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

[Rrussell263]

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".

Except that furmark is intentionally overutilzing the card to push heat to maximum levels across the board.

Nvidia, 10 years ago posted this:

https://www.nvidia.com/en-us/geforce/forums/game-ready-drivers/13/115902/using-furmark-and-other-stress-tests-with-geforce-/

"Furmark is an application designed to stress the GPU by maximizing power draw well beyond any real world application or game."

So a result of 104c in Furmark, of all things, is certainly a non-issue. In the GPU market for laptops the same is true- it can kill boards fairly easily despite its nature as a stress test. I would not recommend anyone use that software even for a stress test. There are far more realistic and better utilizing apps out there for one, for two, there is a good amount of warning about that from the manufacturer that the use of such an app would void the warranty- this again is cut down to another use of the card outside its intended purpose.

I've been on both sides of the GPU world in relation to consumers and professional use, for over 10 years.

These numbers would be an issue if it were simply a game, but the tests you've put forth are highly suspect because of their usage of GPU resources to literally run the card at intensities it wasn't designed for. If that is the case, it is not a failure on the part of Nvidia for that reason, but the person putting the card on a load that isn't in line for the use of the card.

It is essentially no different than someone putting their card in the oven, and expecting a refund when they burn the board.

To clarify I am not saying that this is not an issue, but I am saying it is outside the expected use for the card. This means that the burden of improving cooling is on the consumer, not Nvidia, as the application of the card in Furmark/ethereum mining is outside the normal usage of the card on an everyday basis, and as such is not guaranteed to function within limits during that use of the application

 

[yzonker]

Except that furmark is intentionally overutilzing the card to push heat to maximum levels across the board.

Nvidia, 10 years ago posted this:

https://www.nvidia.com/en-us/geforce/forums/game-ready-drivers/13/115902/using-furmark-and-other-stress-tests-with-geforce-/

"Furmark is an application designed to stress the GPU by maximizing power draw well beyond any real world application or game."

So a result of 104c in Furmark, of all things, is certainly a non-issue. In the GPU market for laptops the same is true- it can kill boards fairly easily despite its nature as a stress test. I would not recommend anyone use that software even for a stress test. There are far more realistic and better utilizing apps out there for one, for two, there is a good amount of warning about that from the manufacturer that the use of such an app would void the warranty- this again is cut down to another use of the card outside its intended purpose.

I've been on both sides of the GPU world in relation to consumers and professional use, for over 10 years.

These numbers would be an issue if it were simply a game, but the tests you've put forth are highly suspect because of their usage of GPU resources to literally run the card at intensities it wasn't designed for. If that is the case, it is not a failure on the part of Nvidia for that reason, but the person putting the card on a load that isn't in line for the use of the card.

It is essentially no different than someone putting their card in the oven, and expecting a refund when they burn the board.

To clarify I am not saying that this is not an issue, but I am saying it is outside the expected use for the card. This means that the burden of improving cooling is on the consumer, not Nvidia, as the application of the card in Furmark/ethereum mining is outside the normal usage of the card on an everyday basis, and as such is not guaranteed to function within limits during that use of the application

Except that I can push my VRAM temp just as high with Cyberpunk 2077 as Furmark. I even see higher power draw from Cyberpunk than Furmark.

 

[VTXcnME]

I can only hope the absolute worst for anyone mining with any of these cards.

Dumb comment. Like anyone is going to tell you they've been mining on it!

I don't buy used tech components for the same reason I won't buy used Motorcycles or Pick up trucks....

You don't have any idea how badly the previous owner beat the ever loving crap out of them over their lifetime.


Some folks have luck with it, I don't trust my money to used and (usually) abused items.

 

[JarredWaltonGPU]

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.

It has very little to do with not having done the research, and more to do with knowing that there's a big difference between the cards. Even looking at multiple 3080 and 3090 cards, similar settings on each don't produce the same hash rate, and the memory temps can vary a LOT. EVGA FTW3 is, literally, the best card I'm aware of with regards to cooling the GDDR6X. It's also theoretically more expensive, and due to miner demand I imagine that's only going to get worse. Show me an unmodified 3090 Founders Edition, doing 120MH/s with the memory at less that 100C, and fans not running at 100%. I've done a lot of poking at this in the past few days, and the best hash rate I've seen with the memory not at 110C is ~100MH/s. Power limit at 70%, GPU clocks -450 (the maximum reduction), GDDR6X at -502 (the maximum again). That gives me 102C on the memory with 65% fan speed.

