> At 04:50 PM 10/26/98 +0000, Todd Lewis wrote:
>
> >The first 32bit intel processor was introduced around '88 -89'.
> >the 486.
>
> Actually the 386 was fully 32-bit. The 286 was partly 32-bit.
The 386 (released in '86? Or was it '84?) was 32 bit. The 386sx (386
sucks) had a 32 bit internal path and a 16 bit external.
The 286 had an odd 24 bit bus...odd that. I believe it was still 16 bit
externally though.
> >It wasn't really fully utilized for 5 years.
>
> Depends on the software. I was using some 32-bit stuff just
> after I got my 386.
True enough. Once the 386 extenders became more popular, lots more DOS
based programs were 32 bit capable. If by "fully utilized" you mean on the
OS level, true enough. Longer really. Win 3.x had the Win32s come out
after a long long while. NT 3.1 (NTAS) came out in '93 I think. Win32s
couldn't really be called "OS support", but it did allow 32 bit programs
under Windows.
As for OS/2, a fully 32 bit version was also a long time in the making.
Fortunately, we have a situation where Intel is producing the Merced with
the history in mind. Microsoft, for one, is working closely so that they
should have 64bit NT not long after the Merced comes out. I hope :-)
> There are several ways to measure the chip - internal data, external data,
> addressing, and instruction length. It would help us a lot to
> have 128-bit (or
> more) arithmetic. 128-bit addressing would be vast overkill for the
> foreseeable future. AFAIK, 64-bit instructions will be fine,
> until you get to
> the point of putting several instructions into one word. So it
> seems possible
> that we would have 128-bit arithmetic and data, but 64-bit addressing and
> instructions.
That's how a lot of video cards work. They might use 128 bit internal
addressing for moving around large chunks of memory at a time. They might
shrink down to a 32 or 64 bit external bus though. The Matrox Millennium
for instance is 128 bit internally, 64 bit externally (if my memory isn't
totally failing me).
