RTX 4090 Hotspot Temperature > 100C

[estarolas3]

Hello all,

I bought an ASUS TUF RTX 4090 a year and half ago and it has performed fine until about 2 months ago when I noticed the fans immediately ramping up to 100% in most games and the hotspot temperature reaching 110C and above. This temperature is then maintained while the load on the GPU continues (and so are the fan speeds). The card is therefore extremely noisy and, in the medium to long term, this cannot be good for the card itself.

I know this is not normal behaviour and as such I have sent the card for repair/replacement under warranty with the reseller (which is not OCUK), but they have dismissed my claims and initially said that the card was behaving normally and saw no such high temperatures. Now, after shipping the card back, they have changed their tune (I have since sent them more evidence by testing the card on different systems and observing the same overheating behaviour) and claim that this behaviour is completely normal for an RTX 4090. This is an excerpt from a recent email:

"These hotspot temperatures aren't a cause for concern, looking through the speeds you're experiencing the clock speed is only dipping by a small amount which wouldn't indicative of a fault per say, especially when the machine is in use and the card is undergoing GPU intensive tasks. the temperature being so high also isn't necessarily an issue, especially with 4090 cards as they run significantly hotter than the previous versions."

This is one of the many screenshots I have sent them of the card reaching 110C/100% fan speed shortly after starting a game or a benchmark (and again they had the chance to experience this themselves as they had the card in their possession).

CyberpunkScreenshot

Since I have been going back and forth for with them for a while on this and they are refusing to admit that this is a fault, what is your advice?

Thank you in advance!

PS: The specs of the systems I have tried this card on are as follows:

AMD Ryzen 7800x3d
Arctic Liquid Freezer III 360
MSI MAG X670E Tomahawk
32GB (2x16GB) DDR5 Corsair Vengeance
Crucial T500 M.2 SSD 2TB
Corsair RM1000x PSU

AMD Ryzen 5800x
NH-D15 chromax.black
Asus X570 ROG Crosshair VIII Hero
32 GB (2x16GB) DDR4 Teamgroup
Sabrent Rocket 4 Plus M.2 SSD 1TB
Corsair RM850x PSU

 

[Phixsator]

At first glance it isn't a fault in the strictest meaning of the word. Sounds to me that the card will be solid again with acceptable temps after a repaste.

EDIT: Of course, your warranty will most likely be called into question if you do it yourself. BS in my opinion but nevertheless important info to have on hand. Hotspot is very far off the very acceptable edge or normal temp suggesting to me that you have a spot in the paste on the core that needs replacing(by replacing all the paste btw).

EDIT2: If you decide to go ahead and do a repaste but are feeling a bit insecure about it, you could go with a graphene thermal pad like the kryosheet from Thermal Grizzle or one of the alternatives from other manufactorers. The end result wont be as good but it should solve the problem.

 

[wookiee87]

What kind of gpu temps are you seeing if the hotspot is 110c, i would say you have thermal paste leakout more common on 4090s due to the temprature diffrences the card experences under load and then idle conditions. However case airflow is most important with 40 series (4080-4090) the amount of heat these cards can expell is scary sometimes, if all the heat has no where to go it's trapped and will heat up the card and most other components in your pc.

Does your room get super hot after a while, if so you need to open a door or window to bring down the ambient temp, that way the pc can breath and in turn will reduce operating temps.

 

[Ch3m1c4L]

Just repaste it, there was/is a thing about Asus using cheap rubbish paste on their very premium priced GPUs.

 

[Dg834man]

Gpu temp - 60.9 hotspot - 109.3, 50° delta that's not normal or ok.

 

[Finners]

I would stop going back and forth by email I assume and get on the phone and kick up a fuss. Speak to someone higher if you have to.

In this day and age of social media you've got them now being scummy and trying to fob you off, the likes of Reddit/twitter/Facebook will lap these excuses up and will do them more harm than just to accept your RMA.

Your problem is they will want to send it back to an ASUS repair centre and you might have issues all over again from what I read about Asus customer

 

[wookiee87]

Gpu temp - 60.9 hotspot - 109.3, 50° delta that's not normal or ok.

thats very wrong, the delta should be no more than 20-25 between the core and hotspot. i would say you have thermal spill out on part of the die which is making the hotspot temp read very high indeed

 

[tommybhoy]

Don't know vram tolerance on 4090s but 100% fan speed within a few minutes while gaming isn't within spec of any gpu.

They could be testing open bench, air con workspace, testing via synthetic benchmarks etc and it's 'fine' their end.

If your GPU is orientated on a riser, remove riser and check temps again.

As above, if the room heats up, it's only recirculating hotter air and your system heats up.

Use your phone to externally capture booting into windows and launching/playing (multiple) games with OSD info on screen, upload to YT and send them that info, see what they say.

 

[Joxeon]

Just tell them you will send your data and the RMA refusal response to gamers nexus.

 

[RavenLunatic]

I had a 2.5-year-old Zotac 3090 that I bought from OCUK that developed a problem with the hotspot temperature going over 100 degrees. Sent it back to OCUK and they replaced it straight away. Hotspot temperature is definitely a valid concern. Have you tried to contact ASUS direct? Personally I would not re-paste until I had exhausted all other avenues.

 

[estarolas3]

I would stop going back and forth by email I assume and get on the phone and kick up a fuss. Speak to someone higher if you have to.