 

[Rrussell263]

Except that I can push my VRAM temp just as high with Cyberpunk 2077 as Furmark. I even see higher power draw from Cyberpunk than Furmark.

But at the load levels for Cyberpunk 2077, are you still seeing unacceptable temps for the VRAM? Furmark Still has a unrealistic draw and just because power draw doesn't creep to CP2077's levels, does not mean that the load is less intensive. There are several instruction sets for GPUs, just as there are several instruction sets for CPUs, like AVX or AVX-2 or AVX-512, each stresses the CPU in a different way, and power draw can be considerably higher when dealing with AVX vs AVX-2.

The point here is that personal experience =/= objective fact.

Besides, having higher sensor power draw doesn't mean anything, have you tested at the wall?

A basic shunt mod tricks the power sensing circuit on the GPU to think it's drawing less power than it is, and sensors are far from perfect on any board.

 

[UnCertainty08]

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.

Sorry I didn't see this message at first, I looked at one immediately after this one, didn't notice this one until later.

I understood you were speaking to the cards you have tested, as I said originally I was not trying to brag or argue just posting my results to add to the data pool. I appreciate the article you Wrote Jarrod and the work overall you and your co-workers at Tom's Hardware do everyday. I've been a regular to the Tomshardware website since the very beginning. I am extremely thankful for the information you provided this article. I had no idea how high the actual VRAM temp was. I was very suspicious of the readings from ICX because I could not hold my hand on the back of the video card for very long at all. The GPU was reading <50°C and the highest mem temp was ~60°C. So from those numbers I would have expected the back of the card not to be as hot as it was. I was even considering buying one of those laser thermometers. I tried to measure the surface of the card with a high end "instant" chef's thermometer but it did not work like I wanted.

There is always going to relatively large differences in cards. Main contributing factor to the added cost of some of the AIB cards. Like the EVGA 3090 FTW3 Ultra I bought. Also not every single card has perfect placement of thermal pads or best possible contact with heatsinks/heat pipes.


For the record I am purely a gamer. I've been a PC Gaming Enthusiast since 1991 when I got my first PC, and I started building my own high end Gaming PC's in 1995.
This is the first time I have ever mined crypto. After seeing what I could profit per day I made the decision to attempt to recoup some of the cost of this card. I live in canada so the everything included price to my door was almost $3000CDN.

 

[UnCertainty08]

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.

Yes I am running fans at 100%. I spent all my teenage years with a PC in my bedroom running 24/7 so I'm immune to fan noise.

This is my first time mining, so I appreciate any advice you have. I will try to optimize and get the power usage down a little. At first I didn't really know I was doing and I tried a couple different mining software until I got one working.
Once I learned a little more I got my current mining software working, Phoenix Miner, it was worse before.




****
I lowered my OC to -320 core and 70% power my hashrate dropped from ~122MH/s to ~117MH/s while my power dropped from 343W to 296W. My GPU temp dropped to 40°C from 47-48°C, HWinfo reports Mem TJ as 76°C down from 82-84°C.
That's definitely better without losing much performance. I left the Mem OC as +1298.

 

[Deleted member 2838871]

For the record I am purely a gamer. I've been a PC Gaming Enthusiast since 1991 when I got my first PC, and I started building my own high end Gaming PC's in 1995.
This is the first time I have ever mined crypto. After seeing what I could profit per day I made the decision to attempt to recoup some of the cost of this card. I live in canada so the everything included price to my door was almost $3000CDN.

... and how much is that per day? Serious question.

(Gaming enthusiast since 1988 when my Dad got me a Commodore Amiga 500)

 

[UnCertainty08]

... and how much is that per day? Serious question.

