James_Bell@f6.n3601.z1.fidonet.org (James Bell) (05/02/91)
> Can anyone here explain (in a simple > manner) what the 486SX is? Whats is the difference to the 386? > It's just a 486 without the built-in floating point unit. JB
Joachim_Kainz@p11.f11.n310.z2.fidonet.org (Joachim Kainz) (05/02/91)
> 486SX == 486 w/o "built in" coprocessor
486SX == 386 ?
Morrie_Wilson@p16.f8.n343.z1.fidonet.org (Morrie Wilson) (05/06/91)
>> Can anyone here explain (in a simple >> manner) what the 486SX is? Whats is the difference > to the 386? > > > It's just a 486 without the built-in floating point unit. > JB > > Actually, the skinny I've heard is that there are problems with the yields of the 486 processors with the built in co-processors. So Intel came up with the neat idea of being able to disable the co-processor and sell the result as a 486sx. In actuality, the co-processor does not lend a whole lot of speed to most applications. E.G. Lets say an application is 10% floating point intensive. Let say the coprocessor can speed up that portion 10X. The result is that with the co-processor it runs in 91% of the time, for a total savings of 9%.
melling@cs.psu.edu (Michael D Mellinger) (05/20/91)
In article <funpack111518@f1.n6000.z2.fidonet.org> Morrie_Wilson@p16.f8.n343.z1.fidonet.org (Morrie Wilson) writes:
Actually, the skinny I've heard is that there are problems with the yields of
the 486 processors with the built in co-processors. So
Intel came up with the neat idea of being able to disable the co-processor and
sell the result as a 486sx.
In actuality, the co-processor does not lend a whole lot of speed to most
applications. E.G.
Lets say an application is 10% floating point intensive. Let say
the coprocessor can speed up that portion 10X. The result is that
with the co-processor it runs in 91% of the time, for a total savings
of 9%.
Like a speadsheet recalculation? How about graphics? Are all the
calculations done in integer math? Do Adobe Type Manager or True Type
fonts require floating-point calculations? I'm not sure, but I think
that there should be enough "real" world applications that it would be
nice to have a math co-processor. And when the application does
require a co-processor it's going to be several(20? 30? 100 times?)
times slower w/o it.
-Mike