RE: [TCLUG:3420] MIPS is dead

[Brad DeJong <1796@mn3.lawson.lawson.com>]

Intel does cater to the "high-end" server and workstation market.  SGI   
and the others based their business models on a definition of "high-end"   
that no longer provides a large enough market to sustain them.  Their   
fundamental problem is that the definition of "high-end" keeps changing.

In most businesses, the reality is that bottom line dollars spent is easy   
to calculate while money wasted waiting for a slow machine isn't.  Once   
the burden of proof shifts from the purchasing manager having to prove   
that a PC can do the job to you having to prove that it can't,   
workstation sales go down.  Once it crosses the line to you having to   
prove that the PC can't do the job efficiently, even more sales are lost.

You used to need a high-end workstation to display real-time 3-D   
graphics.  Today, the Nintendo64 can provide the same level of   
performance that required a $20,000 workstation not too long ago.  For   
development use, that same $20,000 machine is available on the used   
market for less than $2,000 (SGI Indigo II with Extreme graphics and 250   
MHz R4400).  Makes it tough to sell new hardware.

Even Intel is facing this problem.  Once the average business user has a   
system that can be upgraded to Windows 2000 and Office 2000 without   
buying a new system, Intel's business model (everybody buys a new system   
every 2 - 3 years) takes a serious blow.  How can they convince you to   
keep buying new systems if the one you have does what you need.  It used   
to be easy.  Working in 64 KB of RAM was painful.  Users saw an immediate   
benefit after moving to a 256 KB system.  How many users with 64 MB,   
would see the same kind of benefit in moving to a system with 256 MB?   
 I'm sure that eventually the software folks will piss it all away again,   
but for the moment hardware has taken the lead.  In short, there is no   
"killer app" for a desktop system (for the average business user) with   
256 MB of RAM and 8 x 4 GB 7200 RPM SCSI drives.

On the server side, a manufacturer who employs 5,000 people can   
(probably) run their accounting and payroll systems on a 4-way Pentium II   
machine.  That means that a lot of customers who used to need a   
mainframe, a mini-computer, or a Unix server no longer do.  After all, as   
long as everyone gets paid on time and the month-end close finishes over   
the weekend, the system is meeting their business needs.  I have an   
aquaintance who still runs his auto parts store (replica parts for   
vintage roadsters) on an 8088.  As long as the inventory list prints, he   
sees no need to replace the machine.  And more importantly, there is no   
need for him to replace it.

I'm getting some other replies now in a similar vein to this, but with   
caveats about system bus and such being used to distinguish "high-end".   
 PCI may be commodity, but I never been able to saturate a 133 MB/sec   
bus.  Again, the question becomes can you really justify spending 4 times   
as much for a machine that will provide 20% better performance?  You may   
think so, but does the person who authorizes your purchases think so?



Serge M. Egelman wrote:

I agree with you, Apple and SGI deal with completely different markets.   
 It
would make sense that Apple should suffer because they are competing with   
Intel
on the desktop pc market.  However, one would assume SGI should be doing   
even
better since DEC is almost dead.  And furthermore, Intel does not cater   
to the
high-end workstation and server market.  So it doesn't make much sense to   
me
why this is happening.  I suppose that IRIX is dead too then?