AMD views Ryzen 5000 CPU temperatures up to 95C as ‘typical and by design’

That's interesting, I had heard 90C for Zen3, as opposed to 95C for Zen2.

I'm still not comfortable with temps over 80C personally. Setting PBO to "motherboard" (essentially unlimiting power draw) hits 90C instantly under load, which is where it's capped by default.
 
My 5950X hit 85C under load in Cinebench r20, and I have a very robust watercooling system (triple 480mm radiators). However it never exceeded that threshold, and I don't really see a problem considering my performance was right in line with expectations (~10,950). Single core boosts up to 5GHz as well, but obviously that runs much cooler vs an all-core workload.

If it's truly by design then I don't see an issue.
 
My 5950X hit 85C under load in Cinebench r20, and I have a very robust watercooling system (triple 480mm radiators). However it never exceeded that threshold, and I don't really see a problem considering my performance was right in line with expectations (~10,950). Single core boosts up to 5GHz as well, but obviously that runs much cooler vs an all-core workload.

If it's truly by design then I don't see an issue.
85c with triple 480 rad?? I haven't seen 80c with my 360 Corsair capellix Aio. I would think you should be running waay more cooler then me. BTW-I am running stock though with no overclock, just letting the cpu boost itself. R20 score is only 9700 as well. Got pretty crappy Nanya corsair mem
 
85c with triple 480 rad?? I haven't seen 80c with my 360 Corsair capellix Aio. I would think you should be running waay more cooler then me. BTW-I am running stock though with no overclock, just letting the cpu boost itself. R20 score is only 9700 as well. Got pretty crappy Nanya corsair mem
I think even with watercooling, Cinebench is AVX loads right? I think it'll slaughter any CPU no matter what. What would be more interesting is if on a similar system with air cooling instead you saw lower clock boosts and such overall as AFAIK, the way that precision boost works is that it'll boost up to some limits, thermal being one of them so if you lower that, then everything else gets boosted.
 
My 5950X hit 85C under load in Cinebench r20, and I have a very robust watercooling system (triple 480mm radiators). However it never exceeded that threshold, and I don't really see a problem considering my performance was right in line with expectations (~10,950). Single core boosts up to 5GHz as well, but obviously that runs much cooler vs an all-core workload.

If it's truly by design then I don't see an issue.
Thanks for posting temps on a 5950! Running the Default Multi core cinebench r23 test just using PBO, I’m hitting 78 on mine with 28c coolant temp and 2 rads. Not sure if there is much difference between r20 and r23, but it’s at least a comparison point.
 
I really feel that the key variable missing here is time. To say that 90-95C is "typical and by design".... For how long? Minutes? Hours? Days? Weeks? Months? If I had a 5900X CPU at 90C for 24 hours a day in perpetuity, would I never experience a temperature related failure or even long-term degradation?
 
85c with triple 480 rad?? I haven't seen 80c with my 360 Corsair capellix Aio. I would think you should be running waay more cooler then me. BTW-I am running stock though with no overclock, just letting the cpu boost itself. R20 score is only 9700 as well. Got pretty crappy Nanya corsair mem

Thanks for posting temps on a 5950! Running the Default Multi core cinebench r23 test just using PBO, I’m hitting 78 on mine with 28c coolant temp and 2 rads. Not sure if there is much difference between r20 and r23, but it’s at least a comparison point.
I should note that I enabled Precision Boost and set the limits to "Motherboard". The CPU was pulling 100% of my motherboard's allowed EDC load (which is notably higher than AMD's default), so I suspect the temperatures were higher than they would have been at totally default settings or with the standard PBO limits. Coolant temp was ~27C.
 
Okay, just tested because I was curious.

5950X default settings
Cinebench R20: ~9700
Temp: Low 60s

5950X PBO enabled:
Cinebench R20: 10,950
Temp: 85C

So yeah, the CPU will pull significantly more power when PBO is enabled and power limits are raised, and performance goes up a lot as does temperature.
 
Last edited:
Okay, just tested because I was curious.

5950X default settings
Cinebench R20: ~9700
Temp: Low 60s

5950X PBO enabled, motherboard power limits:
Cinebench R20: 10,950
Temp: 85C

So yeah, the CPU will pull significantly more power when PBO is enabled and power limits are raised, and performance goes up a lot as does temperature.
Thanks for doing this Rizen. At least it validates that the 5950x seems to boost well and at stock, scores and the way the cpu boosts seems to be equal across chips
 
Made a tweak to my above post. Seems the default "Motherboard" and AMD PBO power limits are basically the same. Pretty sure I can manually raise it, but given temps and power draw I don't think that's a wise move. I'd rather allow the processor to manage itself.

In my testing, the Default AMD EDC limit is 140A. Precision Boost Overdrive EDC limit is 215A.

PBO will use 100% of that if temps are in check. So there is a very significant power increase and subsequent temp increase from enabling PBO, however in my case the multi-core performance improved ~12% which is notable.

YMMV based on chip, motherboard, cooling, etc :)
 
Made a tweak to my above post. Seems the default "Motherboard" and AMD PBO power limits are basically the same. Pretty sure I can manually raise it, but given temps and power draw I don't think that's a wise move. I'd rather allow the processor to manage itself.

In my testing, the Default AMD EDC limit is 140A. Precision Boost Overdrive EDC limit is 215A.

PBO will use 100% of that if temps are in check. So there is a very significant power increase and subsequent temp increase from enabling PBO, however in my case the multi-core performance improved ~12% which is notable.

YMMV based on chip, motherboard, cooling, etc :)
EDC limit on my hero viii is 200A. My clocks are sitting right around 4.6ghz. Usually pops between 4610 and 4601. Wonder if I could get a little more edc room, or if that just the limit on my MB.
 
Oh yeah putting PBO on motherboard is a substantial power draw increase. It immediately makes my 5950X hit 90C under load, but I'm on high-end air, not high-end water like you guys.
 
Anyone know what setting a negative offset in the curve optimizer actually does? Is it more voltage? Less voltage? It didn’t seem to have any effect to go from 0 to -10 temp wise for me, but I picked up 50-100mhz when I set it.
 
Yes, each step reduces voltage by between 3 and 5 millivolts, depending on load. 3mV at high load and 5mV at low. So -10 is -0.03V to -0.05V.
 
Yes, each step reduces voltage by between 3 and 5 millivolts, depending on load. 3mV at high load and 5mV at low.
So a negative offset reducing voltage also reduces total power draw, which in turn gives PBO a little more room to boost. Looks like the ideal method is then to reduce offset until it becomes unstable?
 
Yes that is exactly correct.

If you want to get fancy, try reducing your best cores (stars and dots in Ryzen Master) by less than your worst cores. The bad cores aren't used in lightly-threaded work so they'll never boost high and won't require as much voltage.
 
I was trying a lot of things as combinations, the key for me was when I touch to lock my CPU temp up to 75C.
Near everything on Auto + 110% system and Extreme (DIGI+VRM menu)

cinebench-23-12000.png
Super-PI5-54.png

CO is crap for me, it make my system unstable even with -5.
 
Last edited:
Back
Top