Only way to know is through structured testing.
I have an idea of what the test procedure would look like, then. I'm no expert, so I'm using off-the-shelf (i.e whatever I'm aware of that I can take from elsewhere) tests.
Part 1: Adjusting FCLKTo widen the range of possible testers, I'll adjust the memory clock speed down to 3200 MHz. I think it's a reasonable speed to run at, since tech websites tend to recommend this speed for Ryzen. With that in place, I'd use FCLKs of 1600, 1400, 1200, 1000, and 800. This equates to memCLK:FCLK ratios of 1:1, 7:8, 3:4, 5:8, and 1:2. Just keep your original memory timings; I don't think it'll have much bearing on performance.
Then, in BIOS, you set the memory clock speed to 3200 MHz, then you set the ratio to get your desired FCLK.
I sense a problem here, though. Do Ryzen motherboards actually let you set any ratio you want, or are you only allowed to use a predefined list of ratios? Or even better, can you just set the FCLK directly so that we don't have to deal with this ratio nonsense?
Part 2: Testing DF performanceTo test for FCLK's effect on DF performance, I'd use
vosgren's DFmark, running just the TESTSMALLGEN preset to test for worldgen and TESTSMALLEMBARK for embark (it tests for an entire simulated year, which would take a while on the big version). If using Windows, I'd go and use this:
cd df_44_12_win
ECHO DFmark > Test_Results.txt
CALL :gentest TESTSMALLGEN
CALL :embarktest TESTSMALLEMBARK
move Test_Results.txt ..
cd ../
EXIT
:gentest
ECHO. >> Test_Results.txt
ECHO Test = gentest, %1, First Run >> Test_Results.txt
TimeMem-1.0.exe Dwarf_Fortress.exe -gen 101 3498 %1 >> Test_Results.txt
ECHO. >> Test_Results.txt
ECHO Test = gentest, %1, Second Run >> Test_Results.txt
TimeMem-1.0.exe Dwarf_Fortress.exe -gen 102 3498 %1 >> Test_Results.txt
ECHO. >> Test_Results.txt
ECHO Test = gentest, %1, Third Run >> Test_Results.txt
TimeMem-1.0.exe Dwarf_Fortress.exe -gen 103 3498 %1 >> Test_Results.txt
ECHO. >> Test_Results.txt
ECHO Test = gentest, %1, Fourth Run >> Test_Results.txt
TimeMem-1.0.exe Dwarf_Fortress.exe -gen 104 3498 %1 >> Test_Results.txt
cd data/save/
rmdir /S /Q region101, region102, region103, region104
cd ../../
EXIT /B 0
:embarktest
ECHO. >> Test_Results.txt
ECHO Test = embarktest, %1, First Run >> Test_Results.txt
TimeMem-1.0.exe Dwarf_Fortress.exe +load-save %1 >> Test_Results.txt
ECHO. >> Test_Results.txt
ECHO Test = embarktest, %1, Second Run >> Test_Results.txt
TimeMem-1.0.exe Dwarf_Fortress.exe +load-save %1 >> Test_Results.txt
ECHO. >> Test_Results.txt
ECHO Test = embarktest, %1, Third Run >> Test_Results.txt
TimeMem-1.0.exe Dwarf_Fortress.exe +load-save %1 >> Test_Results.txt
ECHO. >> Test_Results.txt
ECHO Test = embarktest, %1, Fourth Run >> Test_Results.txt
TimeMem-1.0.exe Dwarf_Fortress.exe +load-save %1 >> Test_Results.txt
EXIT /B 0
And it does this 4 times. Since we're looking for sustained performance here, the first run of each will be discarded. As usual, the test system should be left unattended for the entire test to prevent user activity from contaminating the results. The small versions are used to prevent it from taking far too long; we need to repeat this performance testing 5 times for each FCLK setting. One round of this would take around an hour on my i5-9300H, dual-channel DDR4-2666, based on my preliminary testing.
The "Test_Results.txt" file created after each run should be saved somewhere, and the filename changed to something like "[CPU]_FCLK_xxxx", xxxx being the FCLK setting used for that run, and [CPU] being the CPU used, like "Ryzen 5 5600X".
Part 3: Compiling DataThing is, I'm not sure I'd trust people to compile their results in a consistent manner, so I think I'll ask them for the raw test data first so that I can compile it later. I have my own methods. I'd use Google Sheets so that people can contribute their results, but my account name is now my real name, so I don't think I could use that. Not sure I want yet another account to manage, too, so... eh.