Hardware & Software improvements (was: historical note)

Paul Fernhout pdfernhout at kurtz-fernhout.com
Sat Jun 23 15:41:51 UTC 2001


Tim Olson wrote:
> Bob Jarvis wrote:
>> I'd like to ask two related questions:
>>Why and how does hardware get faster?
> Hardware (specifically, the microprocessor) has achieved a factor of
> roughly 500x performance improvement over the past 20 years.  By far the
> largest factor in this is raw clock frequency, going from ~16MHz to
> greater than 1GHz (60x).  Most of this is directly due to process
> improvements ("process" here means the feature size and series of steps
> used in fabricating devices on the silicon wafer).
> 
> As more and more transistors were able to be packed into the same area,
> due to the same process improvements that gave us higher frequencies,
> microarchitectural improvements were able to achieve another factor of
> 10x.  These improvements, such as pipelining, superscalar and speculative
> execution, exploit various levels of parallelism, and trade off increased
> die area for decreased time.  Larger caches also exploit spatial and
> temporal locality found in almost all code.

Tim-

Not to quibble with the thrust of your argument, but I see more like a
10,000x to 100,000x speedup at a constant dollar cost over the last 20
years especially if one looks at system costs (although your figures may
be closer for the same chip area). For example, around 1980 I paid $1000
for a 1Mhz Commodore Pet with around 16K (and only a couple years
earlier a little less for a KIM-1 with only 1K). Now $1000 will buy me
around a 1.0+Ghz system with 128MB of memory. So CPU speed is 1000, plus
a 10 to 100 fold (or more) improvement from other factors you mention
(including being able to do execute complex operations like floating
point calculations in a single clock cycle) resulting in a 10,000x -
100,000x speedup. Memory is also up 10,000 times (and is faster). Back
then I bought a dual floppy drive for about $1200 that could store a
total of around 360K or  0.36M. Now, that amount of money can buy about
close to terrabyte of disk space storage (1,000,000M), so about a
1,000,000x increase. Communications bandwidth is harder to measure, (and
with modems has held constant longer) but we are in for some major
sudden increases as DSL and fiber and 3G wireless become more available.
People in the lab have demonstrated over 1 Terrabits per second per
optical fiber and expect continued imporvements.

I expect the improvements to computing power at a constant cost in the
next twenty years to be closer to 1,000,000x or more, as essentially the
rate of increase is also increasing. Especially for bandwidth when we
get fiber to the home. Even if limits are reached for current
technologies, there are many other technologies waiting in the wings (3D
chips, optical computing, DNA computers, nanotech computers, quantum
computers, self-assembling computers) which will likely keep the
exponential trend moving. In fact, this has already happened about five
times in computing to keep up exponential growth with a progression from
relays to tubes to discrete transistors to ICs to VLSIC and beyond.

See for example:
   http://www.kurzweilai.net/articles/art0134.html
   http://nanodot.org/articles/00/06/28/0045201.shtml
   http://www.transhumanist.com/volume1/AI_power_075.jpg
   http://news.cnet.com/news/0-1006-200-6343049.html?tag=owv

It is common go think in linear terms instead of exponential terms.
Kurzweil talks about this in detail in his essay. Kurzweil and Vinge
think we are heading (without unexpected disruption like war, plague,
etc.) for a post-scarcity technological singularity around 2020-2040 as
$1000 worth of computer power greatly exceeds the computational power of
a human brain and then soon all human brains.
     http://www.ugcs.caltech.edu/~phoenix/vinge/vinge-sing.html
I've also written an essay on this topic as well:
  http://www.bootstrap.org/dkr/discussion/0126.html

My take on what drives "Why ... does hardware get faster?" is that such
progress is the result of a technological arms race between capitalistic
corporations (and individuals) on several motherhood-and-apple-pie
fronts (lowering costs, better products to consumers, better market
analysis, better R&D via simulation). The mindset is that organizations
and individuals need better technology to more effectively compete as
their competition similarly ratchets up in computer "arms" levels.
That's one reason such technology has had limited benefit in the
developing world (the "Digital Divide"). The technology isn't being
created today mainly to be about helping people -- it's about fighting
an (economic) war (even though visionaries like Alan Kay, Doug
Engelbart, and Ted Nelson and others may have had other values in the
1960s and likewise people may have an interest in free software or open
source software like Squeak for similar reasons).

