B550 Steel Legend.
  • Added monitoring of Effective GPU clock on NVIDIA GPUs.
  • Improved reporting of serial number on NVIDIA GPUs.
  • Enhanced sensor monitoring on MSI Z590 and B560 series.
  • Enhanced sensor monitoring on GIGABYTE Z590, B560, H570, Q570 and H510 series.
  • Added monitoring of Aquacomputer farbwerk 360 and highflow NEXT.
  • Added fan speed monitoring on some MSI notebooks.
  • Enhanced sensor monitoring on some ASRock Z590, H570 and B560 series.
  • Added monitoring of GDDR6X Memory Junction Temperature on NVIDIA RTX 30-series.
  • Phaaze88
    If this is really the case with gpus being mined on, this could present a problem to miners selling off their old stock.
    Reply
  • Krotow
    Fine. Don't buy used GPUs from miners anymore.
    Reply
  • Clarence_Darrow
    "We also don't know exactly what "GDDR6X Memory Junction Temperature" means "

    Well, article writer might want do some research on that one.

    And any miner worth their salt keeps temps MUCH lower than this... Ive got some 3090 cards mining, and if your memory temps are anywhere near that high you are doing it wrong. Can't mine on stock settings and expect to get the best results. Can easily hit 120 hashrate on 3090 without compromising temps.
    Reply
  • JarredWaltonGPU
    Clarence_Darrow said:
    "We also don't know exactly what "GDDR6X Memory Junction Temperature" means "

    Well, article writer might want do some research on that one.

    And any miner worth their salt keeps temps MUCH lower than this... Ive got some 3090 cards mining, and if your memory temps are anywhere near that high you are doing it wrong. Can't mine on stock settings and expect to get the best results. Can easily hit 120 hashrate on 3090 without compromising temps.
    I'm gonna call bunk on this one. I did some testing, playing with clocks, fan speeds, and power limits. Unless you're modding the cards to improve cooling, or voltage modding, I'm not seeing how you get significantly higher performance than what I measured (100 MH/s on 3090) while keeping GDDR6X temperatures below 100C, never mind 110C. What's even better is the claim that "any miner worth their salt keeps temps MUCH lower than this." Really? And how do you know what temperature your GDDR6X memory is running at? Since the other GPU monitoring utilities don't report GDDR6X thermals AFAIK. MSI Afterburner doesn't report GDDR6X temps, neither does EVGA Precision X1. So what utility were you using, prior to the latest HWiNFO64, that supported GDDR6X temp reporting? For the record, the GPU was running at around 60C with the results I measured, so GDDR6X at 110C, GPU at 60C.

    As for the first part, put the whole quote in: 'We also don't know exactly what "GDDR6X Memory Junction Temperature" means, as far as the other GDDR6X chips are concerned.' We're saying we don't know if what HWiNFO64 says actually equates to what Nvidia and Micron would say, or what temperatures the other nine (3080) or 23 (3090) chips are running at. I've been around long enough to know that software can and often does differ in how sensors are interpreted. You would think that if Micron says the GDDR6X chips are rated for up to 95C TjMax, that the cards would throttle at 95C and not 110C. So perhaps HWiNFO64 doesn't know about an offset from the sensors, or there's something else at play. (Just one example: Ryzen 7 1700 as an example had a 15C offset that it reported to the fan connector, so the mobo would think the CPU was at 85C when in fact it was at 70C.)

    What I can say, unequivocally, is that at stock the exteriors of all the RTX 3080 and 3090 cards I've tested can get very hot to the touch -- 70C or so. It physically hurts if you put your hand on a card that's running at that temperature. Hold it there long enough and it would burn you. The backs of the cards are often much hotter than the fronts where the fans are. Aiming a large fan at the cards would help, and if I were planning on doing 24/7 mining with any of these cards, I'd plan on some serious external airflow at the very least.

