Configurable curve offset value for non-tested cores when using setVoltageOnlyForTestedCore = 1

[mkoehmst]

Currently the non-tested cores are all set to 0, but having a configurable option might be useful.

For instance, none of my tested values are under -25 even after a good bit of testing. So I could set all the non-tested cores to say, -15, and not have any concerns of non-tested cores being the source of instability.

That would lower the overall Vcore and heat and maybe allow the core being tested to hit higher frequencies?

Or maybe a slightly more sophisticated option would be a non-tested offset option. Set it say 10 or 15, and all the non-tested cores would be set +10 or +15 from their current test curve value. That would lower Vcore even more while still giving a big margin to alleviate concerns that non-tested cores are causing instabilities.

This offset approach would be especially useful for dual-CCD Ryzens because CCD1 runs at higher voltages and thus can take higher curve values. Like right now all my CCD0 offset are in the -25 to -35 range, but CCD1 is in the -40 to -49 range. So even with a flat non-tested curve value of -15, that's still a lot of extra voltage compared to -30 to -39 with a +10 non-tested offset.

A flat 0 curve value for non-tested cores is good for the first stages of testing when you don't have any idea what range of curve values you can use, but once you start dialing in rough values after initial testing, having an offset instead would help in nailing down those final, truly stable values.

[mkoehmst]

I just tried this out (one line change in the script!) on the 00-x86 program with a +12 offset on non-tested cores. Rough estimate is the tested core is running 6-10C cooler due to less radiant heat from the other cores running at lower voltages. That seems like a significant improvement!

[sp00n]

I'm going to add this in the next version with this config setting:

# If setVoltageOnlyForTestedCore above is enabled, you can define which Curve Optimizer / voltage offset value you want the other,
# currently not tested cores to be set to
#
# Note: If the "current value" for a core is higher than what is entered here, e.g. derived from the "startValues" setting or
# from errors during testing that caused an automatic adjustment, the higher value for this core will take priority over this setting
# to avoid instabilities
#
# Default: 0
voltageValueForNotTestedCores = 0

[ironmannFB3]

What would be the entered value to disable this function entirely if we think we are withing +/- 1 stability across the board?

[sp00n]

I'm not fully sure I understand.
You can just disable the entire functionality by setting setVoltageOnlyForTestedCore = 0 if you think your other values are fine.

[ironmannFB3]

Oh... I thought setVoltageOnlyForTestedCore=0 set the CO of all other cores during the test to 0 rather than whatever they are currently set at... i.e. -30, -20, -25, -15, -29, -31, -36, -21.

[mkoehmst]

The basic idea is that at testing start, you don't really know what the range of stable values is, so having non-tested all set to 0 works well.

Then after some initial testing you would have a pretty good idea of what's somewhat stable, but not totally stable range. So then you could set your non-tested cores maybe +6 to +10 from their current value.

Then, as you really start getting close, you can move non-tested down to +2 to +5, or just set setVoltageOnlyForTestedCore = 0 to leave them at their current values.

But in my opinion, you don't know what's truly stable, so leaving the non-tested cores at their current values always leaves the possibility that crashes might come from a non-tested core. So it's a good idea to be increasing the voltage at least a little on non-tested cores to mitigate that possibility. But only a little, so you're not unnecessarily increasing Vcore and the radiant heat across the package and causing the tested core to run at a higher voltage or potentially at a lower frequency than it could otherwise hit.