Re: workstation speed

[Kenneth Forbus <forbus@p.cs.uiuc.edu>]

>the hard numbers: we had an explorerII here for two weeks before AAAI.
>the simplest raw benchmark, compiling one of our products ~400k lines
>of lisp, takes 90 min on a 16mb expII, 4.5 hours on a 4MW (16mb) 3620
>(thus comparing apples, as they both have the same disks) and to fuel
>the 68k/unix/lucid/etc.. debate, 18 hours on a 8mb apollo 3000.

That's about right, I think.  On one of my qualitative simulators
(about 4K lines of code, with run times ranging from minutes to hours
depending on the example), I found a 16MB Explorer-II with their 512MB
disk about 2.8 times faster than a 3670 w/16MB & 474MB Eagle, running
6.1.  Unlike the first explorer, which broke when I poked around in
its innards during AAAI-84, this one responded well to user abuse  
(from the keyboard, that is).

The user-interface felt much, much snappier than my Symbolics machine
as well.  I'm now dreading Rel 7 conversion, given that windowing will
be even slower...I admit to being somewhat depressed by the path Symbolics
was taking.  Rel 7 feels like a cadillac.  What I want is a formula jeep
(i.e., extremely fast, but able to take hard jolts and knocks).

>Contrary to Don Mitchels report, the explorerII does not have a new
>disk controller, or disks. (they were going to upgrade to the 190mb
>maxxtors, but the larger capicity drives are too shock sensitive, and
>the QA people won't let them use them until they pass muster) 6-8
>months ago they did start selling a SMD controller and drive, (tho we
>didn't get one for our trial) which would make things even faster.
>(maxxtor 140 drives and the scsi interface are blindingly slow,
>compare a 3620 to a 3650 on a problem big enough to page, the only
>difference between the two is the disk drives (and box))

Well, the Explorer-II I typed on at AAAI didn't use a Maxtor, and their price
list includes non-Maxtor disks.  The lower-price configurations still use
them, but the more expensive ones don't.

>The biggest negative of the explorer(s) (of which the lambda is also
>guilty) is the small address space, (yes we have bumped into it)..

It certainly is a very short-sighted decision.  Marketing people at TI
and other companies have been saying that they don't see the demand
for more yet.  I expect part of the problem is the marketing hype of
the stock hardware/AI on PC vendors, who are trying to convince people
that "small is better".  But what you can get by with today is not
what will let you get to what you want to do tomorrow.

Just yesturday I ran into a collegue who bought a $5K Sun, having been
told it would run Lisp.  Now he is buying an $18K Sun, still diskless,
but one that has a bus so he can go beyond some small amount of RAM.
And he is still spending most of his time worrying about how to make
his code fit in such a small machine.  I find this utterly horrifying,
precisely because I see it over and over again, at many places.  We
are experiencing Gresham's law, applied to symbolic computing.

>The explorerII list price is currently higher than the base 3620
>system, (36k vs 45k) of course when you adjust memory sizes (the
>explorerII comes with 16mb) the prices are not all that different, and
>TI is a lot more aggressive about discounts.  (basicly a TI
>salesperson has a pre authorized discount schedule, with 'bolics all
>exceptions to list had to be appoved from above)

And worse yet, at Symbolics the prices don't always stick.  I had
negotiated a fixed price for 3640's when they came out that was, let's
say, pretty reasonable.  In writing.  I went out and raised over
$0.25M based on that price.  Symbolics then came back and said that,
unless I paid $8K more per machine, they would delay delivery "for
several years".  Given that my department was not exactly convinced
that these machines were a good idea anyway, this behavior was the
kiss of death.  I'm the only one who has kept buying Symbolics
machines (I'm a hostage to performance), everyone else has bought Suns
(or in some cases, Explorers).

>I was initially somewhat embarassed to be coming to TI's defense,
>given all my friends at symbolics, the 3640 on my desk, all the
>cursing I did when it was first time to move our software onto their
>machine, (pre-release H of the operating system), and the general
>sluggishness of a 4mb explorerI; but they have done a very good job on
>both the new hardware, and the new release of the software, and while
>I have complimented some of them privately, some public recognition is
>in order.
	
Add my voice to this.  I still view our Symbolics machines as crucial
to getting our work done, and a fine investment.  For the past few
years they have given us the best environment we could find, bar none.
However, TI has turned the "serious" lisp market into a real horserace
with the Explorer-II.  Given that Symbolics has fallen behind on
performance, the next machines I'm buying are probably going to be
Explorer-II's.
  
I don't see Symbolics out for the count, however, as some people seem
to think.  Everyone has been deeply disapointed that Ivory has taken
so long.  But it seems to be coming, so maybe a year or two from now
Symbolics will be back in the lead in the high-performance market.
It's an empirical question.