News GDDR6X in 3080 and 3090 Hits 110C While Mining Ethereum

[UnCertainty08]

Postimages.org
Imgbb.com

Thank you for the help
Damnit! When I click the pic icon and enter the url from imbb it says "Something went wrong. Please try again or contact the system administrator. I've tried numerous times. It did this before when I tried instgram I think.

 

[UnCertainty08]

I got it to work with postimages.org!
Thanks again for the help Phaaze

 

[UnCertainty08]

Here's one with PX1-

Doing anything on my computer slows the hashrate down from Max at dead idle of 123MH/s. Evening just moving the mouse drops it 3-4MH/s. Playing 4K video in PowerDVD drops to about 100MH/s. I don't game AAA titles while mining but running an older game like wolf ET drops it to 100MH/s as well.

****
I have a Corsair Obsidian 800D with ~14 total fans right now, including a couple 140x38mm 5200rpm fans. I live somewhere with horrible weather it's cool almost the entire year. Ambient temp is 19°C.

 

[yzonker]

That's seems effective ! definitly appealing ! I have the exact same card and it run at 105°C at 80% but the fan speed is only 90% (set to default so not throtling until 110 as stated). I would consider that solution, thx for the post ! what is your components ? Do you had issue with the thermal paste application ? can it be removed without risks for RMA ? I don't have more than 1.2cm space (cpu air cooler). Can i move the GPU to the 2dn slot without compromise on the X570 (bandwith important ?) ?
since i have 2 spare 120mm fan and a GPU holder it will be a nice use !

it would be nice of you if i get even a short reply for the main part. I just started mining 3 weeks ago and i don't want to ruin prematurely this 1760€ gaming card.

I used thermal tape to attach the heatsink. I run just 70% power. Fans are on auto in the 60-70% range. You probably could move the card down

 

[yzonker]

Here's one with PX1-

Doing anything on my computer slows the hashrate down from Max at dead idle of 123MH/s. Evening just moving the mouse drops it 3-4MH/s. Playing 4K video in PowerDVD drops to about 100MH/s. I don't game AAA titles while mining but running an older game like wolf ET drops it to 100MH/s as well.

****
I have a Corsair Obsidian 800D with ~14 total fans right now, including a couple 140x38mm 5200rpm fans. I live somewhere with horrible weather it's cool almost the entire year. Ambient temp is 19°C.

When I zoom in on my phone at least, the image isn't readable.

 

[UnCertainty08]

When I zoom in on my phone at least, the image isn't readable.

Sorry, don't know what to say when I check it on my computer I can zoom in quite a bit and it's still legible. I did originally resize the image to make the file size smaller because I assumed there would be limit for uploading but was able to upload original screen shot saved as jpg through MS Paint.

 

[JarredWaltonGPU]

Here's one with PX1-

Doing anything on my computer slows the hashrate down from Max at dead idle of 123MH/s. Evening just moving the mouse drops it 3-4MH/s. Playing 4K video in PowerDVD drops to about 100MH/s. I don't game AAA titles while mining but running an older game like wolf ET drops it to 100MH/s as well.

****
I have a Corsair Obsidian 800D with ~14 total fans right now, including a couple 140x38mm 5200rpm fans. I live somewhere with horrible weather it's cool almost the entire year. Ambient temp is 19°C.

I suppose I should clarify that I was speaking to the cards I have tested (Asus 3090 TUF, 3090 FE, 3080 FE, MSI 3080 Gaming X) regarding the GDDR6X temps. It looks like EVGA, at least on the FTW3, does a much better job at cooling the VRAM. I'm sure some of the other AIB cards do better, but many don't. The 3080 and 3090 FE, as well as a couple other 3080/3090 cards I looked at, are not very happy with mining, at all. I'm also not getting anywhere near 120MH/s on 3090, or 100MH/s on 3080 (I get at best ~100 and ~90MH/s, depending on settings), but then I was running NiceHashMiner (because it's super easy to configure, but likely very much not optimal).

The main point is to actually check your VRAM temps if you're mining. We still don't know how 110C will affect GDDR6X long term, and Nvidia doesn't appear to throttle until that point, but it's quite high. 24/7 at those temperatures definitely won't improve card longevity. I suspect shims or some other mod to make the GDDR6X have better contact with the heatsinks would do a lot to help GDDR6X temps.

 

[JarredWaltonGPU]

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.

This isn't a good argument in favor of Nvidia.

