From: jack@solucorp.qc.ca (Jacques Gelinas) Subject: Re: How does Linux compare to SUN IPC? Date: Wed, 29 Sep 1993 05:01:17 GMT
slater@gandalf.nrlssc.navy.mil (Rick Slater) writes:
>a228dhal@cdf.toronto.edu (Dhaliwal Bikram Singh) writes:
>|> It has seemed to me that my Linux system at home (X and GCC running in
>|> a 15mb partition, on a 386-40, with room to spare) is faster than the
>|> SUN IPC workstations I use at school. I can only offer subjective
>|> speculation though, ie. time for a xterm to open, etc...
>|>
>|> I was wondering, if anyone has done benchmarks between the two for
>|> various processors. I am not saying that Linux is better, it still
>|> has a ways to go before it can match the all around appeal of the SUN.
>|>
>I've run several "benchmarks" on both a 486DX-33 and a Sun IPC with the
>overall result that the PC was in the same ballpark as the Sun. The
>benchmarks were: Drystone test, whetstone test, and a large TeX file
>that came with the SLS distribution (gentle.tex). The PC went from
>a .tex file to a .dvi file, using the same version of tex, in a little
>more than half the time that it took the Sun. OTOH, the Sun was slightly
>faster on the first two tests.
Those benchmark are funny. I think they highly depend on the
application used.
Those who compare Usenet news unpacking will
have a winner with linux. SUNOS is does synchronous file
creation, not linux. This means that if you creat a large number
of files, linux will goes faster than sunos, which sync the directory
to disk for each new file. SUNOS will be more reliable however.
Off course, this is apple and banana.
One real life test I am doing often is zip and gzip of large
file. The ELC (33 mhz sparc) is about twice as fast as a 486dx33
(both with 16 megs ram). gzip is 30% faster on the Sparc.
On a 486 66, zip is about the same speed, but it seem that SUNOS
I/O (SCSI) is faster than IDE I/O on Linux. Much faster, so to
crunch about 40 megs of DOS .obj file is 50 % faster on a Sparc ELC
than the 486 66.
Again, I guess it hardly depend of the type of the application
you are using.
One thing that really suprise me, it that people are getting
better benchmark on Linux for floatting point. This is difficult
to believe. Linux as little to do here. A 486 is not a screamer
for floatting point ... by far. Benchmark publish by Unix Review
show that a Sparc Classic outperform a 486 50 everywhere by almost
2 except on some integer test where the sparc was sligtly
slower. For disk I/O, the Sparc was much faster. They used a High end
486 50 PC with retail for about the same price as the Sparc.
Just more confusion I guess!
--======================================================== Jacques Gelinas (jacques@solucorp.qc.ca) Maintainer of US4BINR jacques@us4binr.login.qc.ca