• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

ASUS ProART RTX 4070 Ti OC Thermal Pad Thickness?

MisguidedGhost

New Member
Joined
Jan 5, 2024
Messages
6 (0.01/day)
I just ordered an ASUS ProArt RTX 4070 Ti OC last night and it arrives tomorrow. I heard that it doesn't have the best thermals so I was looking to do a repaste and repad of the GPU. The only problem is I have no idea what thickness thermal pads I will need for the repad. I am posting here in hopes that someone else has this GPU and has done a repad. I've posted this question on 4 different subreddits and so far no one has gotten back to me except for a few GPU newbs telling me not to do it because it will void the warranty of the GPU. Bottomline is if the thermals suck and I can fix it by doing a repaste and repad then I'm going to do it even if I am voiding the warranty, which actually shouldn't be a problem as long as I keep the old pads and put them back on before RMAing it if I ever need to do that.
 
You don't need to repaste anything. Leave it alone.
 
You don't need to repaste anything. Leave it alone.
but if I do? If I get temps over 75C be that gpu, hotspot or mem junction I am 100% repasting and repadding. I don't get temps that high on my RTX 3080. I'm only asking this because I watched a review last night of the gpu and the reviewer got a hotspot temp of 87C, which for me is unacceptable.
 
If I get temps over 75C be that gpu, hotspot or mem junction I am 100% repasting and repadding.
Based on what? The fact that you think you're smarter than the NVIDIA engineers who built the GPU with thermal protections?

My 2080 Ti regularly hits over 90 °C hotspot with a core temperature of under 70 °C - and I literally don't care, because it has never throttled for any temperature-related reason. Unless it does, I will continue to not care.

The thing is, nobody apart from NVIDIA engineers actually knows what the hotspot temperature even means, or what a "good" or "bad" hotspot temp actually is. You will find a lot of people on the internet making claims about this, e.g. that the difference between core and hotspot should be under 15 °C, but the truth is none of them actually know - they're just parroting back something that they read somewhere, that someone else made up.

I don't get temps that high on my RTX 3080.
A completely different GPU with a completely different cooler in a completely different system, gives completely different readings? SAY IT AIN'T SO!

I'm only asking this because I watched a review last night of the gpu and the reviewer got a hotspot temp of 87C, which for me is unacceptable.
87°C in a system completely different from yours. And without knowing what hotspot temps other GPUs in the same system record, there is zero comparative evidence to conclude this is an overly high temperature for this particular GPU.
 
87°C in a system completely different from yours. And without knowing what hotspot temps other GPUs in the same system record, there is zero comparative evidence to conclude this is an overly high temperature for this particular GPU.
I pointed out my 3080 because it uses more power and older less efficient VRAM, so saying that doesn't matter because it's a different gpu is just plain stupid considering the 4070Ti uses more efficient VRAM and less power. I do agree with you that I don't know the complete specs or the environment and ambient temp of the environment of the reviewer, but I do know my environment and ambient temp and if it reaches 87C then that is too hot. Again i pointed out my 3080 because that card is less efficient and power hungry and is known to run hot but after I repasted it and repadded it I got/have great temps. I'm not on here to get in an arguement with people. I'm simply trying to find out the thickness of the thermal pads, not get into an arguement of what I think or you think is considered hot for a GPU, we obviously have differering opinions on that matter, but be that as it may I still would like the thickness of the thermal pads incase it does run hot.
I also know that 87C isn't really a big deal, it's considered in spec but if I can get it to run cooler and last longer then why wouldn't I?
 
I pointed out my 3080 because it uses more power and older less efficient VRAM, so saying that doesn't matter because it's a different gpu is just plain stupid considering the 4070Ti uses more efficient VRAM and less power. I do agree with you that I don't know the complete specs or the environment and ambient temp of the environment of the reviewer, but I do know my environment and ambient temp and if it reaches 87C then that is too hot. Again i pointed out my 3080 because that card is less efficient and power hungry and is known to run hot but after I repasted it and repadded it I got/have great temps. I'm not on here to get in an arguement with people. I'm simply trying to find out the thickness of the thermal pads, not get into an arguement of what I think or you think is considered hot for a GPU, we obviously have differering opinions on that matter, but be that as it may I still would like the thickness of the thermal pads incase it does run hot.
I also know that 87C isn't really a big deal, it's considered in spec but if I can get it to run cooler and last longer then why wouldn't I?

