drick@hplvdo.HP.COM (David Rick) (02/04/88)
I have received a number of inquiries concerning a previous
posting in which I badmouthed the ERSO BIOS -- too many to answer
individually, so here is what I've been telling people.
The ERSO BIOS which I used is probably two years old now, so the
situation may have changed (i.e. the BIOS may have improved). I
was assembling a dual-speed XT clone from components purchased
from a variety of sources -- a practice which I don't recommend,
as it was protracted and painful, if quite instructive. I went
through three motherboards before finding one that worked (no, I
will NOT answer inqueries about what motherboard to buy!). It's
likely that some of the problems I noted were with hardware, not
firmware. I report here only problems which I believe (perhaps
without good reason) were BIOS-related.
* System would crash if a DOS disk were not present at boot time.
(Most "reasonable" BIOSs will tell you what the problem is.)
* Much trouble integrating a hard disk and controller. "Creeping
crud" data corruption, and wouldn't work at all with some
controllers. This may be hardware related, but see comments on
interrupts below.
* Keyboard problems including double strike and spurious characters
and spontaneous reboot. Reports are that this is an interaction
between the particular brands of keyboard and motherboard, but I
believe it is fixible in firmware.
* Poor diagnostics. (Can you find the bad RAM chip based on the
error messages?)
* Much difficulty running interrupt-driven communications (FIDO
BBS). I loaned the system to a friend for eight months, and she
used it to run a FIDO node. She had numerous problems which I
can't report in detail, but which went away when we switched to a
Phoenix BIOS. My theory is that the hardware interrupt service
code in the ERSO BIOS wasn't very "tight," hence was not
tolerant of other routines patching themselves into the timer
tick interrupt (a common practice with TSRs and com programs).
This may also be a reason for the difficulty with hard disks and
controllers.
I think the ERSO BIOS will probably work fine for straightforward
use like wordprocessing, provided it works with your hard disk.
For BBS and other interrupt-oriented applications, I would be
inclined to pay extra for a Phoenix or Award BIOS (I base this on
reports in PC Week and EE Times -- I have not used the Award
BIOS).
A way to judge the amount of margin a BIOS leaves for other
programs to patch themselves into the interrupt vector is to try
running several TSRs and perhaps a high-speed com program (e.g
Procomm at 9600 baud) and see what happens (dropped characters,
usually). The shareware package BACKSCR2 (screen back-scroller)
has inefficient interrupt code, and should be a good test case.
Finally, to all those planning to purchase clones soon, here is
my advice (some of it based on superstition, no doubt):
* Avoid dual-speed ("Turbo") XT-class motherboards: few of them
work reliably. If you need speed, it is far better to buy an
AT-class machine. Most of these are based on VLSI chipsets
designed by pretty competant people. Thus there is far less for
the motherboard designer to screw up.
* Get a good-quality BIOS. Phoenix and Award would be my
choices.
* If you will EVER need/want a hard disk, BUY IT WITH THE
MACHINE. Attepting this piece of systems integration yourself
will likely get you a disk that doesn't work in your system, and
an unrewarding position refereeing the finger-pointing that will
occur between computer and hard disk vendors.
* Make sure the computer and accessories have FCC Class B
approval. Getting a computer to pass RFI requirements is no
simple task, so the fact that a machine _does_ pass tells you
something about the engineering and system integration.
Sorry this was so long, and so long in coming. Good luck to all
who wrote.
David L. Rick
hplabs!hplvdo!drick
Disclaimers: These are my own opinions. My employer has no
opinion on these topics. I may be biased in favor of Phoenix and
Award BIOSs because they are used in HP ("Vectra") clones. I
don't work for the division that makes the Vectras, and don't
know anyone who does. I design _voltmeters_ for chrissake! :-)