Path: utzoo!utgpu!watserv1!watmath!att!linac!pacific.mps.ohio-state.edu!zaphod.mps.ohio-state.edu!usc!apple!agate!darkstar!ararat!karplus
From: karplus@ararat.ucsc.edu (Kevin Karplus)
Newsgroups: comp.benchmarks
Subject: bad mini-benchmark
Message-ID: <9540@darkstar.ucsc.edu>
Date: 29 Nov 90 21:04:38 GMT
Expires: 29 Nov 90 21:04:37 GMT
References: <19040001@orac.HP.COM> <4344@awdprime.UUCP> <9516@darkstar.ucsc.edu> <NHZ_D+?@b-tech.uucp>
Sender: usenet@darkstar.ucsc.edu
Reply-To: karplus@ce.ucsc.edu (Kevin Karplus)
Lines: 22

The "benchmark"
	echo 2^5000/2^5000 | /bin/time bc
seems to have poor predictive power for other applications.

Here are results on different Suns (all diskless workstations with at
least 8Mbytes of RAM): 
	Sparcstation SLC	44.5 real	43.4 user	0.1 sys
	3/80			33.3 real	28.7 user	1.0 sys
	4/110			26.8 real	26.3 user	0.1 sys
	Sparcstation 1+		12.2 real	12.0 user	0.1 sys

Note that the 3/80 appears faster than the Sparcstation SLC!  Having
used both machines for months, I can assure you that the SLC is easily
a factor of two or three faster on most applications (except
compilation and linking, which are I/O bound on my diskless SLC).

Does anyone have an explanation for the unexpectedly poor performance
of the SLC on this trivial benchmark?
All three of the Sparc-based machines were running the same copy of
"bc", so it isn't a difference in compilation.

Kevin Karplus