I have one more correction to Ken Fasman's (ken at oscar.welch.jhu.edu) last
posting. (Edited version below) Ken is correct in stating that the original
486SX was cast from the same die as the 486DX; however, he is incorrect in
stating that the 486SX has a 16 bit data path. Same die=same interactions
with the motherboard=same data path.
There was actually quite a bit of furor when the 486SX was first announced
about the fact that the 486SX was being sold for less than the 486DX. Since
the 486SX had to undergo the exact same manufacturing process as the 486DX
with the added step of disabling the FPU, how could the SX be cheaper. (Intel
claims the cost saving comes from not having to test the FPU.) The original
concept for the Overdrive was to allow the user to insert a full fledged 486DX
into the slot, completely disabling the 486SX. (Incidently making you pay for
the main microprocessor and 8k cache twice.) Later incarnations, as Ken
states, no longer contain the FPU. (I guess it was cheaper not to add the
extra silicon after all.) Anyway, sorry about the trivia.
I pulled a couple of magazines off my shelf just to make sure that this is
correct. (PC Magazine V11 N17 p115-257 and PC World December 1991 p 148-167)
Neither was the definative article that I remember reading somewhere but both
contain sidebars about the 486SX. Perhaps Ken was thinking of the Cyrix
486SLC, which is essentially a 386SX (16-bit data path) with a 1K cache and
psuedoFPU. Whoops, this is longer than I intended.
Bottom line. 486SX= 486DX - FPU. 32 bits internally and externally.
Dulaney at uclaue.mbi.ucla.edu
>This is not quite right! It turns out that the 486SX had its floating point
>unit disabled in the original chip, and removed from the die altogether in
>later versions. That is, the 486SX has a 32 bit internal data path, a 16 bit
>external data path, and NO FLOATING POINT COPROCESSOR.