It's kinda hard to hit 75C edge temp (GPU temp) on a reasonably sized 4070 Ti. You have to try pretty hard with shitty airflow. The current gen of GDDR6X runs pretty cool with even minimal effort to cool it with a baseplate, compared to RTX 30 series. ProArt looks pretty capable and big.

I'd let it run and keep an eye on HWinfo and confirm there actually is a problem (VRAM >90C or delta >20C) before opening it up. Asus is not known to be kind in any warranty and RMA matters - even in jurisdictions where Warranty Void stickers are illegal, it doesn't mean that Asus is not going to find other ways to screw you over.

Hotspot is impossible to predict. Every single sample of the same card is different. <10C delta is outstanding, +10C delta from edge temp is excellent (I'm at about 9-11), +15C is pretty common and +20C is on the edge of acceptable. More complex Radeon dies can get up to 25C delta without there being any real problem.

At the end of the day, replacing pads would do absolutely nothing to help your edge temp to hotspot temp delta - both are tied to AD104 and neither have anything to do with VRAM packages.
 
Last edited:
It's kinda hard to hit 75C edge temp (GPU temp) on a reasonably sized 4070 Ti. You have to try pretty hard with shitty airflow. The current gen of GDDR6X runs pretty cool with even minimal effort to cool it with a baseplate, compared to RTX 30 series. ProArt looks pretty capable and big.

I'd let it run and keep an eye on HWinfo and confirm there actually is a problem (VRAM >90C or delta >20C) before opening it up. Asus is not known to be kind in any warranty and RMA matters - even in jurisdictions where Warranty Void stickers are illegal, it doesn't mean that Asus is not going to find other ways to screw you over.

Hotspot is impossible to predict. Every single sample of the same card is different. <10C delta is outstanding, +10C delta from edge temp is excellent (I'm at about 9-11), +15C is pretty common and +20C is on the edge of acceptable. More complex Radeon dies can get up to 25C delta without there being any real problem.

At the end of the day, replacing pads would do absolutely nothing to help your edge temp to hotspot temp delta - both are tied to AD104 and neither have anything to do with VRAM packages.
alright. I'll definitely be checking when i get it tomorrow. Also just for reference I've been processing a video in Topaz Video AI for the last hour and a half with my 3080 at full load and my temps are GPU: 65.3C, Mem Junction: 72C and Hotspot: 76.4C this is the norm for my gpu temps under full load. I can run it for hours at full load and it will stay around those temps. My current GPU is an ASUS Tuf Gaming OC RTX 3080. Honestly I would have gone with the TUF 4070Ti but estimated delivery wasn't until Jan 16th meanwhile the ProArt was 2-days. I only use my gpu for Handbrake, DaVinci Resolve and both Topaz Video Ai and Topaz Photo AI. Sometimes I'm running Topaz Video AI for 10hrs straight so you can see why I want to have the lowest possible temps, unfortunately watercooling isn't an affordable option.
 
alright. I'll definitely be checking when i get it tomorrow. Also just for reference I've been processing a video in Topaz Video AI for the last hour and a half with my 3080 at full load and my temps are GPU: 65.3C, Mem Junction: 72C and Hotspot: 76.4C this is the norm for my gpu temps under full load. I can run it for hours at full load and it will stay around those temps. My current GPU is an ASUS Tuf Gaming OC RTX 3080. Honestly I would have gone with the TUF 4070Ti but estimated delivery wasn't until Jan 16th meanwhile the ProArt was 2-days. I only use my gpu for Handbrake, DaVinci Resolve and both Topaz Video Ai and Topaz Photo AI. Sometimes I'm running Topaz Video AI for 10hrs straight so you can see why I want to have the lowest possible temps, unfortunately watercooling isn't an affordable option.

You can overclock+undervolt the 4070Ti to ~2650mhz/950mV and everything should be fine and dandy, the coolers on these rtx4000 are so overbuilt that replacing thermal paste/pad or even going watercooling is unnecessary.
 
This^
Just undervolt the card, 99.9% that you won't hit 90ºC hotspot even in the summer months. I got my 4070, to 920mV with the same frequency of 2700Mhz as it has default and -12ºC on my GPU hotspot.
 
It's kinda hard to hit 75C edge temp (GPU temp) on a reasonably sized 4070 Ti. You have to try pretty hard with shitty airflow. The current gen of GDDR6X runs pretty cool with even minimal effort to cool it with a baseplate, compared to RTX 30 series. ProArt looks pretty capable and big.