In this day and age of social media you've got them now being scummy and trying to fob you off, the likes of Reddit/twitter/Facebook will lap these excuses up and will do them more harm than just to accept your RMA.

Your problem is they will want to send it back to an ASUS repair centre and you might have issues all over again from what I read about Asus customer

I will try to do that - I did spend some time on the phone with the technician that evaluated my card, but all he elected to say was that he could not see what I was seeing in terms of overheating (quoting that they had used a FLIR camera - which, if true, made me consider the possibility of a bad temperature sensor). It was very hard to get through to him, but I will attempt to do so again (and ask to speak with a manager as well).

thats very wrong, the delta should be no more than 20-25 between the core and hotspot. i would say you have thermal spill out on part of the die which is making the hotspot temp read very high indeed

Yes, this is still the most likely scenario.

Don't know vram tolerance on 4090s but 100% fan speed within a few minutes while gaming isn't within spec of any gpu.

They could be testing open bench, air con workspace, testing via synthetic benchmarks etc and it's 'fine' their end.

If your GPU is orientated on a riser, remove riser and check temps again.

As above, if the room heats up, it's only recirculating hotter air and your system heats up.

Use your phone to externally capture booting into windows and launching/playing (multiple) games with OSD info on screen, upload to YT and send them that info, see what they say.

The second machine I tested on is indeed in an open bench. The ambient and system temperatures (CPU, MB) stay relatively low throughout the tests. The card is not particularly hot during load - the exhaust air is warm, but not too warm - which lends credence to the theory that it is an area in the GPU that just does not have proper thermal contact and where the temperature therefore rises well above the GPU core average temperature. I have actually recorded and sent a video to the reseller showing what happened in an AIDA64 GPU test (including live temperature and fan speed plots on screen, and the noise in the background), but I think they have simply dismissed it as they never mentioned it in subsequent emails.

I had a 2.5-year-old Zotac 3090 that I bought from OCUK that developed a problem with the hotspot temperature going over 100 degrees. Sent it back to OCUK and they replaced it straight away. Hotspot temperature is definitely a valid concern. Have you tried to contact ASUS direct? Personally I would not re-paste until I had exhausted all other avenues.

Yes, actually I contacted ASUS even before going to the reseller. They are quite hopeless I am afraid: after having sent me a generic overheating troubleshooting guide that seemed to be tailored for laptops (?) I insisted and they finally had this to say:

For such products (components, accessories and networking), the warranty applies trough your retailer, not directly to you as end user, therefore, if you would like to take advantage of your warranty, I can only suggest that you contact your place of purchase for further assistance in regards to repair/replacement procedures, strengthened by the fact that you did not acquired the unit directly from ASUS, but through a seller.

Just repaste it, there was/is a thing about Asus using cheap rubbish paste on their very premium priced GPUs.

You are right, and I might end up doing so myself, but I wanted to try and get it repaired/replaced under warranty first.


Another thing that occured to me is that since I purchased the GPU with a credit card, and it is still under warranty, is section 75 of the Consumer Act applicable here? They are effectively refusing to service/replace a product under warranty. The tricky bit is finding evidence/expert opinion that corroborates this not being normal GPU behavior.

 

[Rogueywon]

My Inno3D 4090 was quite toasty at first and in some narrow scenarios (mostly involving Total War games when the CPU was also pumping out heat) the hotspot could hit 100 (highest I saw was 101), with the general temp at around 90. A fresh coat of thermal paste knocked 15 degrees off both.

A 50 degree delta is not normal or healthy and I'm pretty sure 110 is where the capping kicks in, so you may be losing performance.

 

[Phixsator]

My Inno3D 4090 was quite toasty at first and in some narrow scenarios (mostly involving Total War games when the CPU was also pumping out heat) the hotspot could hit 100 (highest I saw was 101), with the general temp at around 90. A fresh coat of thermal paste knocked 15 degrees off both.

A 50 degree delta is not normal or healthy and I'm pretty sure 110 is where the capping kicks in, so you may be losing performance.

edge/normal temps and hot spot shouldn't be more than 10-15c apart imho.

 

[McCarthy]

I’m wondering if the seller is a four letter word maybe? I had to send a full system back and get a refund because they were so incompetent.

You have the evidence to prove you have the issue. If you can prove that this is not normal which should be as simple as your results along with online specs showing what is classed as “not normal” you can take it to your credit card and show what you can. That is your best bet.

I would not re paste it because then you have voided the warranty, and what if it doesn’t fix the issue?


Look like you have exhausted everything else and they are fobbing you off.

 

[estarolas3]

I’m wondering if the seller is a four letter word maybe? I had to send a full system back and get a refund because they were so incompetent.

You have the evidence to prove you have the issue. If you can prove that this is not normal which should be as simple as your results along with online specs showing what is classed as “not normal” you can take it to your credit card and show what you can. That is your best bet.

I would not re paste it because then you have voided the warranty, and what if it doesn’t fix the issue?


Look like you have exhausted everything else and they are fobbing you off.

No, in this case the reseller's name is not a 4 letter word.

The reseller has agreed to an RMA so it is being sent back to ASUS...

 

[wookiee87]

No, in this case the reseller's name is not a 4 letter word.

The reseller has agreed to an RMA so it is being sent back to ASUS...

Shouldn't have taken all the back and forth though fronm the retailer, However good outcome never the less, fingers crossed its tested and fixed, or even replaced with a good card.