I just checked FurMark, letting it run for 30 minutes or so on the 3090 FE, at stock. GDDR6X hit 104C max before settling down to 102C. The card and drivers are fully capable of monitoring these things, so if the memory hits 110C, it's because the card isn't throttling soon enough. FurMark is probably specifically detected, or maybe it's just lighter than mining. Either way, ETH mining at stock settings causes the fan speed to ramp up to 100% after a bit. That's terrible! If the GPU core is at 60C and the memory is at 110C, and the card knows the memory is overheating, it should clock down before that point. Unless Nvidia only considers 110C and above the throttle point, which appears to be the case.

Having checked one more GPU (Colorful 3080 Vulcan), memory temps are much lower (96C max) while mining. I suspect the card just has better cooling contacts on the memory, so keeping the GDDR6X cool can be done. But the first wave of cards, including the Founders Edition, apparently didn't put enough effort into memory cooling. Thermal pads were okay at 14Gbps GDDR6. 19/19.5Gbps GDDR6X? Probably need to go to thermal paste and direct heatsink contact without the millimeter thick pads.

Right now, what I've seen (with a very limited number of cards) is that the 3080/3090 Founders Edition cards are probably the worst of the bunch as far as memory cooling goes. Several other 'typical' card designs also have high memory temps. The fix -- which should arguably be done in drivers, because VBIOS updates are extremely rare -- is pretty simple. Something like:

1 Second Timer:
if (GDDR6XTemp > 100C): DropClocks5Percent
Repeat.

Obviously it would need to track current clocks and previous clocks to see if temps are trending down, but the theory is that the GPU would limit clocks earlier. In fact, I'm sure the cards already do something like this, except it's at 110C right now. So change that temperature to 100C -- or issue a statement that 110C is "perfectly fine".

 

[UnCertainty08]

Having checked one more GPU (Colorful 3080 Vulcan), memory temps are much lower (96C max) while mining. I suspect the card just has better cooling contacts on the memory, so keeping the GDDR6X cool can be done.

My 3090 mem is 82°C, it's been mining the last ~3 days non-stop 24/7. EVGA can cool the memory. That temp with the +129 core +1298 mem

 

[JarredWaltonGPU]

My 3090 mem is 82°C, it's been mining the last ~3 days non-stop 24/7. EVGA can cool the memory. That temp with the +129 core +1298 mem

What mining software are you running? I'm wondering if maybe the NiceHashMiner software is just less ... efficient. But it has multiple miner options for ETH, so it's probably just the difference in cards. EVGA must have put some beefy cooling on the GDDR6X.

 

[UnCertainty08]

I have Phoenix Miner 5.5c currently, I just started using this several days ago and it's significantly better than T-Rex Miner I was using. The mining rate is a very small amount higher with Phoenix, it's much more stable and consistent, it doesn't slow down as much when I 'm doing anything else and the GPU core temp is 4-5°C lower(can't comment on diff of any mem temp, I didn't take note).
EVGA absolutely put a very beefy cooler on the FTW3 Ultra. Also as I stated in my original post I've got a Corsair Obsidian 800D, with 12+ fans running, There's 6 Noctua(3x120-3x140). 2 x 140x38mm 5300RPM and a few more 120&140. The PC is in the basement, where I live has cool/moderate climate, ambient temp is 19°C.
Thanks for the article Jarred, I didn't know about the new feature in HWinfo and would not have learned about without your article. Just about a week ago I added and changed around some fans to get more air moving across the back of the card, the mem temp readings from ICX in PX-1 where >60°C and I couldn't keep my hand on the back of the card for more than several seconds. Now the same readings from ICX are <60°C and I can keep my hand on the back of the card.

 

[Kravon]

I have three 3080 and three 3060 Ti GPU's that mine 24/7. I did notice throttling on the 3080 and temps were reaching about 110c on some of the VRM's. I replaced the thermal pads with 17.0 W/mK pads and it brought the temps down to 80c. The thermal pads were not cheap - like $30 per GPU - but well worth it to me. I did try some cheaper thermal pads but they did not work near as well.

 

[drivinfast247]

I declare a minute of silence for the poor souls who in the coming months will buy those cards used from miners at a price higher than MSRP to see them artifact to death within a week.

I'll pay $200 for a used mining 3080.

 

[UnCertainty08]

I have three 3080 and three 3060 Ti GPU's that mine 24/7. I did notice throttling on the 3080 and temps were reaching about 110c on some of the VRM's. I replaced the thermal pads with 17.0 W/mK pads and it brought the temps down to 80c. The thermal pads were not cheap - like $30 per GPU - but well worth it to me. I did try some cheaper thermal pads but they did not work near as well.

