svehla@zeus.unomaha.edu (09/28/90)
Hello I am searching for information pertaining to the IBM R6000 line of computers. What have been your experiences? Do you have any concerns? Can you provide me any benchmarks? What about AIX? Any information you can provide me on these questions and on tuning AIX would GREATLY BE APPRECIATED!!! Thanks in Advance! Ernie Svehla
jgreco@archimedes.math.uwm.edu (Joe Greco) (09/28/90)
In comp.sys.ibm.pc.rt article <1731.27028f1f@zeus.unomaha.edu>, svehla@zeus.unomaha.edu wrote: :Hello : :I am searching for information pertaining to the IBM R6000 line :of computers. What have been your experiences? Do you have any :concerns? Can you provide me any benchmarks? What about AIX? :Any information you can provide me on these questions and on tuning :AIX would GREATLY BE APPRECIATED!!! : :Thanks in Advance! : :Ernie Svehla Well, the 6000 is one fast little beastie. Since we here deal mostly with floating point math, I devised a little program quite some time ago and timed it on a number of different systems.... here's what I got. Lower numbers are better. Convex C220 convex.csd.uwm.edu 8.1 seconds RS/6000/530 thales.math.uwm.edu 8.8 seconds RS/6000/320 thales.math.uwm.edu 11.0 seconds DEC/3100 berkeley DEC 25.1 seconds Solbourne 5/600 miller.cs.uwm.edu 33.3 seconds IBM RT 6151/125 archimedes.math.uwm.edu 100.6 seconds Unisys 7000/40 csd4.csd.uwm.edu 131.2 seconds IBM 6152 RT banach.math.uwm.edu 165.5 seconds The Solbourne and the Convex are both dual processor machines; if you have two processes running, you can halve their numbers. These systems were all running some variant of BSD UNIX (SunOS, Convex UNIX, 4.3 tahoe, AOS). We were quite impressed that it rated so close to the Convex. However, I'm far from pleased with AIX. It tries to do too much. They've added so many layers of complexity that I'd really just like to scream. I haven't even begun to figure out how the swapping and process scheduler interact. The other day, I let a csh session sit for about 45 minutes. I typed "logout" and it sat there for one minute, sixteen seconds before it did anything. Not sure if anything else was running on the machine, but that's ridiculous. Little things don't work. tset bombs the hft, since it sets -opost and -ff1. I had to hack around that little joy of a problem.... xinit sometimes exhibits that behavior as well, and the X server crashes from time to time. When it does crash, we haven't figured out how to reset the display! It doesn't go down automatically to text mode like the RT does. It just sits there, dead bitmap display. You can bring xinit back up and bring X down, but it's still locked in bitmap mode. The crashes happen more often when you're running X11R4 clients from another host. I'm not pleased with "aixterm" either. I like the enhancements, but it just won't work correctly with a standard BSD "xterm" termcap. I've messed up a couple files in vi on remote machines. There may be solutions to all of these X problems, but I'd prefer not to have to go digging in the first place. IBM's also distributing an earlier release of X (note the famed Xeyes bug with long, skinny Xeyes windows, etc)... Even little things like ps. I'd like a standard BSD ps. I know all the flags already :-) Oh well. It's a good number cruncher though! I just wouldn't buy one for home. ... Joe ------------------------------------------------------------------------------- Joe Greco - University of Wisconsin, Milwaukee - Department of Mathematics jgreco@archimedes.math.uwm.edu USnail: Joe Greco Voice: 414/321-6184 9905 W. Montana Ave. Data: 414/321-9287 (Happy Hacker's BBS) West Allis, WI 53227-3329 ICBM: 43 05 20 N 87 53 10 W #include <witty_and_humorous_saying.h> Disclaimer: I don't speak for the Math Department, the University, or myself.
staff@cadlab.sublink.ORG (Alex Martelli) (10/01/90)
jgreco@archimedes.math.uwm.edu (Joe Greco) writes: ... >However, I'm far from pleased with AIX. It tries to do too much. A fair and concise assessment! As a SW developer, this functional richness has hit me hard in just one place, but a KEY one - linking. Don't DREAM of doing serious sw development with just 16 meg of RAM, your links will CRAWL. If you can afford to stuff the beast with expensive ECC memory, I believe it would be a great environment (MOST unices are rich and complex nowadays... in the workstation business I know of only Sony selling a reallly-vanilla bsd 4.3, and that's not exactly MINIMAL for one coming from an USG background, you know!). -- Alex Martelli - CAD.LAB s.p.a., v. Stalingrado 45, Bologna, Italia Email: (work:) staff@cadlab.sublink.org, (home:) alex@am.sublink.org Phone: (work:) ++39 (51) 371099, (home:) ++39 (51) 250434; Fax: ++39 (51) 366964 (work only; any time of day or night).
mccalpin@perelandra.cms.udel.edu (John D. McCalpin) (10/03/90)
>>>>> On 30 Sep 90 18:48:36 GMT, staff@cadlab.sublink.ORG (Alex Martelli) said:
Alex> Don't DREAM of doing serious sw development with just 16 meg of RAM,
Alex> your links will CRAWL. If you can afford to stuff the beast with
Alex> expensive ECC memory, I believe it would be a great environment
The ECC *memory* itself is not expensive. The rather extensive
circuitry that IBM puts on its memory boards to do ECC (SECDED),
bit-swapping, and chip-swapping is most of the cost.
There are 3rd parties currently developing SIMM's to fit into the
slots on the memory boards that IBM sells. These will be *much*
cheaper than IBM's offering -- maybe $150/MB vs $600/MB (list) or
$420/MB (educational discount) from IBM. I will post more info when I
get my board populated and my supplier tells me that he is ready to
mass produce.
There is also a significant possibility that IBM will lower its prices
when they announce the support of 4 Mbit technology in its memory
boards. As I understand it, the boards already support 256kbit, 1
Mbit and 4 Mbit chips, but IBM is not yet ready to release the 4MByte
SIMM's for those boards --- maybe reliability trouble, but more likely
just a question of availability....
--
John D. McCalpin mccalpin@perelandra.cms.udel.edu
Assistant Professor mccalpin@vax1.udel.edu
College of Marine Studies, U. Del. J.MCCALPIN/OMNET
bass@cs.utk.edu (Vance Bass) (10/09/90)
In article <6654@uwm.edu> jgreco@archimedes.math.uwm.edu (Joe Greco) writes: >In comp.sys.ibm.pc.rt article <1731.27028f1f@zeus.unomaha.edu>, svehla@zeus.unomaha.edu wrote: >xinit sometimes exhibits that behavior as well, and the X server crashes >from time to time. When it does crash, we haven't figured out how to reset >the display! It doesn't go down automatically to text mode like the RT >does. It just sits there, dead bitmap display. You can bring xinit back >up and bring X down, but it's still locked in bitmap mode. The crashes >happen more often when you're running X11R4 clients from another host. It appears that the fix tape which is newly available corrects this problem. If you have not received this tape (or diskettes or whatever), call the IBM software support number and request it. -- Vance Bass I work for IBM. My opinions are my own. IBM The facts speak for themselves. Knoxville, TN I try to err on the side of the latter, if at all.