Running a CPU-intensive task (the calculation of the
area of the Mandelbrot Set using a window of (-2.0,-1.05)-(0.8,1.05), 4000x3000
resolution, and a maximum iteration limit of 1000), I compared the performance
of my workstation to that of an older PC, roughly half as fast, currently
running as my fileserver. My workstation was running QuickBasic 7.1 under
Windows XP; the fileserver, c++ running under Red
Hat Linux 9.0.
The results were nothing short of amazing: the
"slower" system, running Linux, ran the benchmark in 3 minutes, 3
seconds. The workstation, which is essentially an entire generation ahead in
overall architecture and roughly twice as fast, took 29:28 -- almost 30
minutes -- to run the same calculation!!
Therefore, it would seem that c++, running under Linux, is
roughly 19-20 times as efficient as QuickBasic 7.1 running under
Windows XP.
As much as I like Linux, I do feel I need to point out that
QuickBasic was developed roughly 15 years ago for 80286-class systems, and is by
no means a modern programming language. A great deal of the inefficiency is due
to the limitations of QuickBasic, and therefore this benchmark should not be
viewed as providing any meaningful comparison between Windows XP and Linux.
(That's for another day's Question!) Mostly, what it means is, that if I want to
do serious work on evolutionary computation, I should really start using a more
modern programming platform!!
(N.B. I disabled all TSR programs, such as Seti@Home, on the workstation, while the fileserver continued to serve files and run a copy of Seti@Home at low priority in the background, so this wasn't even a fair fight for c++/Linux!!.)