(Gaming enthusiast since 1988 when my Dad got me a Commodore Amiga 500)

The price of Ethereum fluctuates often so it's $9+ USD/day. I live in canada so that's like $100 canadian, lol, it's actually about $13-14CDN and it cost me about $1/day in electricity.
Especially since I've optimized my settings and have the efficiency much higher now, that is worth to me. I could have my computer do nothing all day when I'm study and doing generic windows tasks or I could have it mine in the background all day and all night. Pretty clear decision in my eyes. I have excellent warranty coverage so I'm not worried about my card getting worn out, or whatever people think mining does to video cards.

Here's a new screen shot since adjusting my settings even further, I got it down to 297 Watts @ 120.9MH/s, GPU Temp is 40°C, HWinfo VRAM TJ is still 76-78°C, ICX for what it's worth had mem readings max 50°C.
I'm down from 122-123MH/s @ 343 Watts.

 

[VforV]

I have excellent warranty coverage so I'm not worried about my card getting worn out, or whatever people think mining does to video cards.

Here's a new screen shot since adjusting my settings even further, I got it down to 297 Watts @ 120.9MH/s, GPU Temp is 40°C, HWinfo VRAM TJ is still 76-78°C, ICX for what it's worth had mem readings max 50°C.
I'm down from 122-123MH/s @ 343 Watts.

For the first statement alone I have so many issues with... you're not worried, but what about the unknowingly buyer of your used card 2-3 years from now when he buys it from you? He won't have excellent warranty...

BUT, in the context of having such a ventilated case and one of the best cooled RTX 3000 cards at these current temps (as you say), maybe in this case, there won't be any issues with this particular card.

I can't say the same about other people mining and their cards that will be sold used when the mining craze ends...
So in the end, overall, not only miners are another big issue for all the gamers that can't buy cards because of them (anyone that has bought more than 1 new GPU lately, with very few exception cases), but also the way these cards are used - that is another issue on top of the issue too. The whole situation is worse than it ever was...

 

[320flyboy320]

I have an Asus Tuf 3080 and would like to replace the thermal pads that cover the memory chips on both the front and back of the PCB with better quality ones. Does anyone know the correct thickness of these particular pads? I know Gamers Nexus did a teardown of this card and guessed at its thickness but I'm looking to see if anyone knows for sure.

 

[Deleted member 2838871]

The price of Ethereum fluctuates often so it's $9+ USD/day. I live in canada so that's like $100 canadian, lol, it's actually about $13-14CDN and it cost me about $1/day in electricity.

Cool... was just wondering. I personally don't think it's worth the wear and tear on the hardware for $300 a month. I'd compare it to taking a job delivering pizza putting 100 miles a day on my car to make what I consider to be chump change. All the wear and tear isn't worth it when I can make that same amount in a day on my job.

Someone working a minimum wage job would probably disagree.

Anyway... not looking to debate just wanted confirmation on my initial thoughts... and I was right.

 

[Neil S]

HWInfo64 adds GDDR6X temperature monitoring for Nvidia's RTX 3080 and RTX 3090 with latest version 6.42.

GDDR6X in 3080 and 3090 Hits 110C While Mining Ethereum : Read more

Fixed on my 3080 gaming x trio.
Was mining at 96 deg with large underclock. Now 70deg.
Replaced 1mm pads with 0.5mm copper and 0.5mm pad.

 

[norbert-the-great]

HWInfo64 adds GDDR6X temperature monitoring for Nvidia's RTX 3080 and RTX 3090 with latest version 6.42.

GDDR6X in 3080 and 3090 Hits 110C While Mining Ethereum : Read more

I've found the best solution to this is to apply heatsinks to the backplate and aim a fan at them. Here's how to do it:
http://www.norbert-the-great.com/index.php?entry=Solved:-RTX-3090-VRAM-Overheating

 

[haha not funny]

Im not really sure how to feel about this, currently mining on 3090. But I sort of believe nvidia that if they allow these temperatures, that they are relatively safe. And if not these cards have usually 36 months warranty so.

 

[haha not funny]

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.

basically there is single card that can run like this and it is the evga ftw3, all other cards are running at 100+C and some special ones like gigabyte aorus are thermal throttling at stock settings xD

 

[haha not funny]

For the first statement alone I have so many issues with... you're not worried, but what about the unknowingly buyer of your used card 2-3 years from now when he buys it from you? He won't have excellent warranty...

