They both work, but TSCAdjustReset is specific to X299 and only needs to run its code once, on startup. Where CpuTscSync runs regularly, always making sure the TSCs are in sync across the cores. But on X299, the TSCs are already always are in sync - on X299 just one correction needs to be made to them all, which TSCAdjustReset does at startup. After that, it goes idle - it could even be unloaded, the Github page says (though I don't believe it's possible to unload running kexts any more as of Big Sur.)
So theoretically using TSCAdjustReset should be more efficient, saving a few clock cycles and having one less active kext. But it's probably such a small difference that it's not measurable, or certainly not by standard benchmarks.
CpuTscSync has the advantage that it's actively maintained by Acidanthera, making it as 'official' as anything can be for Hacks. Meaning it will definitely be updated and kept current over time, should any future compatibility changes be needed. We don't know if that will be the case for TSCAdjustReset, though it certainly runs fine for now.
I use TSCAdjustReset on the basis that if I can have fewer unnecessary instructions being executed, why not. But I really doubt there's any significant difference either way.