[SOLVED] Should I get active cooling for my 3080 backplate?

[cadyn]

I just saw this video from LinusTechTips about the 3080 and 3090 backplate cooling problem,

View: https://www.youtube.com/watch?v=ToTWaZtGOj8


And after installing HWiNFO, yeah it checks out for my card.

I have the Alienware Aurora R11 PC with I7 10700KF and RTX 3080, I bought it because of the GPU shortage and prebuilts pretty much being the only way to get the 3000 series.

I use my computer for gaming with 1 2560x1440 165hz monitor as well as VR, and when I don't have my GPU in use for gaming, I usually spin it up to mine Ethereum, including overnight and when I'm away from my computer for days on end. I've been using it like this for around a month and a half. I did notice some funny things happening on my PC such as games crashing randomly recently which I attributed to software shenanigans, and to my credit reinstalling windows did seem to fix a lot of those issues. Now, I saw this video from linus, and I'm wondering whether those issues could be from letting my GPU run so hot and whether I should invest in the liquid cooling.

So my question is do you think I should invest in the cooling backplate? If so, how would I even install it, I've never installed liquid cooling before (I built my own PC before, and switched out the CPU etc so I'm not a complete noob when it comes to hardware swaps) If I don't get it, what should I expect going forward? Will my GPU just die on me at random, or slowly degrade over time? I'm pretty concerned, so I'd love it if someone could enlighten me on what all this actually means.

 

[Bassman999]

Oh, I missed this post.

Not surprised underclocking the memory changed little.
-No Vram voltage control exists.
-Vram doesn't pull much power. I recall an older Buildzoid video where he mentions that a single chip pulls power in the single digits.
Plus, on a gpu with total board power of 320-370w? Power limit won't do much for that - it's to keep the core clock low while mining; one does not need Gpu Boost boosting core clock during mining.

-Current state of mining is just hard on Vram.

Hey, if you had a 3090, it'd be worse. Those aren't attractive for mining due to how difficult managing the Vram thermals are, but not everyone who has one knows this...

Yeah the 3090 has more than double the Vram and some chips ae on backplate side I believe making cooling more difficult.
Just to see what the temps are I think Ill try the 6800 mining for an hour tomorrow.
I never did it because of swapping and cleaning drivers back and forth.
But the 6800 uses different memory so it will likely be cooler

 

[Bassman999]

If so, yikes; they couldn't be more wrong.
GDDR5 and R5X can be cooled on air alone for normal tasks. No need for pads or sinks, unless chassis airflow is crap.
GDDR6? Yeah-no, air alone isn't going to cut it. R6X has even higher operating thermals.

@cadyn
One of the main reasons for getting a prebuilt is the customer service, so you don't have to deal with the headaches of troubleshooting that can come from DIY.
It kind of sucks to see you considering getting hands on if you were trying to avoid it, but at the same time, I don't see Dell doing anything to alleviate your concerns.

Supposedly Nvidia states 110c is fine?

https://imgur.com/a/YaWMJFP

View: https://imgur.com/a/YaWMJFP

 

[Phaaze88]

Supposedly Nvidia states 110c is fine?
-CLIP-

You knew it was coming.

"Just because it's 'within spec', doesn't make it ok." Simple as that. There's a reason that guides, or general lower operating temperature recommendations, exist against what the specs for different parts state.
In the rep's defense though, they're likely considering gaming, which the load isn't really going to sit there, unless there's poor cooler contact, or inadequate cooling... so a spike up there every now and then isn't exactly terrible.
Sitting at excessive temperatures for indefinite amounts of time is not good for anything.
Specifically, gaming laptops have been getting toasty too, but they normally don't sit there either.

Msi Afterburner: Power limit and temp limit are unlinked, and temp limit is set to 80C for my 1080Ti(Kraken G12+Celsius S36).
Temp limit for my 7820X is set to 86C - used to be 90C - in bios.
There's really no reason to let one's hardware run that close to their rated thermal limits. If it does, there's clearly something wrong - settings too aggressive, inadequate cooling, etc - and you gotta fix it.

If the user is someone who only keeps a gpu for a year or two, "Who cares, right?" Well, what if you want to:
-hand it down to a relative/friend?
-sell it later? Ok, maybe no one cares once it's in the hands of the other person - screwed up, but that's how it is with some folks these days.

 

[Unsounded]

I have always sworn on a fractal design case for airflow with 3 or 4 fans. But whether that can keep a 30's series card cool i have no idea.

 

[Bassman999]

You knew it was coming.

"Just because it's 'within spec', doesn't make it ok." Simple as that. There's a reason that guides, or general lower operating temperature recommendations, exist against what the specs for different parts state.
In the rep's defense though, they're likely considering gaming, which the load isn't really going to sit there, unless there's poor cooler contact, or inadequate cooling... so a spike up there every now and then isn't exactly terrible.
Sitting at excessive temperatures for indefinite amounts of time is not good for anything.
Specifically, gaming laptops have been getting toasty too, but they normally don't sit there either.