I'm not saying this technological arms race is good in and of itself
(although it does open new possibilities -- space settlement, flexible
cheap manufacturing, robots to assist with drudgery, better services for
the ill or handicapped). Rather, I feel a major issue facing the future
of computing is to figure out a way to transcend technological arms
races because ultimately one usually can't win them without destroying
one's way of life (and likely everything else). For example, several
native American tribes finally did adopt the methods of their genocidal
invaders (horses, guns, capitalism, deceit, slaughter) but it was very
damaging to their own way of life. [See Howard Zinn's works like "A
People's History of the United States"]
  http://www.people.fas.harvard.edu/~sachs/Zinn/online_works.html 
or "Lies My Teacher Told Me: Everything Your American History Textbook
Got Wrong" by James W. Loewen
 http://www.uvm.edu/~jloewen/
With such a history of what happens as a result of advanced military and
technological power in the hands of morally corrupt (or at best amoral)
profit driven individuals and organizations meeting groups without such
military power, 
  http://www.dsl.org/copyleft/11-rules.shtml
there is much that needs to be done to ensure a future for humaneness
and humanity in the face of organizations soon to have supercomputers a
million times faster or smaller (think super-"smart" bombs) if they are
still primarily driven primarily by a profit or profit-defense motive.
That's one problem I have with U.S. military policy -- I want the
defense establishment to be told to by government to defend me from arms
races (including technological ones), whereas they are told by
government to win them (which thus makes the situation worse). :-(
Theodore Sturgeon's short story "The Skills of Xanadu" is a good
illustration of a society who has at least locally transcended the
technological arms race. James P. Hogan's "Voyage from Yesteryear" is
another. Both show what happens when a post-scarcity society meets a
scarcity society, so perhaps there is room for some hope -- assuming
such a post scarcity society is allowed to come into existence in the
first place.

There is a downside to the current approach of technological funding in
the US producing these continuing increases in certain forms of
computing capacity. Many possible futures could be invented, but the
path often chosen is the one of least resistance -- short term profit --
which is kind of like trying to climb the mountain of human spirituality
by picking the easy path down into the valley. The best one can do is
try to find ventures which both help humanity and make short term
economic sense -- but there needs to be a consciousness of making a
decision about what to innovate and an acceptance that short-term
profits might not always be maximized (even if one might still survive
economically). "Excitement" is another bad reason to pick what one
innovates on -- one needs to pick more a convergence of excitement,
humane need, and fundability. Yet, "excitement" is driving much of
technological change, sort of like betting one's house on every role of
the dice -- exciting, but probably not sensible. One of my wife's
favorite books is "Finite and Infinite Games" by James P. Carse
  http://www.wunderland.com/WTS/Jake/fig.html
  http://www.alamut.com/subj/artiface/games/infiniteGames.html
I would argue it is important for individuals and organizations to start
playing infinite games, not finite ones, especially in the context of
technological change. I just saw the movie "Atlantis" last night and I
thought the ending [hopefully not a spoiler considering this was a
Disney film] was an excellent example of some people deciding to
choose the infinite game route in the face of seemingly more profitable
short-term choices relating to finite technological advancement. Way to
go Disney!

