Rx 6800 XT card hotspot temp is 102+ c

[Vya Domus]

Furmark can be dangerous even with default settings.

I really don't think that is the case anymore, I get way lower temperatures and power consumption in Furmark compared to a lot of other benchmarks and games.

In another note, don't drivers "nerf" Furmark nowadays so its less nocive?

They do, which is why it's pretty irrelevant these days.

 

[Nike_486DX]

Check your delta (normal temp sensor vs hotspot sensor), if under full load you see more than 15C difference then you got bad mount or your paste pumped out. Never use silicone based pastes (noctua nth1, nth2, MX4 5 6, Silver 5, Kryonaut, Deepcool z9, CoolerMaster gel maker, etc etc etc) they are all trash and not suitable for bare die application (gpu/laptops), gets pumped out in less than 4 months, so unless you want to repaste 6 times more than you should (once per 2 years), avoid silicone pastes.

Hydronaut is a budget friendly pump-proof option, then there is ptm 7950 honeywell, and then there is conductonaut. On my 6600 i got conductonaut and the hotspot delta is just 7 degrees after long continuous 100% load.

 

[Kodehawa]

Check your delta (normal temp sensor vs hotspot sensor), if under full load you see more than 15C difference then you got bad mount or your paste pumped out. Never use silicone based pastes (noctua nth1, nth2, MX4 5 6, Silver 5, Kryonaut, Deepcool z9, CoolerMaster gel maker, etc etc etc) they are all trash and not suitable for bare die application (gpu/laptops), gets pumped out in less than 4 months, so unless you want to repaste 6 times more than you should (once per 2 years), avoid silicone pastes.

Hydronaut is a budget friendly pump-proof option, then there is ptm 7950 honeywell, and then there is conductonaut. On my 6600 i got conductonaut and the hotspot delta is just 7 degrees after long continuous 100% load.

Can confirm. I repasted my GPU with NT-H1 (stock paste was pumped out too, and I was seeing up to 98C hotspot -- 35C delta sometimes) and it was pumped out in like a month or two. Kind of a bummer, but 7950 has me at ~15C hotspot delta, which is great.

Condoctonaut would give me an anxiety attack honestly, lmao.

 

[Belfaborac]

RDNA (1,2 and 3) have extreme hotspot deltas even under water...

I hardly think that's anything like a Universal Truth. My 6800XT Red Devil had something like a 15C delta with the original cooler and less than 10C when I put a water block on it. That doesn't classify as extreme in my book.

 

[Vya Domus]

It's also related to how hot the card is running to begin with/power consumption. At stock my 7900XT has about 15-18C delta but once overclocked that rises to 25C despite the fact that the edge temperature barely changes.

 

[freeagent]

I could not in good faith let anything run at anywhere near 100c or over in my PC. That is crazy.. normalized

 

[Martin007]

Check your delta (normal temp sensor vs hotspot sensor), if under full load you see more than 15C difference then you got bad mount or your paste pumped out. Never use silicone based pastes (noctua nth1, nth2, MX4 5 6, Silver 5, Kryonaut, Deepcool z9, CoolerMaster gel maker, etc etc etc) they are all trash and not suitable for bare die application (gpu/laptops), gets pumped out in less than 4 months, so unless you want to repaste 6 times more than you should (once per 2 years), avoid silicone pastes.

Hydronaut is a budget friendly pump-proof option, then there is ptm 7950 honeywell, and then there is conductonaut. On my 6600 i got conductonaut and the hotspot delta is just 7 degrees after long continuous 100% load.

Thermal Grizzly - Hydronaut ? and Thermal Grizzly Kryonaut ?​

I ran out of my current thermal paste all do pump-out I will buy thermal grizzly as soon as I could. But the gpu when I open it , it has thermal pad so why thermal pads did not work ? I tried gelid

I found also I can buy COOLERMASTER CryoFuze Violet Ultra-High Performance , Arctic MX-4 , Cooler Master MasterGel Pro V2

Which is better ?

 

[Vya Domus]

This isn't a problem related to the paste/pump out/etc, that stuff can affect the temperatures by a couple of degrees at most, the contact between the die and cooler simply isn't good. This is either because the cold plate is uneven or the cooler isn't screwed in evenly.

It's also possible that the thermal pads they used for the VRAM/VRM are ever so slightly too tall, unlikely but manufacturers do screw this up from time to time.

 

[Martin007]

This isn't a problem related to the paste/pump out/etc, that stuff can affect the temperatures by a couple of degrees at most, the contact between the die and cooler simply isn't good. This is either because the cold plate is uneven or the cooler isn't screwed in evenly.

It's also possible that the thermal pads they used for the VRAM/VRM are ever so slightly too tall, unlikely but manufacturers do screw this up from time to time.

I tried to screw the cooler many times and they should be even but how I know if the cooler is not contacting the chip ?

Tried again now with Adrenaline undervolt but when I do stress test from adrenalin the hotspot still reach 109c when it reach 110c system shutdown.

 

[Kodehawa]

This isn't a problem related to the paste/pump out/etc, that stuff can affect the temperatures by a couple of degrees at most, the contact between the die and cooler simply isn't good. This is either because the cold plate is uneven or the cooler isn't screwed in evenly.

It's also possible that the thermal pads they used for the VRAM/VRM are ever so slightly too tall, unlikely but manufacturers do screw this up from time to time.

I did have a funny thing where my hotspot deltas were closer to 35C for the longest time, and I managed to get them under 15C now with PTM. I imagine its an ok workaround for bad cold plates if its not horrid.

 

[Martin007]

These are my trials with thermal pad and thermal greases

 

[Kodehawa]

These are my trials with thermal pad and thermal greases

Oh my god what is this.

 

[Martin007]

maybe it is best to remove all pads and use thermal paste so the cooler can touch all chips ?

