🚀 go-pugleaf

RetroBBS NetNews Server

Inspired by RockSolid Light RIP Retro Guy

Article View: comp.sys.next.misc
Article #3087

Re: Performance comparison with SUN

#3087
From: Magnus Nordborg
Date: Fri, 16 Oct 1992 05:51
69 lines
3677 bytes
In article <1992Oct13.153938.19926@yang.earlham.edu>  writes:
> OK, I know that this has been trodden through before, but I'd like some  
numbers
> 
> (or a number) comparing the performance of a turbo color NeXTstation (or
> NextDimension) with a Sun 10. I have a poor misguided faculty type  
person who
> can just about handle System 7 but who is desparate for the most  
powerful
> machine in the Universe. No amount of reasoning can persuade him that  
what he
> gains in program execution time he will definitely lose in system
> administration time. So what I need to know is, approximately how many
> Meaningless Indicators of Performance (MIPS) does the NeXT turbo  
generate
> compared with a mighty Sun 10 (or whatever)?
> 
> --
> Mark H Pearson              Coordinator of Academic Computing
> Earlham College, Richmond, In 47374
> regular e-mail:         markp@yang.earlham.edu
> NeXT mail     :         markp@cranham.cc.earlham.edu

We don't have any Sparc 10's but we do have a number of Sparc 2's.  A NeXT  
Turbo is approximately equal to the Sparc 2. I belive the Meaningless  
Iteration Performance Statistics are 25 vs 28 or so.  Better stats such as  
Specmarks usually give the Sparc a bit more of an advantage. The  
difference is noticeable (but pretty unimportant) when doing large  
compilations, and when running numerical C-programs.  The diff is more  
important when lots of disk i/o is involved due to lousy SCSI performance  
of NeXT.  

However, what really matters is the speed with which the machine does what  
you want it to and here the NeXT has a gigantic advantage.  Most of the  
nice, friendly apps that run on the NeXT are available as horrid  
X/Openwindows apps on the Sparc, but due to the complete inferiority of  
the interface, a comparison is not possible. The only software we  
regularly use on both is Mathematica, and when I recently ran the  
"official" benchmark suite for it, the NeXT won!

The conclusion in this lab (we all work in population genetics / applied  
math) is that the Sparc 2 and the NeXT are practically equivalent in  
speed, so that the main differences are:

1.) The NeXT is a pleasure to work on, has great bundled software,  
excellent TeX, etc, whereas the Sparc 2 has nothing.

2.) The Sparc 2 is twice as expensive.

With a Sparc 10 the speed diff will be considerable (depending on which  
model you get -- they start around 70 MIPS).  However, DO ask yourselves  
if you really need speed that much.  You are absolutely right about the  
sys admin overhead -- not only is a Sparc 10 much more expensive, you will  
also need professional Unix help from time to time whereas NeXT's can be  
administered by grad students.

I would NEVER buy a Sparc 10 as only machine. If I had lots of money to  
start a lab, I might buy 10 NeXTs (say) and one Sparc 10 to use as a  
computer and file server. I would put it in the basement so that no one  
had to see its ugly interface.

P.S. If you do decide to go for speed, buy an SGI or HP instead!
--


		Magnus Nordborg
		Department of Biological Sciences
		Stanford University
		magnus@fisher.stanford.edu (NeXT mail preferred)
Technical Details
Message-ID: <1992Oct16.055119.26016@leland.Stanford.EDU>
Path: arc1-ca.pugleaf.net!sparky!uunet!think.com!ames!agate!stanford.edu!leland.Stanford.EDU!news
References: <1992Oct13.153938.19926@yang.earlham.edu>

Show Full Headers
Complete Article Headers
Newsgroups: comp.sys.next.misc
Path: arc1-ca.pugleaf.net!sparky!uunet!think.com!ames!agate!stanford.edu!leland.Stanford.EDU!news
From: magnus@Fisher.Stanford.EDU (Magnus Nordborg)
Subject: Re: Performance comparison with SUN
Message-ID: <1992Oct16.055119.26016@leland.Stanford.EDU>
Sender: news@leland.Stanford.EDU (Mr News)
Organization: DSO, Stanford University
References: <1992Oct13.153938.19926@yang.earlham.edu>
Date: Fri, 16 Oct 92 05:51:19 GMT
Lines: 69