3090 Users, what are your hotspot temps?

So I’ve always wondered, why, when playing MSFS my PC fans sound like a plane taking off. Today I loaded up GPUZ whilst flying around with the drone, and my hotspot was 103 degrees. This seems insanely hot to me. I think the actual GPU temp was 88 degrees or something.

Should I be worried? I have an MSI Ventus 3090.

Both temps are high. While the hotspot is more „for information“ the gpu starts throttling at 84C.

Do you have OC? Whats your power draw and do you hit or stay at the powertarget?

What is your memory junction temp?

I have gpu temps in the 60‘s, hotspot temps in the 70‘s and memory temps around the 90‘s with my evga 3090xc ultra.

Undervolting is highly recommended for the best performance with the 3090.

With those temps you are slowly killing your gpu … it might keep running for a few months, maybe a year or it might even die today but to be sure it will die eventually and I can’t understand how your gpu’s drivers even allow it to get that hot. Military specced hardware has to work for something like thirty seconds with temperatures not much higher than yours.

Memories up to 90-95 °C. It’s normal.
GPU up to 79 °C, it gets throttled beyond that.

depends always from frame rates.
With 60fps cap:
gpu temp: 60s-70s ( but usually below 65c)
hot spot memory temp : 70s-80 ( maximum)
Asus 3090 rog strix OC

https://www.overclock.net/threads/3090-gpu-hotspot-temperature-30c-delta.1777947/

Yes and no - GDDR6X max safe vram temp +105 °C
https://www.micron.com/products/ultra-bandwidth-solutions/gddr6x

Buy thermal pads :wink:
https://www.overclock.net/threads/3080-3090-junction-temp-cooling-thread.1777046/

I will do it after the warranty expires (Jan 2022). It has a warranty seal on one of the screws :slight_smile:

It’s all the RMA 'd miners cards that have caused the the gpu shortage in the first place, they even plan on burning them out before the returns deadline.

It’s a broader supply chain issue impacting not just semi conductors but even simpler components right now.
As manufacturers struggle to source replacements they don’t always get the quality they were after…now is not a great time to buy electronics. Reports already of early failures caused by poor quality replacement parts in a couple of industries.

However, I agree with your point. Things are indeed made worse by Miners who absolutely do thrash the kit until it’s almost dead and then scalp it on eBay…it’s disgusting.

1 Like

Regarding your temps. Yes. Those could be better.

  1. Limit frame rate. MSFS2020 is pretty janky and unlimited frame rates will result in 100% GPU load when updating, in menu’s and splash screens.

  2. I assume air cooled card?
    Check your case airflow. Improve if possible.

I have an Asus Tuff 3090 which was originally air cooled. The air cooling was “ok” - not great, just ok. It kept the GPU cool but the memory junction was way higher than I was happy with.

There is a lot of memory on the back of a 3090 and that is basically getting passive cooled by nothing more than the backplate. (At least on the Tuff and some others I’ve seen stripped down online)

Mine is now part of my custom watercooling loop but there wasn’t a waterblock with an active block for the back of my card so I went with the block from mp5works which sits on top of the backplate. It’s not perfect but it’s a lot better.

Depending on your will to custom cool your PC and the cars you have you might actually find a compatible back and front cooling block. Check EKWB for compatibility.

Ideally your hot spot temperature should be close to your actual GPU core temperature reported, but there’s always a slight difference. On my TUF 3090, my core temp is usually maxed around 65c during gaming with the hot spot temps around 70-75c. From a technical standpoint, the GPU core temp reported is more of an average of all GPU core sensors, while the hot spot temp is the hottest sensor on the die. If you’re having such a large delta between the core and hot spot temperature, I suspect you have a thermal paste issue. Either your thermal paste isn’t covering the die completely, or your heatsink isn’t making good contact completely over the die.

If your actual GPU core temp is near 88c, that’s a big issue where throttling starts become a concern. I don’t usually adovate opening your card up to repaste your die and/or thermal pads. Although fairly easy, lots can go wrong if you don’t know what you’re doing. I’ll leave that choice up to you. There’s quite few YouTube videos for reference.

I run mine at ~1998Mhz Core clock and get about 70C max. Is GPUZ set to Fahrenheit?

Thanks, I think I might have over reported my GPU temp. Running some 3DMark benchmarks its peaks around 73.5 degrees, but I can’t see a hotspot recording on 3D Mark. My hotspot was definitely 103 when running MSFS though. I’ll run some more tests tonight to make sure I’m accurately reporting all my temps.

Running 3D mark, my temps return back to normal within a few seconds it seems.
They currently read GPU = 47.3 Hot Spot = 59.7 Memory 52 which seems to be their usual idle temps.

So they get cool really quickly, but heat up just as quickly. Are those temps too high for idle use? The hotspot does seem the hottest one, and quite a bit hotter than everything else.

My 3D mark scores tend to be about 400 points below average for my setup, although I’m not overclocking anything so I don’t know if that’s normal?

It’s normal for it to quickly go back to normal idle temperatures after 3D rendering. It’s also normal for the temperature to quickly rise once 3D rendering begins. Mine is usually back to idle temperatures about 30 seconds after gaming. For 3DMark, make sure you’re running a benchmark that actually pushes your GPU to 100%. Spy Extreme will do the trick for a 3090.

I’d suggest you try logging the sensors to a file to so you can track exactly how your temperature varies. If you have GPUZ, you can easily log it to a text file in the ‘Sensors’ tab. Do this each time you play MSFS or other games so you can see exactly how hot they get.

I wouldn’t dwell too much on that 3DMark score. There’s always a slight variance. If you’re having high temperature issues, your boosting frequency may be limited and might explain your slightly lower score than the average. It could be as simple as that.

Something to consider / take a look at (that I have seen in other forms) is it neccessarily the card, or do you have cooling issues going on in your case?

The memory ran hot in my 3090 but after allot of poking around found it to be a manufacturer defect. They put thermal paste only where the GPU was, and none on the metal that makes contact with the memory.

Took card apart, applied thermal paste and dropped the temp 20-30 degrees on the mem.

2 Likes

I’ve just seen a couple Reddit posts that have said the same. It seems appalling that they can sell £1500 cards and not apply enough paste to keep it cool. I’m am not a techy person at all. I buy prebuilts and switching GPU’s, adding ram and storage are the most complex upgrades that I’ve ever done.

I certainly wouldn’t be comfortable taking my card apart. Do you know if there are reputable places I could go and have that done? Taking it to PCWorld to be serviced for example?

I realise it will cost me more, but I’d happily pay £50 or so to a professional to do it properly than risk myself killing my card…

I would imagine any reputable PC repair shop could do this for you, there really isn’t much to these cards but I can understand not wanting to take it apart.

1 Like

One of the fans, the middle one I think also has a slight rattle at low fan speeds.

I think the highest I have seen my 3090 FE get to is about 65C.

Depending on ambient temp, I’m 72-76 degrees with maxed out everything using a strix 3090. 103 is insanely high. Yikes!