GPU Temp: 68°C, Hotspot Temp: 100°C (Sapphire Pulse RX 6700XT)

[Macro Device]

Undervolt it. I run my Powercolor Fighter with these values.

View attachment 328956

This is not the fixed value, this is an offset and it's different for every GPU.

My GPU, however, hates undervolting and only lets me shave it off to ~1180 effective mV at 2620—2720 MHz range. At the range of 2660—2760, UV doesn't work at all, it's only stable at max voltage.

It's still extremely hard to meaningfully undervolt 6700 series GPUs, unlike bigger Navi SKUs. Want less heat, go lesser maximum clock.

 

[Ruru]

This is not the fixed value, this is an offset and it's different for every GPU.

My GPU, however, hates undervolting and only lets me shave it off to ~1180 effective mV at 2620—2720 MHz range. At the range of 2660—2760, UV doesn't work at all, it's only stable at max voltage.

It's still extremely hard to meaningfully undervolt 6700 series GPUs, unlike bigger Navi SKUs. Want less heat, go lesser maximum clock.

Interesting. For me, it stays however close to the set value in load, was at around 1.05V during a Superposition 4K run.

 

[Lew Zealand]

Undervolt it. I run my Powercolor Fighter with these values.

View attachment 328956

Those numbers look great! My PC Fighter does not like that much undervolt, it prefers about 1125 and then set frequencies to taste. However I think I got a crud sample as it really takes some finessing to get it to 2600 MHz (reported in GPU-Z) and it really likes to run at 2500-2550 before efficiency is completely tossed out.

However I don't have easy access to this GPU any more as my kid's 6.5 year old 1060 6GB died and the Fighter 6700 XT is the only 2-slot card that physically fits with his Mobo configuration. Considering it's overkill for his uses, I have it underclocked to 2400 MHz and 1150mV and it's still far faster than the 1060.

High five to AMD Adrenalin here as I can have the card maintain those settings after a restart unlike using Afterburner on the 1060 (and 1660 Super the other kid uses), which requires admin to start up.

 

[ChrisM255]

Those numbers look great! My PC Fighter does not like that much undervolt, it prefers about 1125 and then set frequencies to taste. However I think I got a crud sample as it really takes some finessing to get it to 2600 MHz (reported in GPU-Z) and it really likes to run at 2500-2550 before efficiency is completely tossed out.

However I don't have easy access to this GPU any more as my kid's 6.5 year old 1060 6GB died and the Fighter 6700 XT is the only 2-slot card that physically fits with his Mobo configuration. Considering it's overkill for his uses, I have it underclocked to 2400 MHz and 1150mV and it's still far faster than the 1060.

High five to AMD Adrenalin here as I can have the card maintain those settings after a restart unlike using Afterburner on the 1060 (and 1660 Super the other kid uses), which requires admin to start up.

I love AMDs Adrenalin software... don't miss Nvidia one bit in that respect

 

[Macro Device]

High five to AMD Adrenalin here as I can have the card maintain those settings after a restart unlike using Afterburner on the 1060 (and 1660 Super the other kid uses), which requires admin to start up.

It really LOVES resetting everything to default for no apparent reason, that's why I'm using Afterburner for that despite having access to Adrenalin. Full-on clown fiesta.

 

[torkema61]

SOLVED!!!

I was having the same issue with my GPU temps and a 20-30C OR MORE hotspot delta on a Sapphire Pulse 6700 XT. Repasted and new thermal pads, used TG Kryonaut + TG minus pads with minor difference... saw everyone saying "by design/normal behavior" so I reluctantly excepted it. Over the next couple months kept watching the temps that screams bad IHS contact to me and I took one more attempt.

The answer for me was a 25x25mm Kryosheet + Arctic TP-3 thermal pads! 1mm for the mosfets and 1.5mm for the memory.

I also shaved about 0.1-0.2MM off the height of the 2 rubber spacers opposite of the memory modules between the IHS and the board.

