home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!pipex!bnr.co.uk!uknet!doc.ic.ac.uk!cc.ic.ac.uk!imperial.ac.uk!vulture
- From: vulture@imperial.ac.uk (Thomas Sippel - Dau)
- Newsgroups: comp.sys.sgi
- Subject: Re: Indigo R4k/IBM 6k comparison
- Message-ID: <1993Jan11.124442.16695@cc.ic.ac.uk>
- Date: 11 Jan 93 12:44:42 GMT
- References: <8378@news.duke.edu>
- Sender: vulture@carrion.cc.ic.ac.uk (Thomas Sippel - Dau)
- Reply-To: cmaae47@imperial.ac.uk
- Organization: Imperial College of Science, Technology and Medicine
- Lines: 58
- Nntp-Posting-Host: cscgc
-
- In article <8378@news.duke.edu>, rla@canctr.mc.duke.edu writes:
- - Greetings. We are considering a purchase of a set of 4 workstations to be
- - linked together. Objectives are mainly processing (number-crunching) power
- - plus the ability to run the GCG DNA sequence analysis suite of programs.
- - The two remaining candidates are SGI (R4000-based) vs. IBM RS6000s.
- -
- - Anyone with experience on both platforms, including service and reliability
- - and UNIX performance, please let me know your thoughts. Overall system
- - speed is a major consideration. The IBM is not yet GCG-ready, but a
- - beta-version will supposedly be ready soon.
- - Thanks.
-
- On number crunching, the RS/6000 chip will be about five times as fast as
- the R4000 on a cycle basis. This means that for a program that has
-
- do 10 i = 1, n
- 10 sum = sum + x1(i)*y1(i) + x2(i)*y2(i) + x3(i)*y3(i) + x4(i)*y4(i)
-
- as the inner loop where it spends its time, the R4000 will spend about
- 100 cycles on that loop, while the RS/6000 will do it in about 20. You can
- therefore expect a 100 MHz R4000 to need 2.5 times as much time as a 50 MHz
- RS/6000. The actual numbers are data dependent, because the RS/6000 needs
- a cycle more for a floating point instruction if the result after
- normalisation is 0.
-
- Having said that, if you want to do DNA sequence analysis, than I guess you
- need fast disk access and generate graphics interactively with good response
- times, and unfortunately that is where the IBM machines do not shine.
- If there is a number crunching job with reasonable paging in the background,
- the IBMs I have used here feel decidedly heavy going, while you hardly
- notice it on an Indigo.
-
- Regarding service and reliability, both systems come with a fair sprinkling
- of bugs and features you may find undesirable, and hardware reliability
- is similar (especially that of third party disks and memory :-). It more
- depends on whether the actual bugs bite you or leave you alone. The "I wanna
- do it exactly as on my standard BSD 4.x system" approach is guaranteed to
- lead to tears on both.
-
- IBM has tried to make the system safe by conservative design, while SG often
- takes a more aggressive or "risky" approach - maybe it is just that IBM
- systems take longer until the organisation ships them, the old big firm
- vs small firm argument.
-
- We have however found in several cases that when a user maintained system
- suffers an unexplained failure which causes people to react in panic and destroy
- whats left of it by mistaken recovery attempts, then the IRIS will normally
- be up again by the third day, the IBM by the fifth. Notabene this is from
- a small number of incidents.
-
- Hope this helps. Thomas
-
- --
- *** This is the operative statement, all previous statements are inoperative.
- * email: cmaae47 @ ic.ac.uk (Thomas Sippel - Dau) (uk.ac.ic on Janet)
- * voice: +44 71 589 5111 x4937 or 4934 (day), or +44 71 823 9497 (fax)
- * snail: Imperial College of Science, Technology and Medicine
- * The Center for Computing Services, Kensington SW7 2BX, Great Britain
-