However, the only easy way the old guard can maintain control through
scarcity economics in a post-scarcity world (implied by these trends in
computing and other technology) is by draconian legal controls 
  http://freeipx.org/display.php3?id=68
  http://slashdot.org/articles/01/06/21/2010249.shtml
to enforce economic scarcity (think a "Copying War" like the "Drug War"
described in the movie "Traffic" but magnified 100 times). The DMCA
making copying a felony is just the start. Bob Young from above: "We
then approached the question from another angle, saying: if the Open
Source movement failed what would cause it to fail? The answer is that
if left to market forces Open Source will inevitably prevail, simply
because the consumers prefer it over the legacy proprietary binary-only
model. But the technology markets were not being left to market forces.
The government is playing an increasing role at the urging of the major
global publishing organizations. ..."
Microsoft is just starting to flex its muscles in this regard:
  http://news.cnet.com/news/0-1003-200-6352301.html?tag=mn_hd
  http://www.zdnet.com/zdnn/stories/comment/0,5859,2668713,00.html
  http://content.techweb.com/wire/story/TWB20010501S0002

We will be in for a wild ride of conflict between community and vested
special interests also driving technological and social change. The
"Drug War" has over the last twenty years added around 1,000,000 people
  http://www.drugwarfacts.org/
to the US prison population without much of an outcry (at least in the
US -- in fact, politicians and others in the US have built their careers
on promoting the drug war), so if a few million computer users are
imprisoned in the US for non-violent "illegal" copying, who will raise
an outcry? (Especially when raising an outcry may make you a target,
just like any politician who speaks out on the Drug War will be labelled
as "soft on crime".) Of course, such laws are often selectively
enforced, so they provide yet another opportunity for censorship,
authoritarianism, and corruption. For example, here are a couple of
college students facing felony (not civil) charges under DMCA for
distributing software:
  http://www.marist.edu/resnet/unions.html
While I don't condone their actions, the deeper issue is we are creating
a society where sharing is not encouraged but instead is illegal:
  http://www.gnu.org/gnu/thegnuproject.html
  http://www.gnu.org/philosophy/why-free.html
  http://www.gnu.org/philosophy/right-to-read.html
How soon is it before police get civil forfeiture laws allowing them to
confiscate computer equipment and auction it with the proceeds to their
own budgets on just the suspicion of illegal copying, the way they can
with cars and such on even the suspicion of drug use? Considering how
much of the technology and content we use is generally either funded
directly or indirectly by grants or tax subsidies (e.g. R&D tax credit)
often through non-profit organizations like universities, this is all
very ironic. (For example, the patents to MP3 are held by a
"non-profit" organization).
  "Corporate-Sponsored Research Untrustworthy"
  http://slashdot.org/science/01/06/20/1750231.shtml
  "Can University Students GPL Their Submitted Works?"
  http://slashdot.org/article.pl?sid=01/06/18/1622204

Business (and academia) doesn't have to be like that. William C. Norris
in the 1970s
   http://www.digitalcentury.com/encyclo/update/william_norris.html
talked about the responsiblity of business to "meet society's unmet
needs" either alone or through true public-private partnerships. That is
ultimately why the state allows businesses to be chartered. From that
page: "It has been [William C. Norris's] lifelong dream to use corporate
responsibility to eliminate poverty, unemployment, and hunger, and to
revolutionize the educational system to provide the most opportunity to
the most people. Though often misunderstood and sometimes ridiculed for
his beliefs, Norris has had a profound effect on corporate America’s
social conscience." Corporations also need to remember that working in
the public interest can have a big effect on employee morale.

Langdon Winner writes in his 1977 book "Autonomous Technology:
Technics-out-of-Control as a Theme in Political Thought" about the
importance of open systems understandable and modifiable by the end-user
as the basis for a free society:
  http://www.rpi.edu/~winner/
  http://www.anu.edu.au/mail-archives/link/link9810/0203.html
Free and open systems like Squeak can play an important role in helping
keep computer systems accountable to human needs through transparent and
user-adjustable operation to minimize the worst aspects of these
exponential technological arms-race trends. So, really, the future of
answering Bob Jarvis's question "Can we do similar things to improve
software?" hinges on what one means by "improve". In a Langdon Winner
sense, Squeak improves software by making it more open even if it makes
it run slower.

-Paul Fernhout
Kurtz-Fernhout Software 
=========================================================
Developers of custom software and educational simulations
Creators of the Garden with Insight(TM) garden simulator
http://www.kurtz-fernhout.com





More information about the Squeak-dev mailing list