BUT, in the context of having such a ventilated case and one of the best cooled RTX 3000 cards at these current temps (as you say), maybe in this case, there won't be any issues with this particular card.

I can't say the same about other people mining and their cards that will be sold used when the mining craze ends...
So in the end, overall, not only miners are another big issue for all the gamers that can't buy cards because of them (anyone that has bought more than 1 new GPU lately, with very few exception cases), but also the way these cards are used - that is another issue on top of the issue too. The whole situation is worse than it ever was...

I have been gaming for years and just recently started mining on my gaming pc, cuz the price of crypto is kinda crazy. But what I dont understand is ppl constantly hating miners. Tell me how its possible some miners are able to buy 8x 3080 and some poor gamer is not able to get single card? Maybe they are not just putting enough effort. Before I got mine 3090 I had multiple orders on different gpus, I had notifications on all the gpus at like 5 different sites. And numerous times everyday for like 3 weeks I was manually checking all the sites till I got one.

 

[VforV]

I have been gaming for years and just recently started mining on my gaming pc, cuz the price of crypto is kinda crazy. But what I dont understand is ppl constantly hating miners. Tell me how its possible some miners are able to buy 8x 3080 and some poor gamer is not able to get single card? Maybe they are not just putting enough effort. Before I got mine 3090 I had multiple orders on different gpus, I had notifications on all the gpus at like 5 different sites. And numerous times everyday for like 3 weeks I was manually checking all the sites till I got one.

The answer is very, very simple: US is not the entire world.

Do you have any clue on the availability and prices of GPUs in Europe? By the way you're talking I already know the answer, it's no. And even between EU countries there are major differences in availability and prices, in some parts of EU there were never GPUs in stock (like RX 6000s) and when they were (nvidia 3000s, they were single digit numbers and for 2x to 3x the MSRP price, so be my guest and you buy them at that price... *(also most of the people don't use bots to buy things...)

The pandemic and people staying at home, brought a never-seen increase of demand for electronics, especially consoles, PCs and PC parts (CPUs and GPUs the most) and that alone made this whole situation horrible. But on top of that disaster we got an even bigger one added, the mining boom, again... So yeah, **** all that!

It's like we got nuked and after that an asteroid also crashed on us...

PS. You or anyone else that has his personal PC and is mining on that is not the issue, the issue are those buying in bulk GPUs to make mining rigs.

 

[JarredWaltonGPU]

The answer is very, very simple: US is not the entire world.

Do you have any clue on the availability and prices of GPUs in Europe? By the way you're talking I already know the answer, it's no. And even between EU countries there are major differences in availability and prices, in some parts of EU there were never GPUs in stock (like RX 6000s) and when they were (nvidia 3000s, they were single digit numbers and for 2x to 3x the MSRP price, so be my guest and you buy them at that price... *(also most of the people don't use bots to buy things...)

The pandemic and people staying at home, brought a never-seen increase of demand for electronics, especially consoles, PCs and PC parts (CPUs and GPUs the most) and that alone made this whole situation horrible. But on top of that disaster we got an even bigger one added, the mining boom, again... So yeah, **** all that!

It's like we got nuked and after that an asteroid also crashed on us...

PS. You or anyone else that has his personal PC and is mining on that is not the issue, the issue are those buying in bulk GPUs to make mining rigs.

The real issue (IMO) is that the miners can go to a company -- like Asus, EVGA, Gigabyte, MSI, Sapphire, and maybe even AMD and Nvidia direct -- and say, "Hey, you know those graphics cards you want to sell at $700 each? We'll pay you $1050, direct, no distributor markup. And we'll buy 5000 GPUs right now." If you're in charge of that company, do you sell the cards at retail and let Best Buy or whoever take a 15% cut, and the distributor take a 15% cut ... or do you grab the whole pile of cash and then release press statements about how tough times are and that you're working to get more inventory?

 

[haha not funny]

The answer is very, very simple: US is not the entire world.