Temps after are what I am used to 60-70C while gaming, quieter, 10 degree delta or less, no more 95-110C with 20-30+ delta!!!

Just wanted to share since no fix online anywhere. ENJOY!

links to products used
Kryosheet https://www.amazon.com/dp/B0C61NWNZ...3d6c4048dbb11&language=en_US&ref_=as_li_ss_tl
TP-3 1mm https://www.amazon.com/ARCTIC-TP-3-...b593993f31926&language=en_US&ref_=as_li_ss_tl
TP-3 1.5mm https://www.amazon.com/ARCTIC-TP-3-...3d44a68ea4fa3&language=en_US&ref_=as_li_ss_tl

before: 99% utilization = 90C+ hotspot, constant spikes up to 95-110C, 20-30C delta
after: 99% utilization = 70C hotspot, occasional spikes up to 85C, 10 degree delta or less, outstanding thermal recovery time!

in COD MW3 = 75C max hotspot, only 2-5 degree delta, 60C memory on a 1440p ultrawide monitor using a moderate overclock

Manual overclock used:
FREQ = 2500-2600MHz
Volts = 1.20v
MEM = 2112MHz, fast timings on memory = results in 2100MHz
Power = +15%
fan = 75%, zero rpm off



This is what worked for me, do at your own risk.

Is it 75% fan speed or the percentage the fan will reach at its maximum?

If I encounter this problem after the warranty period of my card expires, I will try to repair it as you said. Thank you very much!

My undervolt values are as follows:
Mhz: 2581 (base clock)
mV: 1150

 

[ChrisM255]

Is it 75% fan speed or the percentage the fan will reach at its maximum?

If I encounter this problem after the warranty period of my card expires, I will try to repair it as you said. Thank you very much!

My undervolt values are as follows:
Mhz: 2581 (base clock)
mV: 1150

Just setting fan speed in Adrenaline to 75 so it's the max. 65 is a little quieter, but I prefer the lower temps. Lowers the idle temps a bit as well.

 

[torkema61]

Just setting fan speed in Adrenaline to 75 so it's the max. 65 is a little quieter, but I prefer the lower temps. Lowers the idle temps a bit as well.

This is a very high value for me. Of course, temperatures will be lower at this fan speed.

My current fan RPM speed is on average between 1800-2000RPM. This is approximately 45%, and when the fan is at this speed, the GPU is 70°C and the hotspot temperature is 82°C at default settings. When I undervolt, GPU is 62°C and Hotspot is 73-75°C. Fan noise above 2250RPM is a bit annoying. For now, these temperatures are ideal for me.

My case had 2 120mm fans at the front. These fans have a rating of 45 CFM. Instead of these fans, I bought 2 140mm fans and each has 110 CFM. I will test the temperatures again after installing it.

 

[Ruru]

It really LOVES resetting everything to default for no apparent reason, that's why I'm using Afterburner for that despite having access to Adrenalin. Full-on clown fiesta.

Does that sometimes but it takes literally only maybe a second or two that I load the OC profile.

 

[Macro Device]

Does that sometimes but it takes literally only maybe a second or two that I load the OC profile.

I don't care how easy it is to workaround. What matters is that the workaround is needed and this ultimately means this software is garbage. Overreactive, overprotective garbage.

I can imagine it being possible to damage a GCN GPU only using Adrenalin/Crimson (but that definitely required a lot of effort anyway) but RDNA... They are so hard locked it makes no sense to impose mandatory defaulting. If it burns down it's definitely nothing to do with manual overclocking in AMD software.

This "feature" ONLY makes for an absolute nuisance.

 

[AsRock]

RDNA2 has a huge hotspot delta even under Water.
30°C is not rare. a repaste might lower it to 20.

Thing is, if i am understanding the OP right is that changing the mV screws with the fan, i lower mine -20 and never see it over 81c on the hotspot. although without the -20 mV it will go much higher how ever i can change my fan profile as i see fit were it sounds like they cannot.