 

[Kapone33]

With RDNA cards one of the things I have noticed to affect temps is also the quality of the thermal pads they use on the Memory chips and VRM. As the above picture shows the heat that the memory produces is real. If you use Noctua, I find that NTH2 works way better on GPUs than NTH1. The most important thing when you take a card apart is making sure the pressure is even across the card. I find that the pads that come with waterblocks are fine as well. Alphacool give you most diverse and EK give you well working ones that aare usually thin and a real pita to remove the protective film. You will see people use tweezers on Youtube to remove them.

 

[Martin007]

With RDNA cards one of the things I have noticed to affect temps is also the quality of the thermal pads they use on the Memory chips and VRM. As the above picture shows the heat that the memory produces is real. If you use Noctua, I find that NTH2 works way better on GPUs than NTH1. The most important thing when you take a card apart is making sure the pressure is even across the card. I find that the pads that come with waterblocks are fine as well. Alphacool give you most diverse and EK give you well working ones that aare usually thin and a real pita to remove the protective film. You will see people use tweezers on Youtube to remove them.

Which method to use ? I am so confused. Available to me these ones :

Thermal Grizzly - Hydronaut and Thermal Grizzly Kryonaut ​

COOLERMASTER CryoFuze Violet Ultra-High Performance , Arctic MX-4 , Cooler Master MasterGel Pro V2, NTH2

I really dont know also how to make sure all chips are even with the cooler so should I use paste instead of pads on all chips ?

 

[Vya Domus]

These are my trials with thermal pad and thermal greases

maybe it is best to remove all pads and use thermal paste so the cooler can touch all chips ?

What is that stuff that you used ? That looks like that weird thermal putty not actual pads. Either way you cannot use both pads and paste, if that's what you have been doing all this time that's why you have those temps.

You can't just use any pads, they have to be right height. I found this for that card :

The thickness of thermal pads are 1mm and 1.5 mm.

51.5mm x 6mm x 1.0mm

55.5mm x 10.4mm x 1.0mm

48mm x 14mm x 1.5mm (2 pcs)

43mm x 14mm x 1.5mm

89mm x 10.4mm x 1.0mm

86mm x 6mm x 1.0mm

You have to clean all that out and redo the thermal pads properly, just use paste on the die.

 

[Martin007]

What is that stuff that you used ? That looks like that weird thermal putty not actual pads. Either way you cannot use both pads and paste, if that's what you have been doing all this time that's why you have those temps.

You can't just use any pads, they have to be right height. I found this for that card :

The thickness of thermal pads are 1mm and 1.5 mm.

51.5mm x 6mm x 1.0mm

55.5mm x 10.4mm x 1.0mm

48mm x 14mm x 1.5mm (2 pcs)

43mm x 14mm x 1.5mm

89mm x 10.4mm x 1.0mm

86mm x 6mm x 1.0mm

You have to clean all that out and redo the thermal pads properly, just use paste on the die.

Thanks alot I was searching for these info. I used gelid pads but the original thermal solution was also a pad on the die

 

[Vya Domus]

the original thermal solution was also a pad on the die

I think that wasn't a pad but one of those ptm7950 things, they look like pads when cold.

 

[Martin007]

I think that wasn't a pad but one of those ptm7950 things, they look like pads when cold.

Cant buy ptm7950 so what is the best alternative ?

 

[wNotyarD]

Cant buy ptm7950 so what is the best alternative ?

For a quick fix on the GPU core, you may use any good thermal paste (yes, some may pump out faster, but that's why I said quick fix). You'll still need thermal pads for the VRAM and VRM chips, though.

Do you have some isopropyl alcohol to clean everything up?

 

[Martin007]

For a quick fix on the GPU core, you may use any good thermal paste (yes, some may pump out faster, but that's why I said quick fix). You'll still need thermal pads for the VRAM and VRM chips, though.

Do you have some isopropyl alcohol to clean everything up?

I did thanks. I will try buying couple of good paste and see the result but I am not optimistic because even the two kinds I tried did not solve the problem even temporarily

 

[wNotyarD]

I did thanks. I will try buying couple of good paste and see the result but I am not optimistic because even the two kinds I tried did not solve the problem even temporarily

With how those thermal pads/putty look in your pictures, I wouldn't doubt that they're too thick and thus the GPU can't transfer heat correctly to the cold plate due to incorrect mounting pressure.

 

[Martin007]

With how those thermal pads/putty look in your pictures, I wouldn't doubt that they're too thick and thus the GPU can't transfer heat correctly to the cold plate due to incorrect mounting pressure.

I wish this is the case. I tried thermal paste on the die only and left other chips with old pads as it was but the paste or the pad did not solve the problem. All testing on the die only and if I used paste on all chips it wont touch the cooler. I made an order now to get new paste and pads. Will have to wait several days.

conductivity of 6.0 W/Mk consider to be acceptable for VRAM ?

 

[torkema61]

These are my trials with thermal pad and thermal greases

Dude.. GPU has no paste, paste has GPU.

Clean the paste and pads well with isopropyl alcohol for repaste. And watch the guide videos for this repaste process. As someone who has never experienced it before, it will be a very difficult experience, you need to be careful.

 

[Frick]

even in winter with 18c ambient. my 6950XT can stay at around 90-95°C for maybe 10 minutes until it starts to hit 110-113°C and throttles.
same for my 5700XT (when overclocked), 6800XT Aorus Master, 6900XT Speedster and 6900XT Nitro+. all are at least hitting 100°C on the hotspot after a while. the Nitro + and 6950XT hits 113°C even at 100% fanspeed with a 40°C delta after three repastes.

I'll have to check, but I don't think I've seen my XFX 6950xt Speedster Merc 319 at those temps...