On Wed, 11 Nov 1998, Lawrence D. Lopez wrote:
> I notice that one Pentium II 450 is $600.
> But two Pentium II 350 processors are $249 x 2 or $498.
>
> Obviously single threaded code would run faster
> on the 450 mHz processor.
>
> I was wondering if it was worth the trouble going to
> SMP.
There is basically no "trouble" in most cases, especially if you plan to
run 2.0.X. You uncomment one line in the kernel Makefile, make a
kernel, boot it, and forget it. If a very few cases there are hardware
conflicts or particular applications that tweak one of the very few bugs
that remain in the 2.0.X kernel, and even these can usually be handled
via this list.
> The application is program development.
>
> I, personally, have found 2 processor systems handy
> when large amounts of kernel debugging output are placed
> on a X windows screen. This was in a Sun environment
> though.
I run a largish cluster of dual systems and have a LOT of experience
with linux in code development in both single and dual configurations.
Dual systems can be handy in program development as one can make in
parallel (within dependency limits) and one can also run multiple tests
at once while debugging. However, even single PII's of any clock/flavor
at are so fast and linux so efficient that I doubt you'll be able to see
any difference between single and duals for putting debugging output on
an X screen, unless you are working one processor at 100% while doing
so. Perhaps, as you note, when debugging the kernel, which has a
Heisenberg Uncertainty effect.
In fact, even a 300 MHz Celeron A is perfectly acceptable for a
development environment unless your builds are REALLY long and
complicated (a full linux kernel recompile of one of my configurations
from "clean" takes only 4 minutes or so on a 300A, vs 2+ minutes on a
dual 350 and 2- minutes on a dual 450 -- at some point it ceases to
matter).
Further benefits of an SMP system are twice as many CPU's (and more
total power) for production runs; a second CPU to handle the kernel and
devices when working on development (certain network tasks may be
faster) and, of course, the possibility of developing multithreaded
code.
If it were me and I were ONLY doing program development, I would (and
did) get a 300A system -- it is even cheaper than a 350 PII, perfectly
adequate for running an interactive environment, and (with its
full-speed cache that, although smaller than that of a regular PII, is
still much bigger than most sequentially read source files) likely "just
as fast" for compiles. Only if I were doing program development of
multithreaded programs, beowulf (distributed parallel) programs, had
plans to use the system for production, or just wanted to be able to run
a test job while working unencumbered, would I seriously consider a
dual.
rgb
Robert G. Brown http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567 Fax: 919-660-2525 email:[EMAIL PROTECTED]
