Re: Retiring the sparc32 port
Jordan Bettis wrote:
Like Chris said, new machines generally draw a lot more power overall.
My Ultra 5 that I use as my desktop can draw 200W max, and probably
doesn't really draw much over 100W total. Compare that to a typical
modern PC desktop that has a 400W supply in it and probably draws well
over 200W, mostly to power a GPU so it can display silly bouncing
icons and semi-transparent window edges.
There are two separate things to take into account here. The first is the
quiescent consumption, I admit to not having values from a number of systems
so for the sake of argument I'll agree that this is generally increasing.
However I'd suggest that if a computer is sitting there doing noting you'd be
better looking for ways to power it off or use a shared computing resource-
Sunray or whatever.
The second thing- where I do have numbers to back up my argument- is how much
energy is consumed to perform a unit of work. My figures, by and large, show
that while running a "torture test" a range of computers consume between 60
and 550W, with no overwhelming correlation with their age. On the other hand
the time to complete a unit of work has dropped dramatically over the last 20
years, which leads me to suggest that by and large the energy consumed per
unit of work has also dropped significantly.
Looking at two extreme cases:
SPARCstation 20, 2 jobs, 130W (175VA) 8m12.582s 1,068
Compaq AP550 1GHz, 768Mb, 8 jobs, 135W (180VA) 0m42.730 96
That last column is W-min to complete a given workload, selecting the best
(fastest) figures by splitting it into a number of jobs.
So assuming that the quiescent consumption is equal you're /far/ better off
with a newer system since even if it consumes substantially more power while
working hard it does so for far less time.
Mark Morgan Lloyd
markMLl .AT. telemetry.co .DOT. uk
[Opinions above are the author's, not those of his employers or colleagues]