Re: Testaus

On Dec 16, 2:59 am, "Michael J. Mahon" <mjma...@xxxxxxx> wrote:
Eric Rucker wrote:
On Dec 15, 3:43 am, Gav <g...@xxxxxxxxxxxx> wrote:
On Dec 15, 1:44 pm, "Michael J. Mahon" <mjma...@xxxxxxx> wrote:

The 17-board AppleCrate II consumes about 75W total!
You could be on to something there - from memory a fully populated HP
c-class blade enclosure (16 blades) can suck up to 6KW of juice (or
there abouts) - the answer to the worlds energy problems is now
obvious - I think we should give Apple a call and get them to crank up
the Apple II assembly line... ;-)

The funny thing is, the Apple II is horribly energy inefficient for
what it is, in today's time. With modern process technology, you could
probably make a *MUCH* more power efficient Apple-II-on-a-chip ASIC,
and run it off of a couple AA batteries for a ridiculously long time.

I mean, my calculator has much, much more processing power than any
Apple II, even a GS with a 14MHz TWGS, thanks to its 75MHz ARM920T-
based core, and it uses four AAA batteries, and I think I've only
changed them once in over a year. And, it seems that thing's processor
is made on a 180 nm process (which was state of the art in 1999,
outdated in 2003 or so,) not current 65 or 45 nm processes.

Of course that's true.  A modern implementation of a 1MHz Apple //e
could easily be powered by a coin cell!

The implication is that there's no real reason why a modern computer
should consume more than milliwatts while waiting for a keystroke or
a mouse move.  Evidently, power reduction has not been a priority in
the PC world--even for laptops, where battery life is a huge factor.
(I suppose getting the electronics under the power consumption of the
always-on backlight is not seen as much of a win, just as the 3W-4W
backlight itself is not seen as a big problem for 25W laptops.)

Since semiconductor advances have, for the last decade, been plagued
by leakage and power problems, it's easy to forget that the power per
transistor and, in particular, power per MIP has drastically declined.

The snag is that the clock rate increases have slowed to a crawl, and
intra-thread parallelism has been almost fully exploited, so all the
performance improvements are now coming from increased cache sizes,
higher memory bandwidth, and--the biggie--multiple cores on one die.

To get the performance benefit of this, it's necessary to decompose
applications into parallel chunks that run efficiently on single cores,
and are not too bogged down by inter-core communication.

For some applications, such a decomposition is trivial, but for many
it is very difficult and poorly understood.  As a result, the principal
performance driver for the future is going to be, not semiconductor
technology, but parallel software technology.

If we're lucky, it will only be a decade or so before we develop a
real science of parallel computation.  If we're about average, I'd
expect it will take two decades.  Meanwhile, we're waiting for the
electrician.  ;-(

The fact that this "parallel wall" was predictable (and predicted) for
decades did not hasten our progress in the study of parallel computing
in the slightest.  As usual, humans never address a world-changing
problem until it becomes an "emergency".

That's what I call "using your nose for a wall detector"--it detects
every wall unerringly, but painfully.  ;-)

In the meantime, it might be fun to see how simple we can make it
to "make computing simple".  For example, how much of the advantage
of modern GUI computing has nothing to do with wasteful eye candy
but simply with providing efficient pointing and selection tools?

Could we make a nice "anybody can use it" computer that runs on less
than a watt, with a consequent battery life of several days?  Could
we simplify it to the point that it required less than a half-gig of
DRAM (maybe then it could be SRAM) and a quarter-gig of code, just
to browse the web, play music, and write a letter or spreadsheet?
(Video is a little tougher, but with dedicated silicon it shouldn't
be too bad.)

It's wonderful to have pervasive standards from the point of view of
easy data interchange, but I do regret that that standard has become
such a kludge tower (and MacOS isn't that much better in terms of
complexity and sheer size).

It surprises me that so few efforts have been made to radically simplify
the hardware and software that we use, since such a large part of the
market uses only a few applications anyway.

The computing needs of at least half of the world's users could be
met by a single "blobbed" chip powered for a week by two D cells!
(Kind of like my HP 100LX palmtop that provides a PC XT that runs for
three weeks on two AA cells and is booted less than once a year!)

-michael

******** Note new website URL ********

NadaNet and AppleCrate II for Apple II parallel computing!
Home page:  http://home.comcast.net/~mjmahon/

"The wastebasket is our most important design
tool--and it's seriously underused."

And you've certainly got an interesting point there. Although, HTML
and CSS rendering takes a fair amount of power, especially with
JavaScript in the equation.

Music... what formats? If you're fine with just MP3, WMA, and Ogg
Vorbis, then there are off-the-shelf chips that are designed to have a
microcontroller feed them MP3s from some filesystem, and are basically
dedicated sound processors that take a very small amount of power.

Writing letters and spreadsheets... depends on the complexity of the
documents, but for 99% of what people need, something like AppleWorks
GS is enough.

So, web is the hard one - music is actually the easiest, surprisingly.

I think, if you coded *EXTREMELY* tightly, you could get a browser
with modern HTML, CSS, and JavaScript support running comfortably in
32 megs of RAM, though. But, I do mean *EXTREMELY* tightly - writing
the whole thing in assembly, etc., etc. Take away the JavaScript, and
you could probably do it in 16 megs. (But, the JavaScript IS rather
important nowadays.)

There is also... cheating. Write a MIDP 2.0 implementation for the
thing, and run Opera Mini on it. Let some servers in Norway handle the
HTML and CSS rendering and JavaScript execution.
.

Relevant Pages

  • Re: Testaus
    ... (I suppose getting the electronics under the power consumption of the ... to "make computing simple". ... NadaNet and AppleCrate II for Apple II parallel computing! ... JavaScript in the equation. ...
    (comp.sys.apple2)
  • Re: Testaus
    ... the Apple II assembly line... ... I mean, my calculator has much, much more processing power than any ... where battery life is a huge factor. ... to "make computing simple". ...
    (comp.sys.apple2)
  • Re: Writing an Operating System in Ada
    ... Note that the file system itself is built on a layer of blocks and sectors. ... If we compared the computing power we had in i386 days with the ... "pragmatism" was, if you have a software problem, buy a new computer. ...
    (comp.lang.ada)
  • Re: Alternative to electricity/electronics technology?
    ... about everything, from power transfer to computing, I was curious - is ... You could harden electronics just fine...although micro-mechanical ... I'm not sure how optical computing ... The water mains thus once doubled as a power ...
    (rec.arts.sf.science)
  • Re: Keeping Computers Secure in the Modern Era
    ... Computing in the modern era is so full of chaos advotees. ... >offensive and start shutting down hacker websites and computers/and or ... the facts as I see them. ...
    (microsoft.public.windowsxp.general)