andys@men1.UUCP (Andy Sibre) (06/06/86)
HARDWARE REVIEW (sort of)
BELL TECHNOLOGIES' AT-CLONE ($4995) with:
72 MB (formatted) H.D., 24 ms. access 60 MB 1/4" tape streamer
6 Port (rs-232) board SCO SYS V Rel 2.12 (extra $)
8 MHz Clock 7 MB ram
1.2 MB floppy U.S.Robotics 2400 modem (xtra $)
& all necessary driver software
SPEED: with little serial I/O going on, it beats a 750 VAX in almost every
category. Floating point benchmarks poor until 80287 installed. With lots
of serial I/O going on, such as 3 other users catting bigfiles, speed
degrades noticeably. "Make"s take about as long on clone as on lightly
loaded VAX (1-2 users), again with the caveat that the clone not be getting
beaten up by heavy serial I/O. Serial board is of the extremely dumb type,
no on-board smarts, so use of smarter board (CompuTone, etc) might improve
things a bit. With 1 make, 3 edit sessions, and someone dialed in trying to
mail something to East Chickenswitch, machine is usably quick for all
concerned.
BACKUP: The tape is a wonderful thing. I know that no one else has ever
thought "rm temp*" and typed "rm temp *", but it's da**ed nice when all you
have to do is grab last nite's tape 'n say "tar x <whole_da**ed_directory>",
and actually HAVE it in the machine again within 5 minutes! The data rate
seems to be around 650K / minute when making a tar tape. Tapes readable by
Altos, & a number of other UNI-Boxes.
SERIAL I/O: As said, dumb cards run slow, but don't try to substitute their
own brilliant ideas for mine. (ahem!) The Bell card comes with their driver
which STILL seems to have a few aspects of interest to entymologists. First
and second releases of driver were a disaster, 3rd seems pretty stable if
terminal is decent. Works great w/ vt-100's, vt-220's, horrible with Tandy
dt-100's. Latter gets screen garbaged if I type while clone is painting
screen. Modem I/O perfect.
DISK: Bell sells a Toshiba 86 MB (unformatted) ST-506 interface drive that
is the finest I have ever seen. FAST !!! In 2 months of use, I have had NOT
ONE error of any kind on disk I/O. 4 sec. to copy 300K shar file from 1
filesystem to another. Bell sells w/ drive a modified ROM with altered
drive table entry that supports 830 cylinders and 10 surfaces, so if you get
SCO Xenix instead of MicroPort UNIX, you can still get a semi-reasonable
sized filesystem. Advertised average assess time = 24 millisec., and having
worked on a VAX w/ SI controller + Fujitsu Eagle, I BELIEVE IT !!! The
darned thing is fast, it doesn't make misteaks, and it's reasonably priced !
SCO XENIX: Everything you've heard about their compiler is wrong:
IT'S WORSE !!!
Example; mod.sources has a "TRC" expert-system builder. It compiles just
fine on a Tandy 6000 (Xenix 3.0, mc68000 CPU) and on a VAX (BSD 4.1) and on
an Altos 3068 (SYS V, mc 68020). Try it on SCO w/ default cc options, you
get thru the first few passes OK, then the loader sez "group DGROUP > 64K".
So like a dumbbell you give cc the -M2em (or -LARGE) options. As soon as it
gets past the pre-processor, it starts complaining about undefined symbols
(namely, every one you've used !) and crumps. Hmmm... Rumor Control has it
that SCO will release a compiler by 7/30 that actually works. We'll see.
SCO's disk buffering scheme seems to work very well, speeds up disk I/O very
nicely. Either because they don't know how their products work (when they
do, that is) or because they haven't figured ou