I just flashed to P3.10 from P3.00, it went fine, BUT:

1. H/W Monitor tab still shows T-ctrl for it's temp display (no options to change that).  I'm not that fussed if this is the case, but maybe I misread and thought that it was changeable/did change.

2. The fan control options allow for a few options, one being T-die but is not labelled as such.  Unfortunately it doesn't actually work when in the OS!  I've set it to raise the fan speed from 30-50% from 39°C to 40°C, but it's doing nothing even way over at 50°C!

OpenHardwareMonitor and SpeedFan can both adjust the fan speed, but neither can read Tdie.

It seems that setting to use T-ctrl and offsetting the target temperatures +20°C may get the functionality back, but that is just what I was doing under P3.00.

3. User datonyb has indicated that with their own overclocked 1700 and voltage set to fixed mode, they have scaling voltage which reduces vCore under idle conditions.

Fixed mode is fixed for me, as I would have expected.  I punched in 1.30000v and that is exactly what I get with LLC3 +0.001v.  I am not overclocking.

4. Whilst I can't be sure this didn't happen before, I think the BCLK is moving around in it's own!  On Auto it should stay fixed at 100MHz, and it isn't.  I've observed in HWInfo64 that it was 102.1MHz max.  I set it to 100MHz manual and it still moves to 100.5MHz on occasion.  

I think this is a possible source for system instability right there!!!

#. Is there a complete change log for each BIOS/UEFI update?  I'd like to know just what was changed in P3.10 from P3.00.

2. T-die is still something NOT in the motherboard vocabulary!

The options are:

- Motherboard (the motherboard temp sensor)

- CPU (T-socket)

- CPU T-ctrl (Tctrl or T-die+20)

Nothing has changed if you want fans to follow the current chip temperature, you'll still have to manually input +20°C values into the fan control options.

I have my fan values as such:

T1 59

%1 30

T2 40

%2 50

T3 70

%3 50

T4 85

%4 75

Critical 100

This works nicely for a Noctua NH-U14s in a Fractal R5.  The CPU fan is not constantly jumping up/down (and causing more perceptible noise as a result).  39/40°C is the point at which the fan speed jumps to 50%.  I'd like to have control over hysteresis, and there is an option for this in a deeper advanced sub menu but I don't see it currently working.

1. It wouldn't surprise me to find this is what the temp reported in H/W monitor tab is still T-ctrl (Tdie+20).

3. Whilst Fixed mode voltage is doing what I'd expect it to do rather than what datonyb has it doing on their system, I have dialled in Offset to the same -0.100v as was in P3.00 and all seems stable.  vCore readings seems very similar.

I am not overclocking so cannot copy exactly.

What I washed to achieve is a max vCore ceiling.  There has never been this option in any BIOS for any system as far as I am aware, so why would there be one now.  I was just hoping that "fixed" may do a similar job, it does not.

Offsetting -100mV seems to do the job just fine and the idle voltage drop ~880mV to 790mV is saving about 2w on the kWh meter!  Those will add up over time.

I have LLC set to level 5 too, so I am trying to run the heavy (single thread) load end lower and lower.  It does pull the all-core down from ~1.20v to ~1.10v.

I may run into stability problems at some point, and this will be my first point of call.

Maybe I should start to play with overclocking with P-states, but I have no desire to shorten the life of the chip, but the exact opposite.  I intend for this thing to last around 10 years.

Re -x vs non-x.

Yes, I am thinking the same thing as you.  The extra stress by allowing additional boosting via XFR and turbo - they clock higher, lets face it, means that they'll require more voltage.  This is why the VID is set to 1.55v.  Although it's argued as safe, I don't like it.  I am no AMD engineer though and I maybe just being a doofus over this point.

Off topic:

I am starting to play with the Windows Power Plans.  I have discovered Rightmark PPM Panel which offers some rather interesting information I have not yet seen reported by other s/w.

From what I read, Windows 7 was more multi-core single-thread thinking, rather than multi-thread multi-core.  Windows 10 (and to some extent 7) seem rather badly programmed in terms of actually utilising the processor intelligence and dealing with the distribution of load.

Just run a single threaded load and you can watch the % utlisation bounce around all over the shop.  Forcing a higher percentage of thread parking (or rather lower % unpark) stops this happening so much, but may lead to a more sluggish feel as the chip has to wake the cores up.

Windows is not quick at dealing with load changes, and this could be a reason for the Ryzen Balanced profile setting the CPU to keep itself 'awake'.

I am investigating this now that I am fairly satisfied with my voltage, temps and watts settings.

BCLK is at 101.7MHz...bleh, if it ain't crashing....

I don't really mind at all, as offsetting -100mV is working just fine.

I've started playing with Windows Power Plan 'deep' settings and seem to have effectively forced Win10 to stop being a moron and let a 2-threaded load run over 4 threads but on just two cores.  I think the whole reason why SMT is important is because it's quicker than sending a whole load down one thread to the core.

Ran a 1 and 2 thread benchmark on &zip and utilisation was about 50% over the 2 or 4 threads, I am happy with that.  Before I tweaked the plan, it was bouncing all over the place.  Now the 2 cores jump to 3.9GHz and stay there, doing what CPB and XFR was meant to.

I think I would discover clear failings with the Windows dealing with threads and general load organisation if I was to run some Linux builds.  However, I am not feeling that the system is in anyway sluggish.

Call me sceptical, but buildzoid is the same guy who hooked up numerous points on the board to measure voltages and reached a conclusion.  Another guy wired up (correctly) and reported the UEFI was actually damn close and buildzoid had measured wrong.

Some interesting stuff he does post though.

For those URLs to work, just replace "%3F" with a "?" mark.  Auto Html encoding conversion junk.

Don't worry about the links, it's this damned site.

Yeah, AMD's stance is to burn more power and have a more functional CPU.  I am generally quite happy now that I am adjusting things how I want them.  Of course it may not work for everyone.

I'm gonna try one more thing that datonyb had, so I can at least be content in trying essentially everything that makes sense.