The RX 6000 series Owners' Club

[lunatik]

Are there any simple (not undervolt) ways to lower junction temp on 6700xt(sapphire nitro)? Problem being 55° on normal temp but hitting 100° on junction and crashes..seems like an issue with thermal pads or smth? Never really opened a gpu before. Or it's just normal with air cooled gpu?

 

[GamerGuy]

Are there any simple (not undervolt) ways to lower junction temp on 6700xt(sapphire nitro)? Problem being 55° on normal temp but hitting 100° on junction and crashes..seems like an issue with thermal pads or smth? Never really opened a gpu before. Or it's just normal with air cooled gpu?

How's case airflow for your rig? Also, have you tried adjusting the fan curve via Adrenalin CP? I have the Nitro+ RX 6900 XT and don't have an issue with heat with my fans ramped up (only if you don't mind the fan noise, which isn't that bad for me as my rig's on the floor)

 

[Butanding1987]

Got my self a 6900xt LC edition and found out that the original bios with 2310mhz memory clock has horrible timings so my points on 3dm were awful
i flashed with asus LC edition and go these results
i limite the card to 355A and 402w
it stays arround 2750mhz during the test
if i increse the power draw i can go beyond but i dont really wanna break do card

I scored 19 186 in Time Spy

AMD Ryzen 5 5600X, AMD Radeon RX 6900 XT x 1, 16384 MB, 64-bit Windows 11}

www.3dmark.com

It’s pointless to buy an expensive high-end card like that if you‘re not brave enough to push it. My 6800 XT is not far behind your 6900 XT.

 

[Valantar]

Are there any simple (not undervolt) ways to lower junction temp on 6700xt(sapphire nitro)? Problem being 55° on normal temp but hitting 100° on junction and crashes..seems like an issue with thermal pads or smth? Never really opened a gpu before. Or it's just normal with air cooled gpu?

That's definitely not normal. The TPU review showed <20°C between the edge and hotspot temperatures. Either it's poorly mounted, has poor a poor thermal paste application, or has some sort of production flaw (a non-flat cooler surface, for example). My first suggestion would be to RMA it, though given the current situation that might leave you without a GPU for a while. After that, open it and check for thermal paste coverage and contact (i.e. how thinly squeezed the paste is across the die). Mounting pressure is often an issue with AMD cards for some reason, and getting some thin plastic washers and putting them between the screws on the back of the PCB (the four around the die specifically) and the PCB can alleviate this - but then you also need to be careful, as too high pressure can damage the card. The safest route is an RMA.

 

[lunatik]

How's case airflow for your rig? Also, have you tried adjusting the fan curve via Adrenalin CP? I have the Nitro+ RX 6900 XT and don't have an issue with heat with my fans ramped up (only if you don't mind the fan noise, which isn't that bad for me as my rig's on the floor)

Phanteks p500a, 3x 140mm front intake, 1x 140mm rear exhaust, 1x 140mm top exhaust and 1x140mm fan over ram. Gpu runs at 2780-2880mhz, 1110mv set in adrenalin (this is with junction temp 95-98 under full load). Just looking for ways to push it further, since lowering voltage would make it crash or using higher voltage would mean junction temp getting too high.

 

[ratirt]

Phanteks p500a, 3x 140mm front intake, 1x 140mm rear exhaust, 1x 140mm top exhaust and 1x140mm fan over ram. Gpu runs at 2780-2880mhz, 1110mv set in adrenalin (this is with junction temp 95-98 under full load). Just looking for ways to push it further, since lowering voltage would make it crash or using higher voltage would mean junction temp getting too high.

the 140mm fan over ram may disrupt the proper airflow in your case. My 6900xt never hits 100c but i can double check after playing for a while what temp it's going to show.

 

[Valantar]

the 140mm fan over ram may disrupt the proper airflow in your case. My 6900xt never hits 100c but i can double check after playing for a while what temp it's going to show.

Airflow issues will not cause a 45-degree delta between edge and hotspot temperatures. Airflow issues will affect all temperatures roughly equally. There is something wrong with that cooler or its mounting, period.