I'd let it run and keep an eye on HWinfo and confirm there actually is a problem (VRAM >90C or delta >20C) before opening it up. Asus is not known to be kind in any warranty and RMA matters - even in jurisdictions where Warranty Void stickers are illegal, it doesn't mean that Asus is not going to find other ways to screw you over.

Hotspot is impossible to predict. Every single sample of the same card is different. <10C delta is outstanding, +10C delta from edge temp is excellent (I'm at about 9-11), +15C is pretty common and +20C is on the edge of acceptable. More complex Radeon dies can get up to 25C delta without there being any real problem.

At the end of the day, replacing pads would do absolutely nothing to help your edge temp to hotspot temp delta - both are tied to AD104 and neither have anything to do with VRAM packages.
I was looking for any information on people with an Asus 4070 TI Tuf Gaming OC and found this thread and your reply. I have been using mine for 6 months.

My 4070 TI is in a Corsair 5000D with plenty of airflow and an ambient room temp of 21°C. The first 5 months it has been fantastic with a max GPU temp of 65C, memory at 62C and a hotspot of 80C when pulling 290w giving a delta of 15C.

However, over the last month, Temps have increased significantly. At 280-290w load, I'm seeing 72C GPU, 68C memory and 93C hotspot for a delta of 21C. As I'm sure you know, this is due to pump-out.

I am unwilling to RMA my GPU over this. Asus RMA is a total crap-shoot with an emphasis on crap.

I have the new Kryonaut Kryosheet sitting on my desk waiting for me to install it. I do not plan on replacing the pads as the memory Temps are quite acceptable and have likely only crept up in temp due to the GPU increasing in temp. I'll do some basic load testing before and after and report back with my findings.

There are quite a few posts across media platforms of people choosing to repaste their 4070 TI's and seeing Temps drop to more acceptable levels. Whatever paste Asus is using is particularly susceptible to pump-out.
 
Have you checked your PC, case and GPU for dust?
Just a thought...
 
Have you checked your PC, case and GPU for dust?
Just a thought...
It's completely clean, GPU included. All fans working, 5 blowing in, 4 blowing out. My office has a tower air purifier because I hate dusty computers. Watercooled CPU with radiator top mounted blowing out.
 
Any other change in fans function/curve/rpms? (GPU/case)
Room temp change?
Any significant change in system behavior or usage or software that affect cooling?

Sorry for all the (stupid) questions, but I find it hard to believe that in 6 months something gone wrong with the card.
I never came across in a such situation in 20years of PC fiddling with many GPU cards in my systems, some working for a few years.
 
If you care so much about temperatures, watercool, otherwise leave it alone.

This is with a 400 W PL.

1705497009388.png
 
I was looking for any information on people with an Asus 4070 TI Tuf Gaming OC and found this thread and your reply. I have been using mine for 6 months.

...
There are quite a few posts across media platforms of people choosing to repaste their 4070 TI's and seeing Temps drop to more acceptable levels. Whatever paste Asus is using is particularly susceptible to pump-out.
Undervolt it. Got my 4070 Asus to 920mV and temperature drops by about 7º and hotspot by more than 10ºC. Asus is saying "it is withing parameters" so you either use it as is, undervolt it. Or void the warranty(depending where you are in the World) and use some good paste.
PS: my UV results are in this clip:
 
Any other change in fans function/curve/rpms? (GPU/case)
Room temp change?
Any significant change in system behavior or usage or software that affect cooling?

Sorry for all the (stupid) questions, but I find it hard to believe that in 6 months something gone wrong with the card.
I never came across in a such situation in 20years of PC fiddling with many GPU cards in my systems, some working for a few years.
MW2, Diablo, Sniper Elite 4, all send my GPU into the 90's for a hotspot. MW2 and Diablo especially. 95°C is seen after 5 minutes playing either of those two. No other games I play do that.
Undervolt it. Got my 4070 Asus to 920mV and temperature drops by about 7º and hotspot by more than 10ºC. Asus is saying "it is withing parameters" so you either use it as is, undervolt it. Or void the warranty(depending where you are in the World) and use some good paste.
PS: my UV results are in this clip:
I've tried undervolting once and it was extremely unstable once loading a game. I may do more video watching and see what others have done.

Not worried about the warranty from repasting. If it blows up, so be it. I'd rather it not, though, and last a long time. I've reposted more GPU's than I can count and have seen excellent results in the vast majority of them.
 
