in the sbc-bench results (sbc-bench/Results.md at master · ThomasKaiser/sbc-bench · GitHub) there is an older entry for the star64 identifying itself as visionfive2 (so i guess it was an early vf2 given to the star64 crowd to do some early work on) which has cpu opp points up to 1.75ghz and as a result also a slightly higher (about 20%) max performance - see: http://ix.io/4a3s
looking through the old commits in the vf2 kernel tree one can also find commits removing those extra opp points and cpu idle states as well like:
and merges like:
does anyone here maybe have any information or idea if those were removed for stability reasons or because they were buggy or only availabe in early samples of the vf2 or jh7110? the sbc-bench results look like they seem to had been working at least at some time in the past …
Skimming the commits, and with a limited knowledge of CPU internals and timings etc, the real question I have is: Define Unsupported !?
Eg; Does this need to be changed because the hardware does not support it; in which case things may go really wonky if you revert the commits.
Or; are StarFive just keeping within the product definitions/spec and removing functionality not needed to meet that; especially ones that might push the CPU out of thermal spec.
The latter is quite a legitimate thing to do; the hack/overclock community can recompile and work around these changes, even improve them, but StarFive do not have to devote resources to testing and supporting use outside the specs. (It’s the testing that costs the most by the way, double the number of freq/divider combinations you need to support, double the time/equipment/human resources needed to test them properly for each release sprint/cycle)