I'll concede that I may have overstated the problems with PCI cards and a
well designed card should work in additional systems as well as X86.  (I
still maintain that not all cards are well designed)  However, there is
still an issue with peripheral hardware that is processor specific.  Also, a
third party vendor must make a conscious decision to support their products
on a given platform.  This can have significant cost even if it is just a
"re-compile" of the drive.  Cost which many companies will not pay to sell
in to what they consider a small market.  I spent months trying to get
National Instruments (www.ni.com) to support one of their products on an
Alpha/WinNT machine.  They have full support for X86/WinNT and it should
have been simple to recompile for the Alpha, but they wouldn't do it.  We
even offered to do the port ourselves.  We lost a multi million dollar
contract, in part, because NI was so stubborn.   
  Issues like this are a significant reason for the continued dominance of
the X86/PC architecture.     

-----Original Message-----
From: Michael Torrie [mailto:[EMAIL PROTECTED]
Sent: Friday, February 21, 2003 6:04 PM
To: [EMAIL PROTECTED]
Subject: Re: [uug] Industry standards processor


On Fri, 2003-02-21 at 17:19, Soren Harward wrote:
> Tucker, David wrote:
>  > They may not be concerned about running
> > Windows app (for a time, I had an SGI X86 PC that did run windows) but
there
> > are plenty of peripheral that run on the X86.  Your SCSI card or fancy
new
> > video card use X86 bios code.  Plug one into the PCI bus of a sparc, and
you
> > wont get far.  The PCI card bios can be redone, but that's expensive and
you
> > will loose the advantages of commodity goods.
> 
> The statement about PCI hardware being tied to he processor isn't 
> entirely true.  There's a ton of hardware I've used in both Alpha and 
> x86 machines, from SCSI controllers to sound cards.  Right now I have a 
> graphics card in my Alpha that used to be in my K6/166.  There might be 
> some weird platform-dependent cards, but in every case I've seen, as 
> long as the driver compiles on the architecture, the card will work.

I second that.  I have an ultrasparc 30 at work, and I plan on putting
in a promise ide controller and an ide disk, since scsi are so expensive
(and western digital is now selling 10k rpm enterprise-grade ide
disks).  Most video cards should work.  Sound cards, etc.  There are
endian issues (With DMA and so forth), but that's largely a software
issue.  The driver sorts that out.

Most PCI devices have no real dependency on the BIOS per se, as long as
someone initializes them along the way, like the operating system.  The
PCI spec itself provides for a lot of the configuration. On systems that
support hot-swapping the pci bus, there is no bios there at all.  It's
all up to the OS.  Hopefully Linux will support this in 2.6.  Only
Solaris does currently.

Michael


-- 
Michael Torrie <[EMAIL PROTECTED]>


____________________
BYU Unix Users Group 
http://uug.byu.edu/ 
___________________________________________________________________
List Info: http://phantom.byu.edu/cgi-bin/mailman/listinfo/uug-list

____________________
BYU Unix Users Group 
http://uug.byu.edu/ 
___________________________________________________________________
List Info: http://phantom.byu.edu/cgi-bin/mailman/listinfo/uug-list

Reply via email to