MW2, Diablo, Sniper Elite 4, all send my GPU into the 90's for a hotspot. MW2 and Diablo especially. 95°C is seen after 5 minutes playing either of those two. No other games I play do that.

I've tried undervolting once and it was extremely unstable once loading a game. I may do more video watching and see what others have done.

Not worried about the warranty from repasting. If it blows up, so be it. I'd rather it not, though, and last a long time. I've reposted more GPU's than I can count and have seen excellent results in the vast majority of them.

My Eagle hasn't changed at all in 12 months, MW2 and SE4/SE5 included. It sounds like Asus is indeed doing a shitty job of pasting. In your case a simple repaste would be preferable to sending it in. Hopefully that works out for you and the pads stay intact for re-use.

OP was pre-emptively worrying over someone else's experience with the card without even having received the card, and contemplating repasting. Jumping the gun would be an understatement.

If it's unstable then your undervolt needs more work. However, I also wouldn't go into it expecting miracles, 4070 Ti doesn't seem to be producing the same unicorn results as 4070s and 4090s do.
 
alright. I'll definitely be checking when i get it tomorrow. Also just for reference I've been processing a video in Topaz Video AI for the last hour and a half with my 3080 at full load and my temps are GPU: 65.3C, Mem Junction: 72C and Hotspot: 76.4C this is the norm for my gpu temps under full load. I can run it for hours at full load and it will stay around those temps. My current GPU is an ASUS Tuf Gaming OC RTX 3080. Honestly I would have gone with the TUF 4070Ti but estimated delivery wasn't until Jan 16th meanwhile the ProArt was 2-days. I only use my gpu for Handbrake, DaVinci Resolve and both Topaz Video Ai and Topaz Photo AI. Sometimes I'm running Topaz Video AI for 10hrs straight so you can see why I want to have the lowest possible temps, unfortunately watercooling isn't an affordable option.
Just because a 3080 ran at temp X does not mean that temp X is a 'norm'. You're making your life very difficult with these self imposed norms.

A hotspot of below 80 C is very low and I would certainly not use that as a norm for any GPU use case. Its bullshit and you won't lengthen the life of your GPU in any possible way shape or form. Fairy tales. That being said, you do you.
 
My Eagle hasn't changed at all in 12 months, MW2 and SE4/SE5 included. It sounds like Asus is indeed doing a shitty job of pasting. In your case a simple repaste would be preferable to sending it in. Hopefully that works out for you and the pads stay intact for re-use.

OP was pre-emptively worrying over someone else's experience with the card without even having received the card, and contemplating repasting. Jumping the gun would be an understatement.

If it's unstable then your undervolt needs more work. However, I also wouldn't go into it expecting miracles, 4070 Ti doesn't seem to be producing the same unicorn results as 4070s and 4090s do.
So here's what I've done.

I got my GPU up to temp playing MW2. 95°C hotspot and 73°C GPU. Memory was fine at 68.

Shut it down, pulled the GPU apart and there's pumpout going on. Not only that, one spot of the die had no TIM on it. The pads covering the VRAM chips and VRM's and Mosfets all remained intact. Also, the original TIM was getting dried out. Cleaned it all off, used TapeMasters 1/4"x39yd Polyimide high temp tape from Amazon to cover the nearby transistors, then laid the 25x25mm Kryosheet over the die. Put it back together and went back to MW2.

After 30 minutes of playing, same room temp, GPU was 71°C, memory at 65C and Hotspot at 81°C.
 

Attachments

  • 20240129_112532.jpg
    20240129_112532.jpg
    2.3 MB · Views: 444
  • 20240129_112410.jpg
    20240129_112410.jpg
    3.2 MB · Views: 614
  • 20240129_120037.jpg
    20240129_120037.jpg
    2.2 MB · Views: 510
Based on what? The fact that you think you're smarter than the NVIDIA engineers who built the GPU with thermal protections?

My 2080 Ti regularly hits over 90 °C hotspot with a core temperature of under 70 °C - and I literally don't care, because it has never throttled for any temperature-related reason. Unless it does, I will continue to not care.

The thing is, nobody apart from NVIDIA engineers actually knows what the hotspot temperature even means, or what a "good" or "bad" hotspot temp actually is. You will find a lot of people on the internet making claims about this, e.g. that the difference between core and hotspot should be under 15 °C, but the truth is none of them actually know - they're just parroting back something that they read somewhere, that someone else made up.


