Small update in this morning.
First of all I'm dumb as brick. I put the Noctua fan wrong when I dismounted cooler. It was blowing in case.
Now I put another Noctua as intake fan and tweaked a little fan curves (in idle CPU fan has 1000 RPM in idle and intake/exhaust 800 RPM and in load all 3 are at +1300 RPM).
Unfortunately I cannot put second Noctua fan on cooler because RAM clearance. Maybe I'll do some DIY with a 120 fan someday.
Room temperature is the same 22°C .
Something has changed.
Idle temps are down in 30's°C
View attachment 141861
And load temps are 79°C (Small FFT's for almost 10 minutes).
View attachment 141862
At least 7°C showed both in idle and load and looks like PPT/TDC/EDC are up too. Me happy now.
But I want to go further. In my youth I was a big fan of underclock/undervolt. Now underclock doesn't bother me so much but I want to try to undervolt.
Unfortunately I have more than 10 years since I didn't touch an AMD rig so I have no clue about how to do it on Ryzen.
So bear with me and please give me some directions for undervolt this CPU.
Thanks again.
Thats better...
If you want to just under volt you can try with a negative offset to Vcore. I'm not familiar with Asrock BIOS but somewhere there is the voltage control (Vcore) of CPU on auto mode and the offset (Dynamic Vcore) on auto as well.
You must set CPU Vcore to normal and then give the Dynamic Vcore some negative value like -0.00625V or -0.01250V or more.
The -0.00625 is the smallest step.
Some CPU can sustain same boost clocks even with -0.02500V or -0.05000V. My CPU is not so great in terms of silicon quality and looses performance(clock) even with -0.01250V.
Your 3600X is definately better silicon and you may achieve some further temp reduction without loosing clock/perf by undervolt.
But I dont have high temps anyway (max60~65C and 50~55C gaming) so it didnt matter to me that I couldnt use undervolt without loosing performance. So I tryied to do the opposite, and see if I can gain some clocks. Not with positive Dynamic Vcore but with an entirely different approach.
Through manual PBO settings of PPT/TDC/EDC and Scalar.
R5 3600 PBO limits/defaults
AutoPPT
88W (Package Power Tracking)
AutoTDC
60A (Thermal Design Current) = max current on thermally constrained scenarios ~95C)
AutoEDC
90A (Electrical Design Current)
Auto Scalar X1
I was watching with HWiNFO those numbers during CB-R20 and it was like this:
PPT
87.5W
TDC
49.1A
EDC
78.5A
CPU avg(column) average effective clock = ~
3971MHz
Because the temp was good enough (~63C) the CPU was doing all it could and it was capped by PPT limit of 88Watt. Raising the PPT limit alone does not help. It was still around 88W. So I tried to lower the current (EDC) starting with 73A.
Next R20 run:
PPT
87.5W
TDC
49.1A
EDC
73A
CPU avg(column) average effective clock = ~
3972MHz
Keep reducing EDC until 65A.
Next R20 run:
PPT
89W
TDC
49.7A
EDC
65A
CPU avg(column) average effective clock = ~
3985MHz
Then PBO Scalar into the game...
Next R20 run:
PPT
91W
TDC
49.7A
EDC
64A
Scalar
X2
CPU avg(column) average effective clock = ~
4000MHz
Lowering EDC alone has raised a bit the Vcore by it self. Setting Scaler from auto to X2 raised it a bit more so thats why the increased clock. Temp went up 3~4C.
Reducing EDC to 63A and maximizing cooling (AIO extreme profile and TIM change to liquid metal = drop 6~7C) has benefit even more the clock.
Next R20 run:
PPT
93.7W
TDC
51.3A
EDC
63A
Scalar
X2
Max cooling
CPU avg(column) average effective clock = ~
4033MHz
Your CPU has the following limits.
AutoPPT
128W (Package Power Tracking)
AutoTDC
90A (Thermal Design Current) = max current on thermally constrained scenarios ~95C)
AutoEDC
125A (Electrical Design Current)
Auto Scalar X1
It appears that you dont hit any of those limits because even tho you drop the temp, its still high enough for it to not raise clock/voltage by it self.
These limits are monitored by the CPU FIT controller (silicon FITness controller) and are regulated in conjunction with temperature.