That's exactly what I noted in my first FLEX_RATIO post. Patching that code was how I found that this MSR mishandling was the root cause of my mach_kernel failing to boot.
Anyway. Looking at the mach_kernel of Lion I see this:
Where WESTMERE, NEHALEM, and SANDYBRIDGE all jump to the same spot reading the MSR that caused you problems.
I'd say no, it's definitely *not* safe to assume undocumented fields are guaranteed to be set the way apple's code happens to be testing them, when using non-apple hardware. Relying upon undocumented hardware behavior is not good practice. If intel has registers whose semantics are only shared under NDA with a few vendors (and not the linux folks for example), then there isn't really anything concrete to complain to the bios vendors about.
And since Apple is Intels biggest client, and since Apple has Intel engineers working for them... I'd say that it is safe to assume that MSR 0x194 is part of what should work. Not to mention that more MSR's are missing from the docs/white papers, and that Asus engineers cite NDA restrictions when people do ask for info about them.
I do however think that having the osx bootloader clear bit 16 of that MSR when bits 8 thru 15 come up 0 is a very reasonable fix. Works for me . If anyone else needs a chameleon boot file with this fix let me know.
I have found a few FLEX_RATIO references in intel's documents but nothing that details the semantics.