Sweet, I wish I could get my hands more 3000 series cards. I'm still on the waiting list for 3080, I signed up one week after launch. The demand has just grown since they were released. Everyone knows the mining performance, so everyone and their dog wants a 3000 series card now.
I used to hate on miners like others but not anymore. I wish I had mined back when I first learned of Bitcoin. I had a 2 x EVGA 580 SLI, installed and had working Bitcoin miner when BitCoin was $12. I didn't know much about it, shut it down after a couple hours and forgot about bitcoin for a long time. Hindsight's a bitch sometimes right?

Would you mind please sharing the thermal pads you used and where you purchased them? I live in Canada which makes it difficult and expensive to get anything. I've been replacing the thermal pads on all my motherboards and most video cards for a number of years now. It's more difficult with video cards to find the correct specs and corresponding high performance thermal pads. There are some 12W/mK thermal pads I've been using for a while now that have made surprisingly large improvements for temps on motherboards and video cards. Even high ASUS motherboards sometimes have garbage thermal pad placement and performance. I currently have an ASUS Maximus Apex and I've had 2 Maximus boards prior to this one including a Maximus Extreme-Z. The pads I order only come in 0.5, 1.0, 1.5mm sizes which is usually fine for Motherboards but Video Cards are a whole different ball game sometimes. Like my current EVGA 3090 FTW3 Ultra, it uses several different sizes including 3.0, 2.2mm and others. I have not been able to find ones I can order that size with >10W/mK performance for a reasonable price. Sometimes 1 pad of particular thickness might cost something crazy like $191USD. By the time I add shipping, conversion, import charges and duty I'm honestly looking at close to $300. Yes that 1 pad would last me for at least a couple video cards but it's still a ridiculous price for a thermal pad

 

[Kravon]

Sweet, I wish I could get my hands more 3000 series cards. I'm still on the waiting list for 3080, I signed up one week after launch. The demand has just grown since they were released. Everyone knows the mining performance, so everyone and their dog wants a 3000 series card now.
I used to hate on miners like others but not anymore. I wish I had mined back when I first learned of Bitcoin. I had a 2 x EVGA 580 SLI, installed and had working Bitcoin miner when BitCoin was $12. I didn't know much about it, shut it down after a couple hours and forgot about bitcoin for a long time. Hindsight's a bitch sometimes right?

Would you mind please sharing the thermal pads you used and where you purchased them? I live in Canada which makes it difficult and expensive to get anything. I've been replacing the thermal pads on all my motherboards and most video cards for a number of years now. It's more difficult with video cards to find the correct specs and corresponding high performance thermal pads. There are some 12W/mK thermal pads I've been using for a while now that have made surprisingly large improvements for temps on motherboards and video cards. Even high ASUS motherboards sometimes have garbage thermal pad placement and performance. I currently have an ASUS Maximus Apex and I've had 2 Maximus boards prior to this one including a Maximus Extreme-Z. The pads I order only come in 0.5, 1.0, 1.5mm sizes which is usually fine for Motherboards but Video Cards are a whole different ball game sometimes. Like my current EVGA 3090 FTW3 Ultra, it uses several different sizes including 3.0, 2.2mm and others. I have not been able to find ones I can order that size with >10W/mK performance for a reasonable price. Sometimes 1 pad of particular thickness might cost something crazy like $191USD. By the time I add shipping, conversion, import charges and duty I'm honestly looking at close to $300. Yes that 1 pad would last me for at least a couple video cards but it's still a ridiculous price for a thermal pad

Sure. They are available from a major site that is named after a rain forest.


Fujipoly/mod/smart Ultra Extreme XR-m Thermal Pad - 60 x 50 x 1.5 - Thermal Conductivity 17.0 W/mK

Also, I use two of the cards for occasional gaming as well. One of the 3080 rigs I have attached to my Oculus Quest 2 via the link cable and another rig I have attached to a 165hz 2k IPS monitor that I use for games like Call of Duty Cold War and CyberPunk 2077.

 

[UnCertainty08]

Sure. They are available from a major site that is named after a rain forest.


Fujipoly/mod/smart Ultra Extreme XR-m Thermal Pad - 60 x 50 x 1.5 - Thermal Conductivity 17.0 W/mK

Also, I use two of the cards for occasional gaming as well. One of the 3080 rigs I have attached to my Oculus Quest 2 via the link cable and another rig I have attached to a 165hz 2k IPS monitor that I use for games like Call of Duty Cold War and CyberPunk 2077.

Thanks for the info! You were able to do all your cards with nothing thicker than 1.5mm? You're lucky! LOL
I'm really loving the new COD as well. My 3090 got delivered the same day as COD Cold War was released-it was a pretty awesome day. I did fresh wipe of my boot drive with a clean install of Win 10 Pro for the 3090 install as well.
I tried CyberPunk at launch but played for 20 mins and got a refund from steam. I've since reinstalled it but haven't been able to get into it. I just keep going to back to Cold War!