    If I'm incorrect, I'd love to know how. Please prove me wrong by posting a video showing your mining software running at 120MH/s with GDDR6X temperatures "MUCH lower than this" and let us know the secret. Until then, I'll just toss this out there. RTX 3090 FE, sitting next to an open window (outside is maybe 40-45F right now), minimum clocks set on GPU and RAM, fan speed at 65 percent. The GPU core is running at around 1350MHz and 53C. The GDDR6X is running at 18Gbps and 108C "GPU Memory Junction Temperature." I've had this running for an hour or so doing testing, and I'm done with that now. I'll take that huge $0.77 in earnings (which is actually me testing mining and GDDR6X temperatures on multiple PCs with different GPUs over the past few hours), and call it a day. Because running consumer focused gaming GPUs like this for months at a time doesn't seem like a safe bet. "Look, I made $500 off of Ethereum mining over a two month period ... and then burned out my $1500 GPU that you can't even find in stock!" Thanks but no thanks.
    79
    Reply
  • Nero1024
    Hopefully, they will burn them down and then nobody would buy them used
    Reply
  • jpe1701
    JarredWaltonGPU said:
    I'm gonna call bunk on this one. I did some testing, playing with clocks, fan speeds, and power limits. Unless you're modding the cards to improve cooling, or voltage modding, I'm not seeing how you get significantly higher performance than what I measured (100 MH/s on 3090) while keeping GDDR6X temperatures below 100C, never mind 110C. What's even better is the claim that "any miner worth their salt keeps temps MUCH lower than this." Really? And how do you know what temperature your GDDR6X memory is running at? Since the other GPU monitoring utilities don't report GDDR6X thermals AFAIK. MSI Afterburner doesn't report GDDR6X temps, neither does EVGA Precision X1. So what utility were you using, prior to the latest HWiNFO64, that supported GDDR6X temp reporting? For the record, the GPU was running at around 60C with the results I measured, so GDDR6X at 110C, GPU at 60C.

    As for the first part, put the whole quote in: 'We also don't know exactly what "GDDR6X Memory Junction Temperature" means, as far as the other GDDR6X chips are concerned.' We're saying we don't know if what HWiNFO64 says actually equates to what Nvidia and Micron would say, or what temperatures the other nine (3080) or 23 (3090) chips are running at. I've been around long enough to know that software can and often does differ in how sensors are interpreted. You would think that if Micron says the GDDR6X chips are rated for up to 95C TjMax, that the cards would throttle at 95C and not 110C. So perhaps HWiNFO64 doesn't know about an offset from the sensors, or there's something else at play. (Just one example: Ryzen 7 1700 as an example had a 15C offset that it reported to the fan connector, so the mobo would think the CPU was at 85C when in fact it was at 70C.)

    What I can say, unequivocally, is that at stock the exteriors of all the RTX 3080 and 3090 cards I've tested can get very hot to the touch -- 70C or so. It physically hurts if you put your hand on a card that's running at that temperature. Hold it there long enough and it would burn you. The backs of the cards are often much hotter than the fronts where the fans are. Aiming a large fan at the cards would help, and if I were planning on doing 24/7 mining with any of these cards, I'd plan on some serious external airflow at the very least.

    If I'm incorrect, I'd love to know how. Please prove me wrong by posting a video showing your mining software running at 120MH/s with GDDR6X temperatures "MUCH lower than this" and let us know the secret. Until then, I'll just toss this out there. RTX 3090 FE, sitting next to an open window (outside is maybe 40-45F right now), minimum clocks set on GPU and RAM, fan speed at 65 percent. The GPU core is running at around 1350MHz and 53C. The GDDR6X is running at 18Gbps and 108C "GPU Memory Junction Temperature." I've had this running for an hour or so doing testing, and I'm done with that now. I'll take that huge $0.77 in earnings (which is actually me testing mining and GDDR6X temperatures on multiple PCs with different GPUs over the past few hours), and call it a day. Because running consumer focused gaming GPUs like this for months at a time doesn't seem like a safe bet. "Look, I made $500 off of Ethereum mining over a two month period ... and then burned out my $1500 GPU that you can't even find in stock!" Thanks but no thanks.
    79
    Jarred it is good to see a Toms hardware writer/editor in the comments. Thank you for being a member of the community. If it is anything like gddr6 the junction temp is the internal temp and actual surface temp is around 14c lower, at least that is how it is for the gddr6 on the 5700xt. I was unfortunate enough to have bought the asus tuf rx 5700xt first revision that had a tiny piece of metal for the memory heat sink and it wasn't attached to the main heat sink and I had to send it back to asus twice before they fixed it so that the memory wouldn't shoot up to 105c after only a few minutes. Now it peaks at 94 according to hwinfo64.
    Reply
  • Rrussell263
    JarredWaltonGPU said:
    I'm gonna call bunk on this one. I did some testing, playing with clocks, fan speeds, and power limits. Unless you're modding the cards to improve cooling, or voltage modding, I'm not seeing how you get significantly higher performance than what I measured (100 MH/s on 3090) while keeping GDDR6X temperatures below 100C, never mind 110C. What's even better is the claim that "any miner worth their salt keeps temps MUCH lower than this." Really? And how do you know what temperature your GDDR6X memory is running at? Since the other GPU monitoring utilities don't report GDDR6X thermals AFAIK. MSI Afterburner doesn't report GDDR6X temps, neither does EVGA Precision X1. So what utility were you using, prior to the latest HWiNFO64, that supported GDDR6X temp reporting? For the record, the GPU was running at around 60C with the results I measured, so GDDR6X at 110C, GPU at 60C.