 

[lunatik]

the 140mm fan over ram may disrupt the proper airflow in your case. My 6900xt never hits 100c but i can double check after playing for a while what temp it's going to show.

Yeah, might be..thought about it too while typing. Altough when i pushed this card to its current "max" limits, i didnt have that fan over ram. There's also my 5600x, running on a be quiet tower, cant remember the name, 1 fan tho. I'll test if it has any impact on temp with or without later tonight.

 

[Yraggul666]

HELL->o! I finally joined the club.

 

[lunatik]

Airflow issues will not cause a 45-degree delta between edge and hotspot temperatures. Airflow issues will affect all temperatures roughly equally. There is something wrong with that cooler or its mounting, period.

I'll also try to tighten the screws. Don't really care that much about warranty, i'm just bored out of my mind with another lockdown. And if needed, i'll order thermal paste and pads too.

 

[Valantar]

I'll also try to tighten the screws. Don't really care that much about warranty, i'm just bored out of my mind with another lockdown. And if needed, i'll order thermal paste and pads too.

Most likely the screws can't be tightened sufficiently on their own, and overtightening can risk stripping or shearing the screws, in which case you're up that proverbial creek without a paddle. If you don't have any nylon washers handy, try cutting some yourself from some stiff plastic? If not, any decent hardware store should have nylon washers. 1mm thick should be plenty.

 

[ratirt]

Airflow issues will not cause a 45-degree delta between edge and hotspot temperatures. Airflow issues will affect all temperatures roughly equally. There is something wrong with that cooler or its mounting, period.

Junction temps can go high on 6000 series. The 100c is quite extreme but you can't disregard all other options as a reason for that. He has this card overclocked so it is normal that from Idle to full load the temp will go up rapidly and high. If the hot air is not being removed with a certain pase due to disturbance in the airflow it may cause overheating over time and rise of temperature.
The difference between the core temp and the junction can reach these temp differences. Especially if the airflow is not OK. He has an air cooled card and this can happen. It is different with your case having a liquid cooled card which probably has a full pcb cover plate.
I'm not saying the mounting is all good but worth to try anyway the easy things and see if it improves things. I've got a Vega 64 and the difference between junction and core were 40 degrees sometimes.

He can try removing case side cover and see if the temps are as high. If it will not reach 100c as with the closed case, it is an airflow. Mounting might help as well but lets focus on the easy stuff before we start stripping card apart, which may not be the problem.

 

[Valantar]

Junction temps can go high on 6000 series. The 100c is quite extreme but you can't disregard all other options as a reason for that. He has this card overclocked so it is normal that from Idle to full load the temp will go up rapidly and high. If the hot air is not being removed with a certain pase due to disturbance in the airflow it may cause overheating over time and rise of temperature.
The difference between the core temp and the junction can reach these temp differences. Especially if the airflow is not OK. He has an air cooled card and this can happen. It is different with your case having a liquid cooled card which probably has a full pcb cover plate.
I'm not saying the mounting is all good but worth to try anyway the easy things and see if it improves things. I've got a Vega 64 and the difference between junction and core were 40 degrees sometimes.

He can try removing case side cover and see if the temps are as high. If it will not reach 100c as with the closed case, it is an airflow. Mounting might help as well but lets focus on the easy stuff before we start stripping card apart, which may not be the problem.

No. TPUs review of that specific GPU, the 6700 XT Nitro+, which I linked above, showed a <20-degree delta. Different workloads can affect the deltas differently, but seeing significantly more than that is a clear indication that there is something wrong with the cooler. A near 50-degree delta is not normal under any circumstance for any GPU, period. GPUs generally spread their heat generation across the majority of the die, which leads to relatively even thermals. A 20-degree delta is fine, 30 is kind of poor but acceptable, 50 is borderline dangerous.

For reference: my deltas with the LDU are not particularly good - depending on the load they often exceed 20 degrees, and I've seen it creep towards 30. This is typical of PowerColor, who tend to have questionable QC on their cooler mounting. As the absolute temperatures are low it ultimately doesn't matter, but with a delta as high as what is described here, you essentially have zero thermal headroom for the die. That is a bad situation to be in.

