386 Bugs

Davig Gonzales davidg at uns-helios.nevada.edu
Fri Sep 8 00:36:13 AEST 1989

It seems that I sent the following message to some poor soul via E-mail.
Sorry about that....

Anyway, here is the re-post.


	In the current issue of ESD (The Electronic System Design Mag.), under
Technology Trends, there appears to be more bugs with the current crop of 386's
(including 20/25 and 33 MHz versions). While the first two bugs are not too
serious the third one can could some problems with OS designers. And I quote:

	" The third and most serious problem is corruption of the prefetch
queue. This happens when: pipelined bus mode is used; paging is enabled; system
is running with 1 to 3 wait states; code queue is full; instruction decode is
full; the two prefetches that filled the code queue pipelined bus cycles; and
the next code prefetch causes a TLB miss."

	It should be noted that this bug is in the newer versions of the 386
and does not appear in some of the older 386's (including the 16 MHz version).
This "newer" version was to overcome some of the problems with the older 386's
(paging problem, better FP support, etc...).

	My question is how common does the abovementioned condition occur? I
am in the market for a 20/25-MHz 386 running UNIX (Bell Tech?) and cannot see
paying $400+ more for a cached 386 (which overcomes the above problems). Also,
has anyone had any experience with the AMI 20-MHz 386 machines (ie - will they
run Bell Tech, BOIS support, etc).

	If anyone is interested, a new version of the "D1-step" should ship by
4Q/89 according to the article.


					David A. Gonzales
					davidg at uns-helios.nevada.edu


More information about the Comp.unix.i386 mailing list