RX 480 4GB Throttling at 60 Degrees Celsius

[flipped_bit]

I have an MSI RX 480 Gaming X 4GB card.  In the past, its load temperature has been in the low to mid 60s (at ~1,250rpm on the fans).

 

All of a sudden it won't go above 60 degrees.  Either the GPU clock or GP utilization will fall as it approaches 60 degrees.  Undervolt/overvolt, lower/higher power limit, nothing I try will get it over 60.  This has led to a roughly 10% decrease in performance - not the end of the world, but annoying (especially since I don't know why it's happened).

 

I've un-installed Afterburner and nuked the system with DDU a couple times.  I haven't re-installed Windows yet, but I may have to try that next.  

 

I'm using the latest Adrenalin 2020 package (19.12.3).  I've gone back and tried older drivers where I had no issue and the problem persists.

 

I have read online about some people that have experienced this kind of problem (temperature based throttling at low temps) and they fixed it by re-applying thermal paste.  This doesn't make sense to me - if the thermal paste was bad, wouldn't the GPU core temp be well over 60 degrees?  

 

I thought perhaps my fan curve is too low, and performance does increase somewhat if I ramp the fans up at lower temps (but not to its previous levels before the problem emerged).  Is it possible something else on the card is responsible for throttling the core?  I don't think this is likely - my card has a 0 fan speed mode up to 60 degrees, so it's designed not to have any active cooling until it hits 60 degrees.  

 

Any ideas would be most welcome.

[muito_gostoso]

check your power profile if you are in high performance, reciently i found windows have bugs with this , go to regedit, an in current control----power ---- cs and set 0, to fully activate high performance, also in adrenalin driver set in power options high performance all tabs to 100% in any case. finally install msi afterburner and set onscreen stats of gpu load and stres the gpu with bench integrated in gpuz to watch the load over time and performance if you hit in 2-3 minutes 100% of your gpu usage you are OK.

 

P.D....intall windows 10 2004 compact build , in confirm that build have 0 issues of throtling, and also is very optimized , just 2 gb of disk usage installed.

[Mr.Meerkat]

Not 100% sure what's causing it but I know for sure that you used to be able to tweak the target temperatures in wattman and once after a driver update, it randomly changed the target temperature to 50C by itself so my RX470 started thermal throttling.

However, in my case, it only had a 6-pin and throttled due to power limitations anyway so the loss of performance wasn't that noticeable. 

[Vishera]

You can change it in MSI Afterburner as long as the AMD drivers don't break the compatibility with the tool.

[flipped_bit]

32 minutes ago, muito_gostoso said:

check your power profile if you are in high performance, reciently i found windows have bugs with this , go to regedit, an in current control----power ---- cs and set 0, to fully activate high performance, also in adrenalin driver set in power options high performance all tabs to 100% in any case. finally install msi afterburner and set onscreen stats of gpu load and stres the gpu with bench integrated in gpuz to watch the load over time and performance if you hit in 2-3 minutes 100% of your gpu usage you are OK.

 

P.D....intall windows 10 2004 compact build , in confirm that build have 0 issues of throtling, and also is very optimized , just 2 gb of disk usage installed.

I've always used the balanced profile with Windows 10.  I did update Windows to the November update recently - now I wonder if that somehow screwed things up.

[flipped_bit]

32 minutes ago, Mr.Meerkat said:

Not 100% sure what's causing it but I know for sure that you used to be able to tweak the target temperatures in wattman and once after a driver update, it randomly changed the target temperature to 50C by itself so my RX470 started thermal throttling.

However, in my case, it only had a 6-pin and throttled due to power limitations anyway so the loss of performance wasn't that noticeable. 

I don't recall ever being able to set a temp limit, but I do have a fan curve set in Adrenalin 2020 (Wattman is no more).  The fans are ramping up according to the curve I set, so I don't think there's any problem there.  I also tried using Afterburner to change some settings to try and push it beyond 60 degrees, but the problem still didn't go away.

[flipped_bit]

Just necro'ing this thread with a solution in case anyone finds it via search engine.

 

The card's performance got significantly worse since my original post, and since the card's warranty is up and it's next to impossible to get a next gen card now anyway, I decided to disassemble it to see what was going on.  Turns out there was next to no thermal paste left on like 3/4 of the GPU. surface area  Given I use this PC for like 2 hours a day on average, this was extremely disappointing/surprising.  I guess I won't be buying an MSI card again if they skimp on the paste and/or do a poor job of applying it.

 

Anyway, I put a generous helping of MX-4 on there, and now she's humming like new.  Load temps back to low 60s, and idle temps at ~20.  And high 60s when I put on a high overclock.  Very pleased with this result.

 

I would guess the reason my card was not going above 60 is because one or more parts of the die were well in excess of 60 due to lack of thermal paste, and the temperature sensor did not capture this.  

 

So if your GPU is stuttering and you're getting low temperature readings, it could be the thermal paste!