🚀 go-pugleaf

JoSH recently mentioned that in 10 to 20 years we will see an
improvment of 1000 to 1 million in computing technology.  While
this is the "standard" figure, it seems to me that a more accurate
rate is 10 to 100 (say, 30) each decade.   Memory has improved a
little faster, crunching a little slower.

Examples.  In 1965 the Linc, a high end "personal workstation" of the
time, had 3K bytes and an 8us cycle time.  Today a RISC has about 10Mb
and 40ns.  This is x3000 and x200, or about x30 and x10 per decade.
One gets similar numbers with home computers, and smaller numbers
comparing a 360/91 and a Cray 3.

So perhaps O(30) per decade is a better choice for back of the
envelope calculations.  Of course, this has no impact whatsoever
on the point of JoSH's note.

Joe

[I can defend the 1000 with examples, but, as you say, it's not the
 point.  The 1000 figure is a quote from CMU's own Raj Reddy.
 --JoSH]