[SOLVED] 5950x temp decreasing when stress testing (high idle)

[anthonyhill86]

Hi Guys,

I have a bit of a weird one here, I have been scratching my head for a few days trying to solve it, but I am stuck and need some help.

Hardware:
I have just build a new PC with a 5950x on a asus rog strix x570-f mobo. with a corsair h100 cooler

Issues:
1. I have noticed the cpu sits at a high idle temp of approximately 70 degrees, I have checked this with several different tools and am getting a similar reading.
2. When I stress test the cpu, using several tools, the temp decreases?

I am using core temp to monitor the cpu temps; At idle it sits at 70 degrees. Then when I stress test it the temp will drop down to about 60 degrees. I know the lower temps isn't necessarily a bad thing, but things just don't add up here.

It's not like the pump or radiator fans are ramping up, I have them set to extreme all the time and I am using iCue to monitor the fans.

A few things that I have checked:

• Cpu pump is running 2500 to 2600 rpm constantly
• No change running radiator fans at 100% 2400rpm. 3 degrees less on the idle (67 deg)
• I've used several different programs to monitor the temp and stress test the cpu with consistent results
• I did consider maybe redoing the thermal paste, I used the original paste that came on the corsair cooler. But this wouldn't fix the decrease in temp, possible the high idle temp.
• Windows power options are set to balanced
• Returned Bios to original state and retested with no difference
• Bios idle temp is at 45 to 50 degrees (I'm not sure if that's package temp, I believe the other measurements are package cpu temp)
• cpu power at idle 75w
• cpu power at 100% stress 139w

- Another thing I have noticed when looking into ctr as suggested by one of the comments below is that 2 of the CCx temps are higher than the other 2. Does anyone have any idea what would be happening here?

Something more interesting, when I monitor the temps in the image below (ccx1 vs ccx2), I notice the actual temps I refer to are the temps in ccx1 (around 70 degreed). I really have no idea what the ccx actually means, maybe a part of the cpu the holds those cores? Anyway, when I run the stress testing the ccx 1 temps come down to match the ccx2, then they slowly increase as you would normally expect. Any one have any ides what's happening here.

I wonder if there is an air bubble in the cpu paste, or the paste not applied properly would cause this?

Idle

Stress test

fan speeds

 

[kerberos_20]

higher voltage in idle (for 4,6GHz boost) is whats rising temperature
btw 75watt doesnt looks like idle to me

 

[JohnBonhamsGhost]

that 4.6GHz(1.45v) vs the 4.1GHz(1.13v) would be what's making the difference.

shouldn't you be ramping down to ~800MHz at idle?
and idle would normally be less than 35\22\14% usage on those certain cores.
there must be something running in the background using a bit of resource vs just "idle".

 

[anthonyhill86]

I will have a look if anything is running but it’s a fresh install. So the only thing running would be the monitoring programs and what started up with boot.

is it normal for the voltage to react this way. It just seems like temp your move the other way. It deceases quick when first stress testing, then it will sill decrease but slow. I’ve never seen this, only the opposite with rising temps.

 

[Bouc]

Hi,
I have the same cpu but a msi x570 unify and a custom loop.
I had the same problem.
But using ClockTuner for Ryzen 2.1 rc5 I've been able to resolve this problem undervolting automatically and overclocking slightly and now my cpu seats at 35-36° idle and boost a little more but 49° when stressed

 

[anthonyhill86]

Hi,
I have the same cpu but a msi x570 unify and a custom loop.
I had the same problem.
But using ClockTuner for Ryzen 2.1 rc5 I've been able to resolve this problem undervolting automatically and overclocking slightly and now my cpu seats at 35-36° idle and boost a little more but 49° when stressed

do you have a capture available of your settings or profile so i can compare?

Thanks

 

[Bouc]

core temp and ctr

I did just follow the ctr guide

 

[Mr.Spock]

whatever software you're using is picking up the high temp of the "boosted" idle cores vs the lower avg temp of having all-cores running (lower voltage and clock). neither idle nor load temps seem to be anything to worry about. even on a new system you have background updates and Windows telemetry running so it's not exactly idle depending on SSD it may be HMB usage as well.

 

[iPeekYou]

The voltage for heavy multithreaded loads is much lower than "idle" or otherwise lighter, lightly threaded loads. Temps do correlate with voltage.

Maybe some reapplying of paste (in X pattern, preferably) and double checking mounting hardware will help. CCX 1 and 2 being hotter could be either due to that, or simply because background task is putting load on cores in those CCXes, where clocks jump instantly along with (reported) temps.

AFAIK, our monitoring softwares only report the highest jump in temps (and clocks) whereas AMD's hardwired control faculty does the monitoring in 1ms interval. Dropped clocks, high temps reported, etc due to missing information between the 1ms and whatever polling rate our software uses.