 

[Lew Zealand]

It really LOVES resetting everything to default for no apparent reason, that's why I'm using Afterburner for that despite having access to Adrenalin. Full-on clown fiesta.

Once I turned off Fast Start, I've had zero problems with Adrenalin setting things to default. That's on these machines:

i7-4790 5600 XT
i7-4790 5600 XT (why have one when you can have two?)
R5-2600 6600 XT
i7-9700F 6700 XT (GPU recently moved to another PC)
R5-5600 7700 XT
R5-5600 6800 XT (same PC as above, just GPU swap)

So no problems in 5 machines but Windows Fast Start needed to be turned off otherwise Adrenalin randomly liked to reset to default in all of them. Afterburner adds nothing except yet more software and needing admin password so I don't use it if I don't need to. Of course it's required for Nvidia GPUs, no better option.

 

[Macro Device]

Once I turned off Fast Start, I've had zero problems with Adrenalin setting things to default. That's on these machines:

i7-4790 5600 XT
i7-4790 5600 XT (why have one when you can have two?)
R5-2600 6600 XT
i7-9700F 6700 XT (GPU recently moved to another PC)
R5-5600 7700 XT
R5-5600 6800 XT (same PC as above, just GPU swap)

So no problems in 5 machines but Windows Fast Start needed to be turned off otherwise Adrenalin randomly liked to reset to default in all of them. Afterburner adds nothing except yet more software and needing admin password so I don't use it if I don't need to. Of course it's required for Nvidia GPUs, no better option.

Eh, I wish I could turn Fast Boot off (it's already disabled).

The only way I don't get everything reset to default is not using AMD software. I tried everything: other OS, other BIOS settings, other BIOS firmware, OTHER MOTHERBOARD, DIFFERENT CPU, etc etc etc. 'haps my GPU is faulty somewhat but I failed to achieve any sort of crash/freeze/blackout during any load at my OC profile so I'm left with a lot of unanswered questions.

 

[torkema61]

Eh, I wish I could turn Fast Boot off (it's already disabled).

The only way I don't get everything reset to default is not using AMD software. I tried everything: other OS, other BIOS settings, other BIOS firmware, OTHER MOTHERBOARD, DIFFERENT CPU, etc etc etc. 'haps my GPU is faulty somewhat but I failed to achieve any sort of crash/freeze/blackout during any load at my OC profile so I'm left with a lot of unanswered questions.

I encounter this problem from time to time, too. My AMD Adrenalin settings are reset at the first startup, at least once a week. I tried every way but couldn't solve it. So I exported my settings. When it is reset, I import the settings I transferred again.

 

[ChrisM255]

This is a very high value for me. Of course, temperatures will be lower at this fan speed.

My current fan RPM speed is on average between 1800-2000RPM. This is approximately 45%, and when the fan is at this speed, the GPU is 70°C and the hotspot temperature is 82°C at default settings. When I undervolt, GPU is 62°C and Hotspot is 73-75°C. Fan noise above 2250RPM is a bit annoying. For now, these temperatures are ideal for me.

My case had 2 120mm fans at the front. These fans have a rating of 45 CFM. Instead of these fans, I bought 2 140mm fans and each has 110 CFM. I will test the temperatures again after installing it.

You obviously do not have the same delta issue the rest of us have... I have a high airflow case with 4 140mm fans and a 120 rear. Airflow ain't the issue.

 

[torkema61]

You obviously do not have the same delta issue the rest of us have... I have a high airflow case with 4 140mm fans and a 120 rear. Airflow ain't the issue.

I had the same problem with my card before I sent it to service. I already created this topic for this reason. They decided to change and gave me a zero card. There is no problem for now, but I feel like I will have the same problem again in the future.

 

[ChrisM255]

Eh, I wish I could turn Fast Boot off (it's already disabled).

