🚀 go-pugleaf

spagiola@frinext.stanford.edu (Stefano Pagiola) writes:

>today's software will simply not run adequately on a 286;

Because today's software is written for 386's and real operating systems,
both can't run on a 286.

>some of it is borderline on a 386.

Well, it's often the lack of memory that slows things down. You can't expect
to run a nice user graphical interface and several tasks on a 386SX@25MHz in
4M memory (the X11 server alone sometimes takes 1.5M on my machine, and I'm
not even talking about things like GCC :). A faster disk is a good idea too.

What I mean is that not all problems can be solved only by buying a fast
processor. I know someone who bought a fast 486 motherboard so that his
FidoNet software would spend less on processing the FidoNet traffic. Gee what
a moron! That's just throwing money in the river! Procesessing FidoNet mail
is 100% *disk* bound. If disk I/O is the bottleneck, then buying a faster
processor won't help a bit (except to empty your wallet).

>If you want to run that software, and newer  
>one on the way, you'll need more power.  If you want a nicer  
>interface, you'll need more power.

What 'we' in the PC world do is to cover up the obscenities in the architecture
with a faster CPU. For instance, with IDE harddisks the CPU transfers *every*
single byte to and from the disk. At the time it transfers to/from disk, it
can't do anything else. So a program which accesses a disk 50% of the time
allows other tasks to run for the other 50% of the time. Thus your machine
'slows' down to 50%. Instead of using DMA, which has an efficiency of 90% or
higher, we buy a 2 times faster CPU so that it can do 2 times more work in
the 50% of the time it can do real work. Get the picture? In short, to get
more work done, you can choose to use your CPU more efficient or add more CPU
power. I don't say that one of the solutions is good or bad, but it clearly
shows that we do not *NEED* more processing power per se. 

Greetings,
Peter Busser