Besides, if airflow was the issue, there's no way the edge temperature would be 55°. The die is not large enough that airflow over different parts of the cooler will affect different areas of the die significantly differently. And besides, the heatpipes run across the length of the GPU, meaning that for your scenario of airflow affecting things to be true, airflow would need to be selectively massively worse for a single heatpipe. That is, well, not possible, when you have three fans blowing directly across them. There's also a copper plate between the heatpipes and die. If airflow was poor, the cooler would be unable to dissipate its heat into the air, leading to the cooler warming up, in turn leading to the entire die warming up. So, there is no realistic scenario in which poor airflow will selectively make one part of the die drastically hotter than another. Poor contact or flatness is the only reasonable explanation, period.

Also, Vega was notorious for poor cooler fitment due to the height differences between HBM and the GPU die. Using that as a reference point for anything is a bad idea.

 

[ratirt]

No. TPUs review of that specific GPU, the 6700 XT Nitro+, which I linked above, showed a <20-degree delta. Different workloads can affect the deltas differently, but seeing significantly more than that is a clear indication that there is something wrong with the cooler. A near 50-degree delta is not normal under any circumstance for any GPU, period. GPUs generally spread their heat generation across the majority of the die, which leads to relatively even thermals. A 20-degree delta is fine, 30 is kind of poor but acceptable, 50 is borderline dangerous.

For reference: my deltas with the LDU are not particularly good - depending on the load they often exceed 20 degrees, and I've seen it creep towards 30. This is typical of PowerColor, who tend to have questionable QC on their cooler mounting. As the absolute temperatures are low it ultimately doesn't matter, but with a delta as high as what is described here, you essentially have zero thermal headroom for the die. That is a bad situation to be in.

Besides, if airflow was the issue, there's no way the edge temperature would be 55°. The die is not large enough that airflow over different parts of the cooler will affect different areas of the die significantly differently. And besides, the heatpipes run across the length of the GPU, meaning that for your scenario of airflow affecting things to be true, airflow would need to be selectively massively worse for a single heatpipe. That is, well, not possible, when you have three fans blowing directly across them. There's also a copper plate between the heatpipes and die. If airflow was poor, the cooler would be unable to dissipate its heat into the air, leading to the cooler warming up, in turn leading to the entire die warming up. So, there is no realistic scenario in which poor airflow will selectively make one part of the die drastically hotter than another. Poor contact or flatness is the only reasonable explanation, period.

Each computer configuration is different and you don't know what's his. Worth to check all but simple thing goes first.
He can open his case and check.

Also, Vega was notorious for poor cooler fitment due to the height differences between HBM and the GPU die. Using that as a reference point for anything is a bad idea.

Vega was just an example. I have a 6900XT Air cooled and I know what happened when I've had a bad airflow and what happens when i open my case.

 

[Valantar]

Each computer configuration is different and you don't know what's his. Worth to check all but simple thing goes first.
He can open his case and check.

Vega was just an example. I have a 6900XT Air cooled and I know what happened when I've had a bad airflow and what happens when i open my case.

Yes, the temperature of the entire GPU changes. That's what airflow does. If changing airflow dramatically changes the delta between edge and hotspot temperatures for you, there is something very wrong with your cooler as well.

But, just to check, can you open HWInfo64, double click the GPU temperature and GPU Hotspot temperature readings to open the graph view for both, run a load on that 6900XT for a few minutes until it stabilizes, and then open your case and let it run a few minutes more, then screenshot the graphs so we can see the change? You're making me curious here.

 

[ratirt]

Yes, the temperature of the entire GPU changes. That's what airflow does. If changing airflow dramatically changes the delta between edge and hotspot temperatures for you, there is something very wrong with your cooler as well.

He's got a fan blowing on ram. this may screw up the airflow. Let him try things.

But, just to check, can you open HWInfo64, double click the GPU temperature and GPU Hotspot temperature readings to open the graph view for both, run a load on that 6900XT for a few minutes until it stabilizes, and then open your case and let it run a few minutes more, then screenshot the graphs so we can see the change? You're making me curious here.