The only way I don't get everything reset to default is not using AMD software. I tried everything: other OS, other BIOS settings, other BIOS firmware, OTHER MOTHERBOARD, DIFFERENT CPU, etc etc etc. 'haps my GPU is faulty somewhat but I failed to achieve any sort of crash/freeze/blackout during any load at my OC profile so I'm left with a lot of unanswered questions.

The only time I've had my settings randomly reset on me on boot when I had an unstable overclock. Usually when my voltage was too low.

The only time I've had my settings randomly reset on me on boot when I had an unstable overclock. Usually when my voltage was too low.

It was always perfectly stable after boot, but something about boot with low voltages causes system instability during post, etc.

I had the same problem with my card before I sent it to service. I already created this topic for this reason. They decided to change and gave me a zero card. There is no problem for now, but I feel like I will have the same problem again in the future.

I looked back and realized what was going on and sounds like your original card did have the same issues ours due which is super common for this gen for some reason and even the 7000 I think. It was just killing me not being able to find a fix online and when I figured it out I just wanted to share. Some other people could do the same that don't feel like RMA their cards. Mine's well within warranty I'm just impatient and don't like people touching my stuff, lol. But looks like your replacement card is more of a winner, but my temps now are still better than yours! Lol. I've been posting on random forums to try to share. Specially since some of the manufacturers are telling their customers that this is normal behavior and they don't need to send their cards in. It's not an AMD problem they made the silicon just fine, I think it's the board partners not making sure the cold plate makes correct contact. Could you imagine if the board partners had to open all of their cards to double check and adjust for this? They wouldn't make any profit so they're just telling people normal behavior or by design. I'm pretty disappointed with sapphire right now I thought they were one of the better ones which is why I chose their brand...

 

[torkema61]

The only time I've had my settings randomly reset on me on boot when I had an unstable overclock. Usually when my voltage was too low.


It was always perfectly stable after boot, but something about boot with low voltages causes system instability during post, etc.


I looked back and realized what was going on and sounds like your original card did have the same issues ours due which is super common for this gen for some reason and even the 7000 I think. It was just killing me not being able to find a fix online and when I figured it out I just wanted to share. Some other people could do the same that don't feel like RMA their cards. Mine's well within warranty I'm just impatient and don't like people touching my stuff, lol. But looks like your replacement card is more of a winner, but my temps now are still better than yours! Lol. I've been posting on random forums to try to share. Specially since some of the manufacturers are telling their customers that this is normal behavior and they don't need to send their cards in. It's not an AMD problem they made the silicon just fine, I think it's the board partners not making sure the cold plate makes correct contact. Could you imagine if the board partners had to open all of their cards to double check and adjust for this? They wouldn't make any profit so they're just telling people normal behavior or by design. I'm pretty disappointed with sapphire right now I thought they were one of the better ones which is why I chose their brand...

Unfortunately, I too was disappointed with Sapphire. They told me that there is no problem as long as the temperature is below 115°C, but running it at these temperatures for a long time will of course reduce the life of the GPU. Frankly, I am very worried about this.

Additionally, I was complaining about GPU fan speed and they blamed AMD software for that. After about 2 months of effort, I learned that the problem was not in the AMD software, but in the RPM speed of the Sapphire graphics card.

The customer representative responds very quickly but gives reluctant answers.

 

[ChrisM255]

Unfortunately, I too was disappointed with Sapphire. They told me that there is no problem as long as the temperature is below 115°C, but running it at these temperatures for a long time will of course reduce the life of the GPU. Frankly, I am very worried about this.

View attachment 329032

Additionally, I was complaining about GPU fan speed and they blamed AMD software for that. After about 2 months of effort, I learned that the problem was not in the AMD software, but in the RPM speed of the Sapphire graphics card.

View attachment 329033

The customer representative responds very quickly but gives reluctant answers.