I've been waiting for VR headsets to get to a certain quality to buy one. I'm really looking forward to it and they are almost to the quality I think will be time for me take the plunge.
How do you find the Quest 2? That's one that doesn't require a PC so I didn't even really consider it. Do you get better performance from it by using a 3080 to drive it?



**
LOL I looked up those pads on the canadian version of the site you mentioned and it's $85+tax and shipping for 1 pad. This is one of the situations when it's cheaper for me to buy from the us and have it shipped to canada.
BTW canada sucks for trying to get high end PC Hardware even in the best of times, the last year with cv it's been total gong show.
The ones I usually buy are ThermalRight 12.8W/mK and cost me less than $20 to my door.

 

[yzonker]

One important thing I've noticed also. Using a custom fan curve in Afterburner will cause the high VRAM temps to be ignored. I couldn't understand what the issue was the first time I closed Afterburner while mining and my fans instantly jumped to 100%.

 

[yzonker]

Sorry, don't know what to say when I check it on my computer I can zoom in quite a bit and it's still legible. I did originally resize the image to make the file size smaller because I assumed there would be limit for uploading but was able to upload original screen shot saved as jpg through MS Paint.

Yea they work on my PC. Phone was down res'ing them I guess. Interesting though. I've seen quite a few people with EVGA cards and high temps. I even saw a couple of Kingpin guys putting heatsink on the backplate that were mining. Would be good to understand what they did better on the FTW3.

 

[yzonker]

Here's one with PX1-

Doing anything on my computer slows the hashrate down from Max at dead idle of 123MH/s. Evening just moving the mouse drops it 3-4MH/s. Playing 4K video in PowerDVD drops to about 100MH/s. I don't game AAA titles while mining but running an older game like wolf ET drops it to 100MH/s as well.

****
I have a Corsair Obsidian 800D with ~14 total fans right now, including a couple 140x38mm 5200rpm fans. I live somewhere with horrible weather it's cool almost the entire year. Ambient temp is 19°C.

BTW, a couple of other comments.

It does appear you are running full fan or nearly so (looks like 3000rpm, that's close to 100% on my Zotac 3090 anyway).

Also just as a side note for mining, maybe you've optimized it already, but I'm pretty sure you can get that hashrate or very close at 300 watts or even a little less. I can get 110-115 @300w.

 

[ern88]

How do AIB cards stack up?

 

[Maxim S]

https://media-www.micron.com/-/medi...rief.pdf?rev=b65e2075b1f2437ab5fa0d4c06e6aa6f

-> To be short on this, micron never defines a junction temperature in their datasheet, only a case temperature(Tc, this is also the case for ddr3,ddr4). If you stay within this limit it's ok (ofcourse the cooler it is the better). juction to case and power are also missing from the pdf that is available for everyone so is the thing you are writing an article about even a problem ??

https://media-www.micron.com/-/medi...apps.pdf?rev=71bbcae453c14b29a736fbb246e1aedb

 

[alexzogh]

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.
View attachment 79

You really need to do a little research in the mining forums to understand how to optimize the power and thermals for maximum output.

attached below is a screenshot of my gaming rig. Jacked up memory +1000, and clocked down the GPU by -250. Also reduced target power consumption down to 77%. This is with an EVGA FTW3 Ultra. Now, the heatsink isn't very hot because the GPU is clocked down, so the airflow of the fan is going right up against the memory. GPU typically sits at 50c or below, and the highest memory temp is typically 64c. You can see all of the memory temps in X1. With this setup, I'll typically get between 112-125 M hashrate. It's variable because I'm typically using the computer.

 

[PapaCrazy]

Between shortages (intended and not), individual scalper scums and company scalpers (AIBs, which are the same) and the gamer fools that buy at those insane prices, and the abhorrent miners, this whole continuing situation makes me vomit. Bleah.

Same here. Decided to put off a new build and ride things out with a 2600k and GTX1080. My PC still puts a smile on my face. And a new build right now will cause me nothing but misery.

 

[Jim90]

Oh those poor miners!
Since the vast, vast majority of comments on intertech forums - relating to gfx cards - are from humble folks simply looking to be able to purchase, wouldn't it be refreshing for all associated tech press to jointly place a ban on showing anything related to this longstanding gfx vapourware - irrespective of reasons.
Clearly, there's a significant volume of forum members someway passed that p#ssed off mark on 'new news' relating to these cards - and their extortionate 3rd party pricing. Something to bear in mind if any value is placed on readerships!

 

[Deleted member 2838871]

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.

Quoting for truth such an epic comment.