I got a good airflow after upgrading so there's no need for that. My GPU doesn't hit 100c junction. But it did get hot some time back when airflow was bad.

 

[Valantar]

He's got a fan blowing on ram. this may screw up the airflow. Let him try things.

A fan blowing on the RAM will either change the airflow patterns for a large part of the GPU, or not at all. Either way, it will not cause high hotspot temperatures while maintaining low edge temperatures. This is simply not possible due to how air coolers work.

I got a good airflow after upgrading so there's no need for that. My GPU doesn't hit 100c junction. But it did get hot some time back when airflow was bad.

I'm not asking you to test it for your own sake, I'm asking you to test it to corroborate your own statements. If you can't do that, then I have no other choice than to veer on the side of logic assume that you've either misremembered or otherwise misidentified the dynamics of your cooler, because there is to my knowledge no way a change in airflow can affect hotspot temperatures significantly more than edge temperatures. So, unless you can actually show some data indicating otherwise, I'll keep assuming that those two temperature readings will change roughly in sync as the environment around them changes. Now, I know the delta is generally smaller at lower loads - my 6900 XT has much higher deltas if I run it at full power than in my ~190W UV profile (~10-15 degrees with UV, mid-20s without), but that's with a relatively poorly mounted cooler from a manufacturer known for this. And, of course, the <20-degree deltas in the TPU review are under full load. Even if some other load triggers warmer hotspots than what they saw, anything above 30 would be deeply worrying in light of that. And that's the point here: some change in the delta across a range of loads is normal. A 45-degree delta is not normal for any card with a well-functioning cooler, period, regardless of environmental factors.

 

[Felix123BU]

If you have a 40-50c difference between gpu temp and hotspot, that is 99% a case of bad mount and uneven pressure.
On another point, a 100c on the hotspot for a OC-ed card on air is really not that bad. I wonder if the difference between the 2 temps is correct, in cases of bad mounts and parts of the GPU not being cooled properly, you would get a lot more than 100c on the hotspot.

 

[ratirt]

I'm not asking you to test it for your own sake, I'm asking you to test it to corroborate your own statements. If you can't do that, then I have no other choice than to veer on the side of logic assume that you've either misremembered or otherwise misidentified the dynamics of your cooler, because there is to my knowledge no way a change in airflow can affect hotspot temperatures significantly more than edge temperatures. So, unless you can actually show some data indicating otherwise, I'll keep assuming that those two temperature readings will change roughly in sync as the environment around them changes. Now, I know the delta is generally smaller at lower loads - my 6900 XT has much higher deltas if I run it at full power than in my ~190W UV profile (~10-15 degrees with UV, mid-20s without), but that's with a relatively poorly mounted cooler from a manufacturer known for this. And, of course, the <20-degree deltas in the TPU review are under full load. Even if some other load triggers warmer hotspots than what they saw, anything above 30 would be deeply worrying in light of that. And that's the point here: some change in the delta across a range of loads is normal. A 45-degree delta is not normal for any card with a well-functioning cooler, period, regardless of environmental factors

I had that problem with airflow. I'm not testing anything but rather speaking from experience. I had a problem with airflow which caused the temps hot spot go wild and that was my advice for one of the members. Check the airflow and see if you can adjust something since maybe something is wrong. Maybe heat-sink needs adjustments but it doesn't have to be that either. I dont know what set up he has exactly so speaking it is this or that will not help him cause there's so many different things that can affect the temp.
Dude, I dont want to argue with you about this nonsense. Is there really a need for me to prove that proper airflow in the case matters? Airflow in the case matters and you will not convince me otherwise. I'm sure there are people who will tell you the same thing.
What I can tell you is, when I have upgraded my fan system, the heat coming out of the case was tremendous. Especially if you run your card at full speed for a while.

 

[lunatik]

So, i did a fresh windows install and also re-installed all the drivers. Put everything in stock/default.
Gpu temp 63, hot spot 97 (front and top silent cover removed, also the ram fan)
Gpu temp 62, hot spot 96 (+ the side panels removed)

Edit: or maybe my testing methods are completely false for this? This was done in time spy stress test 20 loops..