A completely different GPU with a completely different cooler in a completely different system, gives completely different readings? SAY IT AIN'T SO!


87°C in a system completely different from yours. And without knowing what hotspot temps other GPUs in the same system record, there is zero comparative evidence to conclude this is an overly high temperature for this particular GPU.

Hotspot over 90c with core below 70c suggests that the TIM has been pumped out, or wasn't applied properly in the first place.

It was the exact same thing on my gigabyte 4090 gaming oc - took it apart, and half the die didn't even have any TIM. Repasted it, and delta between core and hotspot is now 10c.
 
I agree with @Assimilator, you should listen to him. Even the very first RTX 4070 Ti to roll out of the production line is new enough that its thermal pads and paste haven't yet even possibly begun to degrade. You're very likely going to make things worse, just make sure your case has good airflow and/or bump the fan speed curve a little. Sometimes just disabling zero RPM might improve the thermals enough you don't need to change any other setting for the lifetime of the hardware.
 
Any other change in fans function/curve/rpms? (GPU/case)
Room temp change?
Any significant change in system behavior or usage or software that affect cooling?

Sorry for all the (stupid) questions, but I find it hard to believe that in 6 months something gone wrong with the card.
I never came across in a such situation in 20years of PC fiddling with many GPU cards in my systems, some working for a few years.

Pump-out always happens eventually...

I agree with @Assimilator, you should listen to him. Even the very first RTX 4070 Ti to roll out of the production line is new enough that its thermal pads and paste haven't yet even possibly begun to degrade. You're very likely going to make things worse, just make sure your case has good airflow and/or bump the fan speed curve a little. Sometimes just disabling zero RPM might improve the thermals enough you don't need to change any other setting for the lifetime of the hardware.

Asus usually has fairly good quality control, so it's rare to see terrible TIM applications on their gpus... but you see it fairly often on other vendors, like the gigabyte card i got.
As i wrote further up, i reduced the delta between core and hotspot temp from more than 20c to 10c from just repasting it.
 
Pump-out always happens eventually...
Ok, but of my 10+ GPUs I never experienced that and I had them working sometimes hard for long periods. Some for 1year 24/7 (not loaded 24/7) and others for 2-3years
I quess for people to suggest it it can happen, maybe not so frequently.
 
Asus usually has fairly good quality control, so it's rare to see terrible TIM applications on their gpus... but you see it fairly often on other vendors, like the gigabyte card i got.
As i wrote further up, i reduced the delta between core and hotspot temp from more than 20c to 10c from just repasting it.

Temps also highly depend on ambient. 90 to 100°C hotspot was fairly common to see here in the Brazilian climate without AC on my old TUF RTX 3090. Unless hotspot is consistently in the 100 to 105°C range, I personally wouldn't worry about it.
 
So here's what I've done.

I got my GPU up to temp playing MW2. 95°C hotspot and 73°C GPU. Memory was fine at 68.

Shut it down, pulled the GPU apart and there's pumpout going on. Not only that, one spot of the die had no TIM on it. The pads covering the VRAM chips and VRM's and Mosfets all remained intact. Also, the original TIM was getting dried out. Cleaned it all off, used TapeMasters 1/4"x39yd Polyimide high temp tape from Amazon to cover the nearby transistors, then laid the 25x25mm Kryosheet over the die. Put it back together and went back to MW2.

After 30 minutes of playing, same room temp, GPU was 71°C, memory at 65C and Hotspot at 81°C.

I bet you are glad that you did the repaste - as was i when i did it :)

Temps also highly depend on ambient. 90 to 100°C hotspot was fairly common to see here in the Brazilian climate without AC on my old TUF RTX 3090. Unless hotspot is consistently in the 100 to 105°C range, I personally wouldn't worry about it.

Hotspot temp in itself is meaningless.

It's the delta between the core temp and hotspot that matters. If your core is 90c and your hotspot is 100c, then everything is just getting very toasty, but making proper contact.

However, if your core is 70c, and your hotspot is 100c, then you deffo know that it's either a bad mount, or bad TIM application.

Ok, but of my 10+ GPUs I never experienced that and I had them working sometimes hard for long periods. Some for 1year 24/7 (not loaded 24/7) and others for 2-3years
I quess for people to suggest it it can happen, maybe not so frequently.

Based on the questions you were asking further up, you haven't even been looking at the data, and thus obviously haven't experienced it...
 
Last edited:
Back
Top