    As for the first part, put the whole quote in: 'We also don't know exactly what "GDDR6X Memory Junction Temperature" means, as far as the other GDDR6X chips are concerned.' We're saying we don't know if what HWiNFO64 says actually equates to what Nvidia and Micron would say, or what temperatures the other nine (3080) or 23 (3090) chips are running at. I've been around long enough to know that software can and often does differ in how sensors are interpreted. You would think that if Micron says the GDDR6X chips are rated for up to 95C TjMax, that the cards would throttle at 95C and not 110C. So perhaps HWiNFO64 doesn't know about an offset from the sensors, or there's something else at play. (Just one example: Ryzen 7 1700 as an example had a 15C offset that it reported to the fan connector, so the mobo would think the CPU was at 85C when in fact it was at 70C.)

    What I can say, unequivocally, is that at stock the exteriors of all the RTX 3080 and 3090 cards I've tested can get very hot to the touch -- 70C or so. It physically hurts if you put your hand on a card that's running at that temperature. Hold it there long enough and it would burn you. The backs of the cards are often much hotter than the fronts where the fans are. Aiming a large fan at the cards would help, and if I were planning on doing 24/7 mining with any of these cards, I'd plan on some serious external airflow at the very least.

    If I'm incorrect, I'd love to know how. Please prove me wrong by posting a video showing your mining software running at 120MH/s with GDDR6X temperatures "MUCH lower than this" and let us know the secret. Until then, I'll just toss this out there. RTX 3090 FE, sitting next to an open window (outside is maybe 40-45F right now), minimum clocks set on GPU and RAM, fan speed at 65 percent. The GPU core is running at around 1350MHz and 53C. The GDDR6X is running at 18Gbps and 108C "GPU Memory Junction Temperature." I've had this running for an hour or so doing testing, and I'm done with that now. I'll take that huge $0.77 in earnings (which is actually me testing mining and GDDR6X temperatures on multiple PCs with different GPUs over the past few hours), and call it a day. Because running consumer focused gaming GPUs like this for months at a time doesn't seem like a safe bet. "Look, I made $500 off of Ethereum mining over a two month period ... and then burned out my $1500 GPU that you can't even find in stock!" Thanks but no thanks.
    79
    I registered to add this comment.

    NVIDIA, of the two main GPU designers, is historically less utilized than AMD for mining in general. They likely simply overlooked the capability of these cards to get to high GDDR6X temps in mining due to the fact that they simply don't want their mainline cards used for that purpose. Utilizing the cards memory for high-intensity mining on the memory will raise memory temps past what you would see in most cases, even in server applications.

    The little line in the article assumes NVIDIA simply lets the temperature of the memory be pushed to that point without hesitation. What is more likely is that the memory is stressed beyond what you would see in any case of usual usage, to the point where throttling the GPU does nothing to mitigate it. If utilizing these cards for mining, something NVIDIA has not announced any support for outside the Crypto SKUs, then it is an unsupported application of the hardware.

    Beyond that, the VBIOS is usually prepped by the developer then sent to the board partners for modifications, or made from the ground up by those same partners. The issue would fall on them, then, for not optimizing their cards for mining, not NVIDIA.

    If you use these cards for unsupported applications, it is abundantly clear that these applications might lead to thermal runaway or otherwise. What is most likely here is that the memory pads are insufficient for mining use, so need to be modified for their unsupported use. This is a non-issue.
    Reply
  • Phaaze88
    A880J said:
    3090 Strix here, pulling 102.9 Mh/s @ steady 61C. When not I'm pulling 66FPS on Cyberpunk.

    a/OtrHxYNView: https://imgur.com/a/OtrHxYN

    <3
    61C is gpu core. GDDR6X reading isn't present in that image.

    Still at square one?
    Reply
  • alexandre.nado
    Clarence_Darrow said:
    "We also don't know exactly what "GDDR6X Memory Junction Temperature" means "

    Well, article writer might want do some research on that one.

    And any miner worth their salt keeps temps MUCH lower than this... Ive got some 3090 cards mining, and if your memory temps are anywhere near that high you are doing it wrong. Can't mine on stock settings and expect to get the best results. Can easily hit 120 hashrate on 3090 without compromising temps.
    Gimme a break. On Reddit, NO ONE reported temps below the Memory Junction Temperature on stock cooler, no matter their mining settings.
    Reply
  • Phaaze88
    No one really noticed the Vram thermals unless:
    -they had something like EVGA's FTW3; they can be monitored via Precision X.
    -they used thermocouples.

    It'll be interesting to see/hear what happens to 2nd hand buyers of this gen of mining gpus.
    Reply