Can I ask what was the max voltage applied by CPU power management running prime95 (custom 128K FFTs) when the CPU was under full auto boost settings?
I just re-run it for the sake of getting the baseline straight. Without PBO at full stock settings, the voltage settles at 1.312-1.344V (SVI2) under full load (Prime95 128K FFTs AVX 48 threads) with clocks around 4120MHz. With PBO enabled, voltage is mostly 1.381V with dips occasionally to 1.369V and spikes to 1.4V. Clocks are around 4170MHz.
Since I installed the new BIOS, I thought to revise my overclock and managed to get an additional +25/+50MHz on each CCX and also was able to lower voltages. VCore is set to 1.350V in BIOS and droops to around 1.312V/1.319V (acording to the SVI2 reading). Temps are below 80°C then hovering between 70 and 78 most of the time during Prime95 AVX load.
I guess I'll leave it at that for now with clocks of: 4425; 4425; 4375; 4350; 4450; 4425; 4450; 4425 MHz. This is good for +15.000 CB20 scores.
whether overclocking should be manual, but I insist pbo is just for threadripper power management
In my experience PBO doesn't do much in terms of clock speed increase. What it does is increasing power draw without substantially increasing performance. It also sets PPT to 600W, TDC to 800A and EDC to 960A.
Its a shame you could not fine-tune PBO and use it as your overclock. Retaining boost behavior to get high single core clock speeds while lowering clocks on multi-core workloads would be highly appreciated. Sadly AMD did not intended it this way.
Undervolt and overcurrent is my motto.
You mean keeping voltage as low as possible for a given clock? That's what I intend to do. But you only have vCore and clock to change, current comes only as a result of that setting, it's nothing you can exchange willingly?! Given that, a desired clock speed demands a certain voltage to run stable at, you can't tell the CPU to magically draw more current to get a clock speed stable.
Have you consider degradation?
I dont really know, just asking...
Yes, I'm working on incentivize myself to upgrade to a Zen3 Threadripper when they come out.
I think you would benefit from saving more than 100w(160>52@3.3GHz example). If you consider your whole chip a 2 parts 1800X, take the lower bin CCX'es for a lower limit and open up more room for better clocking ones. Otherwise, you are TDP limited(280W÷1.31=213A, TDC:215A at stock) let alone EDC. At TDC current, you are temperature limited - it overrides everything, you need to be below a certain temperature threshold for it to untrigger 'FID' the master power switch.
If you want to reach peak current limits, you need to save on average consumption. Even if you cut down half of the cores on power(the way you do this is via the voltage setting) you leave the rest to EDC limits. Notice if you had throttled them with EDC instead, neither the good cores or the bad would overclock - they would all clock the same. So, there are ways to go about it. Be safe and smart: maximise EDC, but lower voltage and PPT wherever you may.
I guess I can not follow you there. Dynamic boosting is only possible without PB/PBO as I understand it. As soon as I type in stuff manually, clocks are fixed (perCCX). Maybe that wasn't the case with the the 1800X, but with the 3960X it sure is. As said earlier, PBO does not give anything worthwhile in a performance uplift justifying the power draw it produces.
