Path: utzoo!utgpu!watmath!clyde!att!osu-cis!tut.cis.ohio-state.edu!cwjcc!ukma!gatech!ncsuvx!mcnc!rti!bcw
From: bcw@rti.UUCP (Bruce Wright)
Newsgroups: comp.sys.ibm.pc
Subject: Re: 80x86 numbering (was: 80486)
Summary: Downloadable microcode? Bah!
Message-ID: <2660@rti.UUCP>
Date: 19 Dec 88 06:36:29 GMT
References: <15374@iuvax.cs.indiana.edu> <45900175@uxe.cso.uiuc.edu> <16716@conexch.UUCP>
Organization: Research Triangle Institute, RTP, NC
Lines: 52

In article <16716@conexch.UUCP>, rob@conexch.UUCP (Robert Collins) writes:
> In article <715@ethz.UUCP> zu@bernina.UUCP (Urs Zurbuchen) writes:
> -The 80486 is rumored to have a virtual 80286 mode as the 80386 has a
> -virtual 8086 mode. That's all we waited for, isn't it. So we can run
> -multiple OS/2's on one computer :-) (If you run two OS/2's does this
> -make it a complete Operating System ?)
> -
> -By the way, this virtual 80286 mode caused Intel a lot of problems. It
> -seems to be the main reason why the chip is that late.
> -
> In the latest (or one of the latest) issues of Microprocessor Report,
> says the '486 will have downloadable micro code, and a 6 instruction
> prefetch queue.  As for the bigger prefetch queue, that means that
> relative jumps can be done in 0 clock cycles.  But as to your assertion
> of disbelief that much can be done other than speedups, I think
> downloadable microcode is lightyears more advanced than the '386.

Umm, well, I am by no means entranced with the '386 (for various reasons
...).  My original posting was more along the lines that the '386 has
become too much of a muchness, and that in order to do anything else
with it it seems that there is little choice but to go with something
like virtual 80286 and virtual 80386 modes.  This is hardly an improvement!!!
The architecture is too complex by half already, there is very little 
percentage in piling kludge upon kludge.

As for downloadable microcode, well, I have used a number of machines
that had downloadable microcode;  and although we always thought we would
try to take advantage of it we never found it to be cost-effective.  I
know of other installations which had the same reactions - it's something
which sounds really neat, and it may even be useful so that the Field
Service Engineers can update the microcode on the machine to correct bugs,
but _VERY_ few ever really used it.  Seems likely to me that its only real
use on a 486 will be just that - correcting bugs (assuming that the bugs
are fixable in microcode and aren't embedded in the microengine .....).
It might be _SOLD_ as being something that had broader utility but in my
experience this would be mostly marketing hype rather than solid value.

I also don't quite understand why it is considered so horrible that it
might be a good idea to call a halt to the continual reorganization of
the 80x86 machine architecture and just concentrate on speedups.  As it
is now there is very little software which really takes advantage of the
80386 native mode ... after all, in the real world what matters is how
much work you can get done for how many $.  If you are going to increase
the complexity of the architecture (and hence the difficulty of using the
architecture) then that complexity will have to produce significant
returns.  My concern is that the 80386 with all of its different operating
modes is already complex enough, and provides a reasonable functionality.
Why complicate things further for little additional gain?  If you are
really running into architectural limits I would submit that the time has
arrived to junk the architecture and look at going with a cleaner design.

						Bruce C. Wright