> > Some basic info. The standard IBM hardware setup is:
 > > COM1    IRQ4    3F8
 > > COM2    IRQ3    2F8
 > > COM3    IRQ4    3E8   *
 > > COM4    IRQ3    2E8   *
 > > * Note: COM3 and COM4 were not defined in the original IBM PC
 > > specification, so the last two are a "pseudo-standard".

 > The new (old now ;) defacto-std (which does not conflict) is:

 > COM1 IRQ4   3F8
 > COM2 IRQ3   2F8
 > COM3 IRQ5   3E8
 > COM4 IRQ2/9 2E8

 > Most new Motherboard BIOS's will detect/declare COM ports set
 > as per above. It's a pity SB cards use IRQ5 as their default.
 > Additional COM ports especially for OS/2 systems should be
 > setup as above if possible. If not other IRQ's are fine, OS/2
 > handles it via command line parameters for COM.SYS.

This "standard" must still be fairly recent as it isn't supported
on any motherboard I've ever seen, including the new ('95 Award bios) PCI
board in the machine I'm currently using. 

As you rightly point out sound cards commonly use IRQ5 as a default. Not
only that, but the end user can have much difficulty getting one to work
correctly on any other setting. The sound card manufacturers got away with
using IRQ5  because very few people have the originally assigned second
parallel port, so IRQ5 is already a third-level choice for use in serial
comms. IRQ2/9 is assigned to the second 8259A IRQ controller and is also
typically used for network cards, tape controllers and a lot of other
stuff. In PCI systems it's also the default first slot, which is notable in
that if the CMOS ever gets reset whatever is sitting there will conflict
with a device using IRQ 2/9 in the ISA bus.

In my own case I also have to contend with an EPROM burner, an audio
impulse waveform measuring system, a DSP development system and a scanner,
which goes some way towards explaining my long term advocacy of SASI/SCSI
systems. What it comes down to is the applications, hardware and operating
systems the end-user is likely to want to run during the life of the
machine, which can be difficult to predict, and how many are likely to be
running concurrently.

 > Note that setting more than one hardware device on the same
 > IRQ on the PC ISA bus (without hardware support for same) is
 > an ERROR condition - one which DOS (being single tasking)
 > fairly safely ignored. OS/2 will NOT ignore such a fault, and
 > there's NO excuse for today's new machines being shipped that
 > way.

I agree entirely, and the only excuse there _ever_ was, was the early lack
of available IRQs on 8-bit machines. The use of COM3/IRQ4 and COM4/IRQ3 was
rare anyway outside BBS usage, and really only became a serious issue when
PCs started having internal modems fitted. I standardly set them to
COM4/IRQ3 on the basis that COM2 is seldom used, and to avoid conflict with
a mouse typically installed on (and expected by most software to be seen
at) COM1.

In fact what I was trying to point out to David Smith was that the use of
COM3/IRQ4 for the mouse was, erm, unusual.

 > > If all else fails, put the 586-133 back in. The P100 won't be
 > > much faster anyway.

 > I'd guess enough to notice. The Pentium (64bit) data path
 > gives the system some grunt under load.

It would depend greatly on whether the application utilised the Pentium to
its full potential. Most still don't as yet.

Rob

--- FMail 0.94