 

[Valantar]

Dude, I dont want to argue with you about this nonsense. Is there really a need for me to prove that proper airflow in the case matters? Airflow in the case matters and you will not convince me otherwise. I'm sure there are people who will tell you the same thing.

I never said anything even remotely to this effect. I said poor airflow doesn't cause hot spot temps specifically to rise dramatically while edge temps are normal, as this is simply not how physics and air coolers work. The absolute worst case scenario for this is if the fins around one of the heatpipes got no airflow while the rest did, which would cause the die to selectively heat up, but, again, this isn't possible with how heatsinks are laid out. If all heatpipes get roughly equal airflow - which they will due to the fans mounted to the heatsink - then all heatpipes will dissipate roughly equal amounts of heat, and all areas of the die will get roughly the same cooling, barring unevenness in the mount, flatness issues in the cold plate, production errors causing poor heatpipe-to-cold plate contact, or other such issues. If one or more of the GPU fans gets too little cool air, is recycling hot air, or is obstructed to such a degree that it underperforms, then this affects all heatpipes roughly equally. The scenario you are describing does not make sense. That is why I'm asking you to provide some data. Not because I'm somehow denying that case airflow matters.

It's no wonder that a 330W+ GPU needs a lot of ventilation. What would be a wonder would be if adding more cooling affected hotspot temperatures significantly more than overall GPU temperatures.

So, i did a fresh windows install and also re-installed all the drivers. Put everything in stock/default.
Gpu temp 63, hot spot 97 (front and top silent cover removed, also the ram fan)
Gpu temp 62, hot spot 96 (+ the side panels removed)

Edit: or maybe my testing methods are completely false for this? This was done in time spy stress test 20 loops..

Unless you're controlling fan speeds you can't really compare the two, as the side panel removed test might have the GPU running at lower rpms. I'd still say >30 degrees indicates a poor mount or something similar, as that's an unusually high thermal delta. You also kind of need to monitor clocks, as GPUs these days adjust those very actively.

 

[lunatik]

Edit: discard all that, adrenalin just bugged out and didnt actually apply any settings global or warzone

Edit: i remembered that i had some thermal paste left from my previous computer and changed it on both cpu and gpu. From the same default/stock settings and time spy loop my gpu temp increased from 63 to 65 but hot spot decreased from 97 to 87.

Thank you all for quick replies and help

 

[ratirt]

Edit: discard all that, adrenalin just bugged out and didnt actually apply any settings global or warzone

Edit: i remembered that i had some thermal paste left from my previous computer and changed it on both cpu and gpu. From the same default/stock settings and time spy loop my gpu temp increased from 63 to 65 but hot spot decreased from 97 to 87.

Thank you all for quick replies and help

Gratz dude you got the thing sorted out Happy playing.

 

[Valantar]

Edit: i remembered that i had some thermal paste left from my previous computer and changed it on both cpu and gpu. From the same default/stock settings and time spy loop my gpu temp increased from 63 to 65 but hot spot decreased from 97 to 87.

Thank you all for quick replies and help

That's great! Good to hear it was a relatively easy fix.

 

[nolive721]

This is mine, one difference vs yours is that I set a 1800mhz min frequency because I noticed that in some less graphically intensive games, mainly older ones, a 500mhz frequency can cause very erratic framerates, as in the GPU tends to fluctuate a lot between 500 and 2200mhz in some old games where load is very low, and power draw does not differ much anyway if set at 1800 vs 500mhz for min limit.

View attachment 225054

thanks again because your post pushed me to revisit my UV and Oh boy its been rewarding in the last few days of testing

I have managed to decrease my Vcore to 1010mV and still be stable in Heavy Gaming at 2200Mhz core.
Power drains reduced to now 110W in worst case conditions (Simracing titles like PC2 or AC with heavy rain and again Graphics at Ultra), its just an mazing card

Now I am being offered the equivalent of $1000 for it (around $300 what I had paid for 2months ago) so I need to resist to the temptation