There's a mini flame war going on in comp.unix.bsd about the wisdom of buying
an Intel architecture PC to run Linux/FreeBSD/NetBSD vs., say, an old Sun.

After it had done the usual degeneration to "Let's post benchmarks", this
posting came out:

From: gervasio@Ross.COM (Gregorio T. Gervasio, Jr.)
Newsgroups: comp.unix.bsd,comp.unix.advocacy,comp.sys.sun.misc,comp.unix.pc-clone.32bit
Subject: Re: What do people have against BSD (or Linux for that matter)?

>>>>> On Tue, 31 Jan 1995 23:46:47 GMT, peter@nmti.com (Peter da Silva) said:
[in short]
> Friendly local SS-10, no other users:
> 
> % /bin/time sh -c "echo '2^8192' | bc > /dev/null"
>       8.2 real         7.8 user         0.0 sys  
>
> 486 DX2/66, no other users.
>       4.71 real         3.88 user         0.15 sys
>
> Alpha OSF/1 (DEC 4000/710), heavily loaded:
> real   3.8
> user   1.6
> sys    0.1

SPARCstation-10 clone, 100 Mhz hyperSPARC:

% uname -a
SunOS betelgeus 4.1.3 7 sun4m

% /bin/time sh -c "echo '2^8192' | /bin/bc > /dev/null"
        4.1 real         3.9 user         0.1 sys  
% /bin/time sh -c "echo '2^8192' | /usr/local/gnu/bin/bc > /dev/null"
        0.8 real         0.7 user         0.0 sys  

> If performance matters that much, give DEC a call.

Or get better software.

------- End of Forwarded Article

Now, I *know* this is a stupid non-benchmark, but I just *had* to try it:

isolar:1:60 % uname -a
SunOS isolar 4.1.3 2 sun4c

isolar:1:61 % sysinfo +terse -show model
4/60 (SPARCstation 1)

(It's really a clone, actually.)

isolar:1:62 % /bin/time sh -c "echo '2^8192' | bc > /dev/null"
       27.4 real        25.6 user         0.5 sys  

Sigh.  Well, it's not like I didn't know a 20 MHz SPARC wasn't really slow ...

But meanwhile:

netbsd4me:1:26 [/tmp] % uname -a
NetBSD netbsd4me 1.0 NetBSD 1.0 (NETBSD4ME) #1: Wed Feb  1 18:34:24 PST 1995    
 earle@netbsd4me:/usr/src/1.0/usr/src/sys/arch/sparc/compile/NETBSD4ME sparc

netbsd4me:1:27 [/tmp] % dmesg | egrep SUNW,Sun
cpu0 at mainbus0: SUNW,Sun 4/40 (MB86900/1A or L64801 @ 25 MHz, WTL3170/2 FPU)

(In other words, a 25 MHz IPC.)

netbsd4me:1:28 [/tmp] % /usr/bin/time sh -c "echo '2^8192' | bc > /dev/null"
       93.26 real        70.64 user         0.61 sys

Uh ... can we repeat the question, your honor?!?

netbsd4me:1:29 [/tmp] % date ; echo '2^8192' | bc > /dev/null ; date 
Fri Feb  3 07:25:02 PST 1995
Fri Feb  3 07:26:35 PST 1995

Yep, 93 seconds alright.  And Pentium 90's are doing it in under *2*.
It sounds like a dynamically-linked SPECint92 benchmark suite would give an
IPC running NetBSD/SPARC 1.0 oh, ... about a SPECint92 rating of "2"  :-(
(Anyone want to match a dynamically-linked 1.0R NetBSD/SPARC machine against
 an 11/780 ... )

Any hope for a fix to those shared library slowdowns?!?  :-?

(And he begs & pleads for an Official Patch back-port to 1.0 Release ... :-))

	- Greg