mark%umcp-cs@sri-unix.UUCP (07/03/84)
From: Mark Weiser <mark@umcp-cs> Subject: Workstation trends Newsgroups: net.works Distribution: net I am currently in the middle of a big workstation procurement. I have noticed a couple of interesting trends in workstations that I would like to share and get comments on. First, color. Quite a few of the newer workstations come with color standard, even to excluding B/W even as an option (Metheus). The older workstations are only B/W (xerox) or color as a second thought add-on (Sun). (Exception: color is well integrated into Apollo.) Is color the trend of the future even for non-CAD/CAM engineering workstations? Given that no one really knows how to use color except in specialized applications, color is not now a particular advantage in a general workstation. But it may be soon, if color becomes a standard part of unix systems and we learn how to use it. But for the next five years, is color worth the premium in cost ($7-10k) without a specific application? I don't think so. Second, multiple processors. A lot of workstations these days have multiple more-or-less general-purpose processors. (Not counting bit-sliced graphics do-dads.) Generally, one is general purpose and one handles the display. For instance: 3B2 (main processor and blit processor), Metheus (main processor, peripherals and paging processor, display processor), Symbolics 3600 (peripherals processor and main processor), Masscomp (main processor, paging processor, display processor). Graphics look very good on these systems, without the big pauses for process switching that one sees on a Sun when running multiple graphics jobs. Paging and heavy i/o may be a little better, but not more than 10%, I think. For color, apparently, these processors are more-or-less necessary. Xerox and Sun and Apollo seem behind the times in forcing their single processor to do all the work. True? My own opinion is that the extra processors won't help a bit in 99% of what a software engineering workstation is used for: compiles and editing and text processing and software building. That means that they are probably not worth their extra cost at this time.
mark@umcp-cs.UUCP (07/03/84)
I am currently in the middle of a big workstation procurement. I have
noticed a couple of interesting trends in workstations that I would
like to share and get comments on.
First, color. Quite a few of the newer workstations come with color
standard, even to excluding B/W even as an option (Metheus). The older
workstations are only B/W (xerox) or color as a second thought add-on
(Sun). (Exception: color is well integrated into Apollo.) Is color
the trend of the future even for non-CAD/CAM engineering workstations?
Given that no one really knows how to use color except in specialized
applications, color is not now a particular advantage in a general workstation.
But it may be soon, if color becomes a standard part of unix systems
and we learn how to use it. But for the next five years, is color worth the
premium in cost ($7-10k) without a specific application?
I don't think so.
Second, multiple processors. A lot of workstations these days
have multiple more-or-less general-purpose processors. (Not
counting bit-sliced graphics do-dads.)
Generally, one is general purpose and one handles the display.
For instance: 3B2 (main processor and blit processor), Metheus
(main processor, peripherals and paging processor, display processor),
Symbolics 3600 (peripherals processor and main processor),
Masscomp (main processor, paging processor, display processor).
Graphics look very good on these systems, without the big
pauses for process switching that one sees on a Sun when running
multiple graphics jobs. Paging and heavy i/o may be a little
better, but not more than 10%, I think. For color, apparently,
these processors are more-or-less necessary.
Xerox and Sun and Apollo seem behind the times in forcing
their single processor to do all the work. True?
My own opinion is that the extra processors won't help a bit
in 99% of what a software engineering workstation is used for:
compiles and editing and text processing and software building.
That means that they are probably not worth their extra cost
at this time.
--
Spoken: Mark Weiser ARPA: mark@maryland
CSNet: mark@umcp-cs UUCP: {seismo,allegra}!umcp-cs!markben%umcp-cs@sri-unix.UUCP (07/03/84)
From: Ben Shneiderman <ben@umcp-cs> I'm pleaseI'm pleased to participate in a discussion about workstation design issues. Color... looks good but is generally not that advantageous unless there has been ample effort to include it as part of the system (CAD/CAM etc.) If the bulk of the material is text then higher resolution, sharper letters, more screen space, and faster processing are the critical variables. Color is often misused, costs more, generates more heat, requires more volume, and is less reliable. I agree to push for high resolution B&W and get a large screen. The choice of single or multiple processors is secondary to the performance of the system for typical tasks. I think we should prepare a benchmark set of tasks and then measure the performance time for the benchmark. Other issues - advanced windowing concepts to allow concurrent processes and coordination across windows. Communications to allow cooperative problem solving across workstations (allow a user to send a windolw to a consultant, allow one user to execute a command on another workstation and both can watch the effect). Also...pointing devices (mouse etc.) or touchscreen. Also..graphics, animation, fonts. Ans while we're at it whynot Dvorak keyboards?
glennp%umcp-cs@sri-unix.UUCP (07/04/84)
From: Glenn Pearson <glennp@umcp-cs>
There's a certain chicken-&-egg problem to the use of color for workstations:
There's not enough software that uses the color effectively, and there's
not a common-enough availability of color equipment (except moderate
resolution home computers) to encourage software development. Mostly,
this has to do with price. For a particular resolution and precision of
image, color will probably always cost more than B&W; but its reasonable
to expect the cost differential for hi-resolution monitors to become
less significant in the future.
