#MS-DOS
15 messages in this thread
Jamie,
I've been perusing my April `88 copy of CADalyst with unusual interest,
since reading the `Blips on the CRT' column written by David Cohn. In it he
implies that he is privy to info that suggests MS-DOS is not long for the
ADESK camp once Release 10 hits the streets. While this is no shining
revelation, the 640K barrier being an obvious limitation, I would solicit
your comments on his comments. While the company line may be a tight-lipped
response, forewarned is forearmed, and I for one would like such a change
over to be as painless as possible. Maybe Cohn is just blowing smoke, either
way I don't want to be treated like a mushroom. If you don't keep us in the
dark, and Cohn isn't feeding us any crap, then changes be a comin'. If this
is the case my friend, is there any way to find out what the new operating
system is going to be so I can bone up on it? `Igots ta know man, I just gots
ta know'.
Was that the sound of a can of worms being opened?
Thanx, Dennis
We all would like to know what the next OS will be. There is a belief that it
will be OS/2 but there are some hot challengers. We will indeed make it as
painless as possible and forewarn you as far in advance as we can.
The facts are, we haven't made ANY hard core decisions and perhaps he
overheard rumors or "personal" personnel opinions. So to base any actions
on such information would be VERY dangerous.
For now you can trust the fact that 10 is going to be on MS-DOS and from
there, who knows!
Jamie
Not to belabor the subject, but will there be a version or "mode" which
specifically supports the '386 and/or the Weitek 1187 numeric coprocessor
card? B.W.
B.W. —
We have no plans currently to release an AutoCAD that supports the
Weitek, because our tests indicate that we would not realize a significant
speed increase over the 80×87 versions. We have thought about introducing
a 386-387 specific version of AutoCAD, but again, our tests indicate a
speed gain only in the vicinity of 15 to 30% We're not sure we should
commit the resources neccessary to bring a separate product (in the same way
that the Sun, VAXStation and Apollo versions are separate products) for such
a minimal increase in performance.
— Brad
..
Jamie, Only 15-30 percent, I have customers who would kill for that!! If your
taking votes I'll say it again, even if it costs extra a 386/387 version
would be most welcome. My new systems sales are almost 3 to 1 386/387 even
with the much higher price.
Corky Deeds LightSpeed Corp.
Brad, Corky – ABSOLUTELY! 15-30% is nothing to sneeze at! -ds
.
Don —
Again, let me say that my figures of 15-30% for an 80386-80387
build of AutoCAD are incorrect. Duff's initial test results indicate a
possible increase in the 10-15% range, which I don't think justifies a new
build of AutoCAD specific to that chip combination.
In any case, I have no doubt that the vast majority of DOS-based
AutoCAD workstations would realize some major performance gains, far
beyond the 15-30% range, if they were optimized for performance. The
number of systems without BUFFERS=30 (or higher) statements in the
CONFIG.SYS file, without sufficient ext/exp memory for a reasonable
amount of extended I/O page space, without fast math co-processors or fast
hard disk drives, and without optimized disk drives absolutely boggles my
mind and raises my blood pressure.
Besides, can someone please tell me what in the heck is this
fascination with hardware as a solution to everyone's CAD problems as
regards design throughput? I maintain that a client's cash is much better
spent first on high-quality training for the operators and managers who
will use the systems, and second on the fastest hardware this side of the
Pecos.
The point I'm trying to make is that there are a number of ways that
dealers and customers can improve AutoCAD performance right now, without
depending on us to release a chip-set specific version.
— Brad
..
Brad:
Here, Here!
It seems like half of the prospects we deal with already have some form of
AutoCAD. Most of the time the version is out of date by six months, they
have no config.sys file, and all of their files are in the root directory!
One company contacted us because they wanted to use some 2000 drawings to
drive thier CNC controled machining centers. BUT ALL OF THIER DRAWINGS
WEREN'T DRAWN TO SCALE!!!!!!!
Never enough time to do it right, but always enough time to do it twice!
<ME>
Brad – The numbers I've seen indicate 10-15% improvement for a
386/387-specific version, not 15-30%.
.
Duff —
Whoops. My mistake. I was trying to recall a previous thread on
this topic without researching the message base, and unfortunately my
memory proved faulty.
Sorry about that, folks.
— Brad
Duff – I don't understand…… only 10-15%? We were promised *MUCH*
more by the folks at Intel, et al……..
…
Are there no 386 compilers mature enough to provide the performance
increases we were "promised"? Or have I just been reading the wrong
magazines? -ds 😎
.
Don – For certain programs, perhaps the claims you heard are accurate. For
AutoCAD as a whole, however, such does not appear to be the case. AutoCAD is
already using inline floating-point code. The benefits I see to making a
386/387-specific version would be a slight reduction in code size and a
slight speed increase as the result of removal of the "wait" instructions
needed for the 8087, and the (possible) use of the 80387's trig functions.
However, none of this assists in the I/O area, or in integer calculations,
both of which are also high on AutoCAD's execution profile. We'll consider
creating a 386/387-specific version, but (unless we were to drop the older
machines — and I DON'T THINK WE'RE READY FOR THAT) the pros don't appear to
outweigh the cons at the moment. The biggest problem: this would be a "new
machine build" from the standpoint of our Build and QA groups, doubling the
testing time for the DOS versions of AutoCAD (at least the first time).
.
So then really Duff, you are saying that our super-duper '386 boxes
are still not going to be much more than a fast XT as long as DOS
prevails….. (and the AT buss, and ST506 hard drives, etc. etc…)
…
DARN! Well, we'll just sit back and wait for a Unix version.
…
Thanks for the info.
-ds
.
Don – The '386 does have its good points! Even with today's version of
AutoCAD (and other software) designed for the 8088, 8086, or 80287, the '386
really flings those bits around. One thing to keep in mind is that both the
'286 and the '386 support extended memory, which will make it possible to run
programs larger than 640K on OS/2. While extended memory is useful on DOS
for storing large amounts of data, as in AutoCAD's paging area, programs
can't execute from it without additional OS support. LIM Expanded memory,
basically, is a kludge to allow CPUs without protected mode (8088, 8086) to
access more data. To use extended memory, the CPU must be switched to
protected mode; the '386 has an efficient way to switch back to real mode,
the '286 does not and must be reset to get out of protected mode. Using one
of the "DOS extender" packages, it may be possible to utilize extended memory
on a '286 or '386 for program code AND data, and we're looking into this
along with numerous other options. Stay tuned.
.
We have been considering the concept but have not made any decision. It still
remains a wishlist item. As Brad said, we will not be supporting the Weitek.
Jamie