Subject: cpu cycle server machine
To: None <current-users@NetBSD.ORG>
From: der Mouse <mouse@Rodents.Montreal.QC.CA>
List: current-users
Date: 05/10/1998 19:05:51
I'm looking at getting a new machine, specifically as a cpu cycle
server.  Machines that don't run NetBSD need not apply.

What I'm wondering is, does anyone have recommendations?  How can I get
the best crunch for the dollar?  (Obviously, technically, I get the
best value for the dollar by picking up some old clunker free.  Since
what I'm looking for is computrons, this isn't really a contender.)
The three plausible CPUs according to what little I know are the Intel
(x86, Pentium, etc), the PowerPC, and the Alpha.  But I don't know much
about most of the rest, and even these have their problems:

- Intel: there are so many cheap knockoffs it can be hard to find a
   decent machine without gaping holes (including security holes), and
   said decent machine will be more expensive than the really low end
   that makes Intel boxen look so attractive.

- PowerPC: all the PPC machines I know of are Macs, but I don't want
   (or at least I don't want to pay for) a fancy screen and keyboard
   and sound system and all the other goop Macs insist on.

- Alpha: I've heard that alphas don't give as good crunch for the
   dollar as Pentia, this from someone who was buying some for one of
   my employers.  However, I think they concentrate more on floating
   performance than I do; I mostly want integer-type stuff.

I have minor philosophical objections to the Intel architecture, but
would be willing to ignore them if the price differential is
nontrivial.  Having a 64-bit machine would be Cool, but that factor too
probably isn't worth much money to me.

I'm not looking for much of a system beyond CPU crunch: CPU, RAM on the
order of 32-128 megs, Ethernet (10baseT or AUI) for talking to the
world, SCSI for disk, and either a minimal console screen-&-keyboard or
a serial console port.  However I *am* looking for a well-designed
machine.  For example, I consider unacceptable soft-config peripherals
that use undocumented mechanisms to set the config (Intel boxen are
particularly bad for this) - they mean that software bugs can get the
hardware into a state it can't recover from.  Similar remarks apply to
machines with flash (or other electrically modifiable) firmware, unless
writes can be disabled by something like a motherboard jumper.

Any thoughts from anyone?

					der Mouse

			       mouse@rodents.montreal.qc.ca
		     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B