Do you have any clue on the availability and prices of GPUs in Europe? By the way you're talking I already know the answer, it's no. And even between EU countries there are major differences in availability and prices, in some parts of EU there were never GPUs in stock (like RX 6000s) and when they were (nvidia 3000s, they were single digit numbers and for 2x to 3x the MSRP price, so be my guest and you buy them at that price... *(also most of the people don't use bots to buy things...)

The pandemic and people staying at home, brought a never-seen increase of demand for electronics, especially consoles, PCs and PC parts (CPUs and GPUs the most) and that alone made this whole situation horrible. But on top of that disaster we got an even bigger one added, the mining boom, again... So yeah, **** all that!

It's like we got nuked and after that an asteroid also crashed on us...

PS. You or anyone else that has his personal PC and is mining on that is not the issue, the issue are those buying in bulk GPUs to make mining rigs.

Okay, so first thing, Im actually from Europe (Czech republic specifically). So I have reasonably good clue about gpu pricing and availability in the EU Another thing is how do u know they never were in stock? Usually shops wont show history of availability and which cards were sold at specific dates. I have kept a good eye on the pricing and I can tell u that at first these cards were selling near the MSRP, the price is increasing gradually since (which makes sense, its bussiness afterall and they want as high profit as they can get).

I pretty much agree with the second paragraph. I just wanted to say that if u r patient, not picky (if u r not looking for single exact gpu model), and put effort to it, Im sure you gonna get one. Also what is important, dont really look at the mainstream retailers. Try to take a look at local ones or the less known.

 

[haha not funny]

The real issue (IMO) is that the miners can go to a company -- like Asus, EVGA, Gigabyte, MSI, Sapphire, and maybe even AMD and Nvidia direct -- and say, "Hey, you know those graphics cards you want to sell at $700 each? We'll pay you $1050, direct, no distributor markup. And we'll buy 5000 GPUs right now." If you're in charge of that company, do you sell the cards at retail and let Best Buy or whoever take a 15% cut, and the distributor take a 15% cut ... or do you grab the whole pile of cash and then release press statements about how tough times are and that you're working to get more inventory?

Okay I totally agree with this. What I originally meant was the guys mining at home with like 6 or 8 gpus. I have actually no clue about how these huge miners work or get gpus. Good to know.

 

[S20coeur]

Hello,

on my part i'm unable to reach such temperature on my RTX 3090 FTW3 ULTRA Gaming (according UnCertainty08 temperatures)

I use phoenixminer too and I run on average at 117MH/s with the following settings:

Memory clock +1450
Core Clock -200
Power limit 70
Fans speed 65

But my Tj temperature is between 96-98° celsius.

Regards

 

[VforV]

Hello,

on my part i'm unable to reach such temperature on my RTX 3090 FTW3 ULTRA Gaming

I use phoenixminer too and I run on average at 117MH/s with the following settings:

Memory clock +1450
Core Clock -200
Power limit 70
Fans speed 65

But my Tj temperature is between 96-98° celsius.

Regards

If you want to brag, at least read the article first. The issue of overheating affects mostly the Founders Edition card, not yours...

 

[S20coeur]

If you want to brag, at least read the article first. The issue of overheating affects mostly the Founders Edition card, not yours...

Seriously?

Isn't the topic title "GDDR6X in 3080 and 3090 Hits 110C While Mining Ethereum"?

Isn't a forum a place to exchange, share and help each other?

EDIT : My bad, I was actually answering UnCertainty08, because without modifications I also reach 110° with this card

 

[Praetorian Hawke]

I have a Strix 3090 and this is what I did... I installed an Alphacool Eiswolf AIO, a copper plate with two 80mm fans on the Alphacool backplate that comes with the AIO unit. Upgraded thermal pads to Fujipoly pads and used Thermal Grizzly Kryonaut paste on the die and VRAM with 1mm copper shims and .5mm thermal pads for the VRAM. It runs 84C constant when mining on the Vram, 55ish on the die on "High" using Nicehash Quickminer. 121mh/s rate. The Eisewolf AIO isn't necessary, I just decided early on that I wanted to try it on whatever "next" card I bought, and this happened to be it. The pad sizes are good for the backplate supplied with their AIO only. View: https://www.reddit.com/r/gpumining/comments/qqn2bx/my_asus_strix_3090_cooling_85c_on_the_vram_when/