PC 80486

[D Matthew Lehman]

Juan P. Abonia (pablo at acsu.buffalo.edu) wrote:
> In article <1993Jan25.170138.10928 at welchgate.welch.jhu.edu> ken at oscar.welch.jhu.edu (Ken Fasman) writes:
> >In article <1993Jan25.090410.12066 at gserv1.dl.ac.uk>, schnorr at tournesol.versailles.inra.fr (Kirk Schnorr) writes:
> >|> 	We were thinking to get a 486SX based machine since it is the
> >|> cheapest but I do not want to sacrifice a built in  math co-proccessor
> >|> or 32 bit data path it the 486SX is missing these.  Can anyone enlighten
> >|> me?
> >
> >My understanding of the various members of the Intel 80386 and 80486 families 
> >are as follows:
> >
> >Chip		Internal	External	Floating
> >Designation	Data Path	Data Path	Point Unit?
> >-----------	---------	---------	-----------
> >386DX		32 bits		32 bits		No
> >386SX		32 bits		16 bits		No
> >
> >486DX		32 bits		32 bits		Yes
> >486SX		32 bits		16 bits		Yes
> >
> >Please note that Intel uses the "SX" designation to denote the narrower 
> >external data path, NOT the presence or absence of a math coprocessor.

> This isn't always true.  The 486SX has a disabled floating point unit.  The
> upgrade chips (Overdrive) for these machines disable the original CPU when they
> are installed (They have an enabled FPU).  I believe that the 486SX retains a
> 32-bit external data path.
> -- 
> J. Pablo Abonia
> University at Buffalo				 BITNET:    pablo at sunybcs.BITNET
> Roswell Park Cancer Institute			 Internet:  pablo at cs.Buffalo.EDU

Nope, it's only a 16 bit path to memory.  The 486SX was really a backwards
design for Intel.  They had so much success with the 386SX they took the 
486DX and crippled it into a 16 bit path and disabled fpu support, hoping
that everyone would buy the 486SX over a 386 machine (AMD had started 
producing 386 clones and Intel was not pleased).  The basic design of
the 486SX is the same as the 486DX.  My opinion on the 486SX is that
a person is better off with a 486DX.  Beware of benchmarks that say the
chips are almost computationally equal, as they probably don't talk into
account memory reads/writes.

--------------------------------------------------------------------------
D. Matthew Lehman                   Internet: mlehman at vcs10.vetmed.wsu.edu
Veterinary Computing Services       Phone:    (509) 335-0101
Bustad 203
Washington State University
Pullman, WA 99163
--------------------------------------------------------------------------