Msi Afterburner: Power limit and temp limit are unlinked, and temp limit is set to 80C for my 1080Ti(Kraken G12+Celsius S36).
Temp limit for my 7820X is set to 86C - used to be 90C - in bios.
There's really no reason to let one's hardware run that close to their rated thermal limits. If it does, there's clearly something wrong - settings too aggressive, inadequate cooling, etc - and you gotta fix it.

If the user is someone who only keeps a gpu for a year or two, "Who cares, right?" Well, what if you want to:
-hand it down to a relative/friend?
-sell it later? Ok, maybe no one cares once it's in the hands of the other person - screwed up, but that's how it is with some folks these days.

The sad part is he says he wasnt mining but mining during these temps but just gaming. I know the FE 3080s have more thermal mem t=issues, but thats ridiculous and not a well thought out thermal solution. Also using the cheapest Thermal Pads they can find.

I have always sworn on a fractal design case for airflow with 3 or 4 fans. But whether that can keep a 30's series card cool i have no idea.

I took the side on my case off and blew a box fan right at my GPU and the Vram temps went down by 5c only Mine isnt the FE card but still a 3080

 

[Phaaze88]

I have always sworn on a fractal design case for airflow with 3 or 4 fans. But whether that can keep a 30's series card cool i have no idea.

That's fine.
The issue here is mining on 3080s and 3090s, of which good airflow alone isn't enough. The gpu also needs to have a robust cooler, where Vram cooling wasn't an afterthought.

The sad part is he says he wasnt mining but mining during these temps but just gaming. I know the FE 3080s have more thermal mem t=issues, but thats ridiculous and not a well thought out thermal solution. Also using the cheapest Thermal Pads they can find.

?
I did ask the OP in my 2nd post what was going on in the first hwinfo pic, and they replied with Ethereum mining.
I figured the other pic(hwinfo + Afterburner) they posted was also during mining, after which I explained why the memory underclock didn't do anything for thermals.

I took the side on my case off and blew a box fan right at my GPU and the Vram temps went down by 5c only Mine isnt the FE card but still a 3080

That's not an airflow issue.
Some models are built better than others, and I imagine fewer can actually handle mining without having to mod it.

The 3080FE isn't bad on thermals - from what I've seen from GN and HWUB - but I'm sure Nvidia didn't test test their cooler using any kind of miner app.
The cards are good at it(mining), but no one designed their coolers around it. If a particular model handles it well, it's a bonus.
As I said earlier, mining is akin to extreme overclocking; neither is a normal application; both are running the card out of spec.

 

[Bassman999]

That's fine.
The issue here is mining on 3080s and 3090s, of which good airflow alone isn't enough. The gpu also needs to have a robust cooler, where Vram cooling wasn't an afterthought.


?
I did ask the OP in my 2nd post what was going on in the first hwinfo pic, and they replied with Ethereum mining.
I figured the other pic(hwinfo + Afterburner) they posted was also during mining, after which I explained why the memory underclock didn't do anything for thermals.


That's not an airflow issue.
Some models are built better than others, and I imagine fewer can actually handle mining without having to mod it.

The 3080FE isn't bad on thermals - from what I've seen from GN and HWUB - but I'm sure Nvidia didn't test test their cooler using any kind of miner app.
The cards are good at it(mining), but no one designed their coolers around it. If a particular model handles it well, it's a bonus.
As I said earlier, mining is akin to extreme overclocking; neither is a normal application; both are running the card out of spec.

I just mean that while gaming my Core temp is usually under 60c and Vram temp over 80c. Considering the core (die) is where the majority of power draw is, I think it should not get so hot on the Mem in contrast 44watts or=n the 3080

View: https://www.youtube.com/watch?v=HqbYW188UrE

 

[Phaaze88]

I just mean that while gaming my Core temp is usually under 60c and Vram temp over 80c. Considering the core (die) is where the majority of power draw is, I think it should not get so hot on the Mem in contrast 44watts or=n the 3080

View: https://www.youtube.com/watch?v=HqbYW188UrE

1)I believe that's the Buildzoid video I was talking about.
2)Node shrink = higher thermal density. Not sure what else it could be.

We seem to be on this roll of making things smaller and faster, but not implementing ways to handle the heat from the smaller package...
OR, perhaps we already know how, but they require complete platform redesigns, which would be expensive, thus we are choosing to ride this sinking ship until the very last minute when we are absolutely forced to make the jump... IDK.

 

[Bassman999]

1)I believe that's the Buildzoid video I was talking about.
2)Node shrink = higher thermal density. Not sure what else it could be.

We seem to be on this roll of making things smaller and faster, but not implementing ways to handle the heat from the smaller package...
OR, perhaps we already know how, but they require complete platform redesigns, which would be expensive, thus we are choosing to ride this sinking ship until the very last minute when we are absolutely forced to make the jump... IDK.

Totally agree!
My Ryzen 5900x has brief temp spikes when opening a web page or other small tasks based on the power density.