As intel's recent CPUs this APU has own "Turbo-core" technology along with variety of power-saving states.
Cause it's APU, i.e. CPU + GPU in one SoC with shared TDP and heating it adds additional complexity, NB/GPU freq adjustable too.

So, it would be quite interesting to find the best settings for this device.

GPU is "ATi" one here so best suitable for AstroPulse. So, as "peak performance" I would consider fastest possible GPU processing with AP, even with decreased somehow CPU processing.

I used BrazosTweaker to change device freqs/VIDs.
Unfortunately, nothing can change NorthBridge (NB) VID and multiplier, this app can's adjust them as other monitoring tools show.

So I tried to change CPU part params. No luck with CPU VIDs though there were reports on app's forum of successful usage with C-60 CPU.
But some interesting effect observed instead.
When I change VID for particular CPU-Pstate app shows that CPU switches in another P-state.
Changing VIDs of all P-states through 0 to 2 put CPU in some "P3" state (unknown for this app at all). In this state CPU multiplier is severily reduced, x6.15 instead of usual x10 and x8 when system overheated. CPU-Z considers this multiplier as unreachanbe one too (it shows x8-x13 possible range).
But such state allows speedup of GPU (I thing because CPU power consumption drops inner circuit was allowed to overclock GPU part to 400MHz instead of usual 276MHz).
This GPU speedup was real one, not only other monitoring tools show GPU freq rise, AP task was completed considerably faster too (test carried on fully loaded CPU to emulate actuall full load conditions).

Here is the picture of this strange APU state:



Execution time dropped from 416 secs to 358s while CPU time increased from 15s to 22s.
I would consider this state as of higher performance when usual one (though CPU production should drop of course).

Would be interesting to hear from other AMD APU owners, especially C-seria ones. Are you able to undervolt your device? Especially non-unlocked one (w/o "K" in name).

EDIT: as best possible up to day app settings this command line was used:
-hp -cpu_lock -unroll 4 -ffa_block 4096 -ffa_block_fetch 4096

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

Sorry that I haven't checked here for a long while.

From what I understand of what you're trying to do, you're trying reduce the CPU performance to give the GPU more 'breathing space' so to speak? Having no load on the CPU would allow the GPU the most flexibility, but then that would mean no CPU usage at all - so you're trying to find the best compromise to maximise GPU performance while still having at least some CPU usage?

---

Soli Deo Gloria

exactly.
From time of initial post I learnt how to deal with this tool little better, now I'm able to overclock it and undervolt.
So far I found that my C-60 can work on 1.05V with full load @1000MHz and net step in multiplier, but not in next 2 steps.

Now would be interesting to find from what point BIOS (?) allows GPU speed increase. It allows with x6.15 multiplier, doesn't allow with x10. Much space in between to check.

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.