You know my settings were constantly resetting in adrenaline to, the slider would randomly set to 100% on the fan or sometimes zero percent and I'd have like a 50° Delta but I chalked it up to my overclock but I believe it was due to the improper cold plate contact and during boot it was seeing bad things causing it to glitch up the settings or something. Since I have done this PHYSICAL fix I have had zero GPU setting/software issues and I did not DDU or do anything else to my system since this fix. Not even a driver update. So they are full of s***...

 

[VT1185MF_Rosaria]

As far as what I have seen on Multiple AIB reviews of RDNA2 cards, the Sapphire coolers are in the hotter end of the spectrum when it comes to temperatures.

(Sources for these are TPU reviews themselves)

Keep in Mind that although these results are from the more expensive Nitro+ Design, it should prove to be a good baseline as the Pulse is the cheaper version of the Nitro cooler design.

You can also see in internet posts such as on reddit etc. that users are experiencing large temperature deltas from core to junction in these specific AIB versions of the 6700XT.

 

[ChrisM255]

Undervolt it. I run my Powercolor Fighter with these values.

View attachment 328956

Curiosity killed the cat. Tried these settings and with the Kryosheet system and got 72°C with 77°C hotspot, custom fan curve but with the temps it shouldn't have gone above 45-50% fan speed. This was a 2 hour gaming run tracking with hwinfo64.
Gotta say I thought this would've been hotter and before the mod the card wasn't stable below 1180mv, just thought I had a dud card...
The horrible thermals were making my card wildly unstable compared to what I am seeing now

From the factory I could not go higher than 2550 min freq and could not go lower than 1180mv without crashing or stutters

Now it is testing stable at 2650 min freq and 1030mv, have not tested further yet but I will get bored eventually and start tinkering again

 

[Chrispy_]

Does that sometimes but it takes literally only maybe a second or two that I load the OC profile.

Yeah, it resets if Windows isn't shut down cleanly. It doesn't have to be a driver crash - AMD are just using the "did windows fail to shut down properly" as their trigger for resetting to safe values - and that makes perfect sense, since an unstable overclock/undervolt could easily make a machine non-bootable.

 

[ChrisM255]

Yeah, it resets if Windows isn't shut down cleanly. It doesn't have to be a driver crash - AMD are just using the "did windows fail to shut down properly" as their trigger for resetting to safe values - and that makes perfect sense, since an unstable overclock/undervolt could easily make a machine non-bootable.

I experienced this with my AM4 B450 system
For me, I narrowed it down to my negative voltage offset in addition to my curve optimizer settings or if I set the optimizer too aggressive without an additional offset, when both were applied it would reset my GPU settings CONSTANTLY! Was always stable while in windows, but had to re-apply my GPU settings on almost each boot. I assumed it was unstable voltages at some phase of the shutdown. Originally I thought it was something I had set bad on the GPU, but tested this by only changing the fan curve and I made it more aggressive and would still reset to default wattman settings, then I knew it was the system, not the GPU.
System never reported it either. Feel like this should've been a thing... unstable ANYTHING, and I would want the system to throw up a notification! POST screen, something.
Frustrating(not the word I wanted to use) that the GPU can detect that there was an issue with WINDOWS shutdown, but WINDOWS doesn't report it to you. AT ALL!

 

[67Elco]

In my experience the Adrenaline software is garbage...does what it wants no matter how many times you manually input and save. I now just load the driver only and use Afterburner.

 

[CodeError]

PowerColor Red Devil RX 6700 XT - tested on 2K monitor with Virtual Super Resolution ON in AMD driver settings.

Current temperature in my room is 15C. In summer my hotspot reached 100C in some games (uncapped FPS). After that I switched GPU BIOS from default OC mode to Silent mode and put a stronger fan curve. Also between Summer and this I put some additional fans in case and better cooler on CPU. In summer CPU (5600X) was reaching 85C and with new cooler it is now around 65C. That should have positive effect on GPU temperature in some way.

How much lower temperature could I expect with undervolt?