So for the present, if you're talking about a BUNCH of workstations,
why not put some eggs (to further mutilate this metaphore) in both baskets...
say, purchase 75% B&W and 25% color? Many of the vendors (Sun, Apollo...)
offer both types, so in theory you could avoid cross-vendor incompatabilities
if that consideration was important.
Also, in my opinion a workstation without a pointing device
{mouse, trackball, joystick, or touchscreen} is like a car with a starting
crank; vis, OLD TECHNOLOGY. I'm particularly fond of the mouse or trackball
choices.
Oops, I forgot bitpads. They're good too.
GlennMurray.pa@XEROX.ARPA (07/06/84)
Color is necessary for VLSI design and experimenting with color graphics. For these you want a very good tube, not just a warmed over TV set. Unless you are doing serious color work, I think you will be better off spending your money on bigger and better B+W displays. I think you are on the wrong track when discussing extra processors. Who cares how many processors a system has if the memory is the limiting factor? On the other hand, even when you are compling, an IO processor can save a few % of the main CPU by doing the trashy background tasks. Frequently you can save a lot of hardware by making a CPU work harder. This is easier to get away with if you have an idle IO processor. Some 68000 systems have 2 processors because of a design bug in the page fault area - if you got a nasty one there was no way to get restarted correctly. When a page fault happens, they suspend the first processor and run the page fault handler on a second processor. My keyboard has a 6802 in it. Does that count? That whole scene is very complicated. I suggest that you treat each system as a black box and see how well it does the things you are interested in doing and how much it costs. The software environment is much more important that the hardware. I've seen a 64K byte 8085 run rings around a machine orders of magnitude more powerful. On the other hand, the 8085 wasn't good for writing more 8085 code and there wasn't room for any more even if you wanted to write it. Figure out what you want to do, and take each interesting system for a test drive. If your procurement really is big, it's worth a day or so of your time to become familiar with each system. I second the suggestions that a mouse (or such) is essential. I'm addicted to my mouse. (I have not seriously used other pointing devices.) None of them will help much if all your software expects a TTY. Ben: You should be able to fake a Dvorak keyboard today. Just rearange the keytops and run each character through a translation table.
henry@utzoo.UUCP (Henry Spencer) (07/08/84)
Glenn Pearson observes, in part:
There's a certain chicken-&-egg problem to the use of color for workstations:
There's not enough software that uses the color effectively, and there's
not a common-enough availability of color equipment (except moderate
resolution home computers) to encourage software development....
Some of the potential customers recognize the problem. The original "spec"
for the CMU Spice project explicitly said that color and sound capability
must exist from the beginning so that the software would *use* them. If
you want to encourage use of color, you should *not* buy 75% monochrome
and 25% color, because there will be strong pressure to make the software
run on the lowest-common-denominator configuration, i.e. no color.
Personally, I agree with the original suggestion that color is not worth
the (current) price unless you have a specific need for it.
--
Henry Spencer @ U of Toronto Zoology
{allegra,ihnp4,linus,decvax}!utzoo!henrychabot@amber.DEC (Lisa S. Chabot) (07/15/84)
About color: what I've noticed when I'm sitting trying to work and somebody
brings in a tour group of three-piece suits, is that just about the first thing
question asked is "What about color?". If the tour guide is on her/his toes,
the response includes "What applications do you have for color?".
My opinion is that most of the companies represented by the three-piece suits
don't really have applications that demand color. (except, of course, for all
those engineers back home who wish they were doing animation for the movies :-)
But color is sexy. The three-piece suits are usually those sensitive to trends
in selling, and so they ask the question about color to determine if there are
any irons in the fire (and thereby determine that visited company is going to
keep up in the market and not die of stagnation, therefore becoming unable to
support the boxes they've sold). And color is a toy. Look at video games--
they're color (and they've been around for awhile, so color really should be
here now, right?). And _Tron_ and other movies haved also linked fantasy with
color graphics...flight simulators may not be a requirement, but show one as a
demo, watch everybody's eyes light up, and then wonder about what items in your
presentation tickled the buyers (i.e., is it the capabilities demonstrated by
the flight simulator (which is of course the admitted reason)(and is a very
good reason, too), or are you speaking to the fantasy of a desk-tied manager
or programmer).
Hope this doesn't sound like a denigration of color or color-box companies,
because that's not my opinion at all. It's more of a criticism of those who
shop for a box which does more than they need. On the other hand, if only
needed features were bought, then who would take care of engineers who really
*need* to fly a sopwith camel at 3am?
>Curse you, Red Baron!<
L S Chabot
UUCP: ...decwrl!dec-rhea!dec-amber!chabot
ARPA: ...chabot%amber.DEC@decwrl.ARPA
USFail: DEC, MR03-1/K20, 2 Iron Way, Marlborough, MA 01752ksbc@hirst1.UUCP (Annie Brooking KSBC) (07/19/84)
Personally I don't see any need for colour at the moment, the cost is just not worth the effort of writing the software to use it. As for multiple CPU's, having one controling the display is generally very useful, which is where the Apollo fails and the Xerox doesn't. Also if your gonna network this things, EtherNet etc, then another to organise the net is a good idea, which is where Xerox's bit-sliced CPU comes into action.
waynez@houxh.UUCP (W.ZAKARAS) (07/19/84)
What about COLOR BLIND engineers ??? Something to think about. WayneZ...