On my own K6-2+ chip, I find bit-14 (decimal) off the special "instruction
decode control" register aka SR13 is stuck at 1 - other bits are read/write
and CPU obeys their settings.

According to the K6 anatomy book by Bruce Shriver, that bit set would mean :
"MMX/3D (short-?)decode disable" - a debug feature, not desirable for regular
operation :(

At this point 2 hypotheses :
- AMD released defective processors and silently 'hardwired' this as a fix!
- The K6-"plus" chips indeed keep this bit stuck at one, unlike earlier K6 3D
microprocessors ( Neither the Shriver book nor his web site have been updated
to reflect changes introduced with the newest .18u line of K-sixes)

Unfortunately I disposed of my other, non plus, K6-2.

Anybody with K6's and a basic knowledge of programming/debugging (which I presume
means everebody here;) could help by doing a quick read/write test of that register bit?
(No need to be very bold : it doesn't bite ;-) It's five minute's time ...)

TIA to whoever helps!

brief how-to : MOV EDI, 0F543BC8h ; required 'secret code'
; you might set up an int 0Dh trapping routine just in case (not shown),
; but I'm telling you it won't fault anyway USING ABOVE SECRET ;)
MOV ECX, 97A1036Dh ; K6 instr decode control
RDMSR ; real mode, or CPL=0
Result in EAX ( read as 0000 4000h for me; bit 14 always on! )
... ...
;then try to zero out bit 13 : ECX,EDI as from above
MOV EAX, 0 ; or whatever ya wanna play with
CLI ; just in case
WBINVD ; it can't hurt to be cautious
WRMSR ; 32 bits, EDX=don't care
STI

; and read again
RDMSR ; For me no way to reset bit 14 !