mark@intek01.uucp (Mark McWiggins) (01/30/91)
The most recent version of Unix Review contains a rather flamish letter from someone who claims that his RS6000's don't perform appreciably better in the application they're running than do SparcStation 1+'s, and are much harder to port software to, as well. Any comments on this from RS6000 users? Thanks in advance. -- Mark McWiggins Integration Technologies, Inc. (Intek) +1 206 455 9935 DISCLAIMER: I could be wrong ... 1400 112th Ave SE #202 Bellevue WA 98004 mark@intek.com Ask me about C++!
irv@masc35.rice.edu (Irvin Lustig) (01/31/91)
In article <1991Jan30.003004.74@intek01.uucp>, mark@intek01.uucp (Mark McWiggins) writes: |> The most recent version of Unix Review contains a rather flamish |> letter from someone who claims that his RS6000's don't perform |> appreciably better in the application they're running than do |> SparcStation 1+'s, and are much harder to port software to, as well. |> |> Any comments on this from RS6000 users? Thanks in advance. |> Our Fortran application screams on an RS6000 530. It is about 6 to 7 times faster than on a Sparcstation 1+. You must make sure you use the optimization options in the compiler to see this type of performance. As for porting the application from other Unix platforms, it was much easier to port this particular application to the RS6000 than to the Sun (we were porting from a Silicon Graphics box - the Fortran uses no graphics). This is because the Sun Fortran compiler had some bugs in it. We have also found that IBM's level of support for bugs and other problems is far better than anything we've received from Sun. After our recent experience with Sun, we are stunned that they have been so successful in selling workstations. -Irvin Lustig Assistant Professor Princeton University Department of Civil Engineering and Operations Research irv%basie@princeton.edu (Currently visiting Rice University - irv@rice.edu)
jet@karazm.math.uh.edu ("J. Eric Townsend") (02/01/91)
In article <1991Jan31.152640.12173@rice.edu> irv@masc35.rice.edu (Irvin Lustig) writes: >Our Fortran application screams on an RS6000 530. It is about 6 to 7 >times faster than on a Sparcstation 1+. Comparing a 530 to a 1+ is a bit of a biased comparison, IMHO. Try a 320 to a 1+, perhaps. Our 320's are 3-5X faster than our 1's on floating, and about 1/3 faster on integer only character manipulation. Not bad, since the people who bought the 320's wanted to run fortran array crunchers fast and for little bucks. >This is because the Sun Fortran compiler had some bugs in it. What an understatement. If all you do is run fortran code, *don't buy a sun*, or at least buy somebody else's fortran compiler. SMI's fortran bites pretty hard. On the other hand, I've never had to port a single C-based app that I wanted to use on the Sun. :-) -- J. Eric Townsend - jet@uh.edu - bitnet: jet@UHOU - vox: (713) 749-2120 "It is the cunning of form to veil itself continually in the evidence of content. It is the cunning of the code to veil itself and to produce itself in the obviousness of value." -- Baudrillard
jfh@rpp386.cactus.org (John F Haugh II) (02/01/91)
In article <1991Jan31.152640.12173@rice.edu> irv@masc35.rice.edu (Irvin Lustig) writes: >In article <1991Jan30.003004.74@intek01.uucp>, mark@intek01.uucp (Mark McWiggins) writes: >|> Any comments on this from RS6000 users? Thanks in advance. >|> >Our Fortran application screams on an RS6000 530. It is about 6 to 7 >times faster than on a Sparcstation 1+. I can't imagine any Sparcstation out-performing a S/6000 in floating point, considering how poorly a Sparcstation performs relative to a S/6000 in simpler tasks. I've never met anyone who wasn't impressed by the performance of the machine, and frankly, I miss mine. >We have also found that IBM's level of support for bugs and other >problems is far better than anything we've received from Sun. I'm certain the people working in Defect Support here in Austin appreciate the vote of confidence. I know the manager who set up the groups which work on the actual software fixes, and he is very committed to providing a high level of service. > After >our recent experience with Sun, we are stunned that they have been so >successful in selling workstations. Considering that IBM sold $1 billion in S/6000 last year, perhaps IBM has learned something that Sun long forgot. -- John F. Haugh II UUCP: ...!cs.utexas.edu!rpp386!jfh Ma Bell: (512) 832-8832 Domain: jfh@rpp386.cactus.org "13 of 17 valedictorians in Boston High Schools last spring were immigrants or children of immigrants" -- US News and World Report, May 15, 1990
udo@watzman.quest.sub.org (Udo Klimaschewski) (02/01/91)
From article <1991Jan30.003004.74@intek01.uucp>, by mark@intek01.uucp (Mark McWiggins):
> Any comments on this from RS6000 users? Thanks in advance.
Yupp. I agree.
This damned 530 here is a real good number-cruncher.
Ray tracing is fun - but never NEVER let anybody else use this beast
while you are using e.g. vi. You have to wait 20 secs for a response.
As long as theres no IO, its's quite ok, though.
Apart from the bugs :-)
I have this one in 'w' since some weeks, dunno why:
udo@trash<31>w
5:17am %d day%s, 4:537126188, 6 hr|.&8, Runable processes: 4
Usertty login@ idle JCPU PCPU what what
root hft/0 Di7pm 6:12 45 1 -ksh
michaelv pts/2 2:22pm 7:33 18 1 -csh
udo pts/3 Mo2pm 30 0 0 -
udo pts/0 3:51am 0 0 0 -
As you can see, it's 5am - and I fought against that rubbish X-server
the whole night - sigh.
Buy a sparc ...
regards
udo
--
XmTeenageMutantNinjaMessageBoxGetChild(Widget,XmDIALOG_DEFAULT_BUTTON);pack@acd.acd.ucar.edu (Dan Packman) (02/02/91)
RE: reports of slow programs and difficulty porting to RS6000 versus
Sparcstation 1+
If the application used details of system internals or depended on features
not yet present in AIX (eg, diskquotas), then porting might be a problem.
For what it is worth, I have recently finished benchmarking a *real*
application [The application is a line-by-line infrared radiance calculation]
on several platforms. The results were (minutes:seconds):
CRAY XMP ~3:00
IBM RS6000/320 14:45
DEC 5500 23:26
Solbourne 5E/802I 51:30
Sparcstation 1 86:46
Standard fortran programs port easily. System throughput is excellent.
PS: Standard disclaimers.
Dan Packman NCAR INTERNET: pack@ncar.UCAR.EDU
(303) 497-1427 P.O. Box 3000 CSNET: pack@ncar.CSNET
Boulder, CO 80307 DECNET SPAN: 9.367::PACKpack@acd.acd.ucar.edu (Dan Packman) (02/02/91)
Oops. In my previous posting the machine was an RS6000/530,
not a model 320. Sorry.
Dan Packman NCAR INTERNET: pack@ncar.UCAR.EDU
(303) 497-1427 P.O. Box 3000 CSNET: pack@ncar.CSNET
Boulder, CO 80307 DECNET SPAN: 9.367::PACKpavlov@canisius.UUCP (Greg Pavlov) (02/03/91)
In article <1991Jan30.003004.74@intek01.uucp>, mark@intek01.uucp (Mark McWiggins) writes: > The most recent version of Unix Review contains a rather flamish > letter from someone who claims that his RS6000's don't perform > appreciably better in the application they're running than do > SparcStation 1+'s, and are much harder to port software to, as well. > "flamish" = what he said | how he said it ? - he's not the only one who has come to this conclusion. See comp.arch, among others - greg pavlov, fstrf, amherst, ny pavlov@stewart.fstrf.org