If Moore Law on Computing holds, how fast will a computer run in 2248?

Hi there spacebloom! Great question!

Moore's law stems from an observation by Gordon Moore (who co-founded
Intel) that transistor density would double every year. He then
amended this to every 18 months (See here for more:
http://www.webopedia.com/TERM/M/Moores_Law.html )

Most people use the law to suggest that computer *performance* will
double every 18 months. So working on that basis:

The fastest single "computer" is probably NEC's Earth Simulator, no
slouch with a peak performance of 40 teraflops:
http://www.nec.co.jp/press/en/0203/0801.html

Assuming Moore's revised 18 month law, it will have the opportunity to
double in speed 163 times before 2248 rolls around, so if all goes
well, it will be breezing along at a mere
467680523945888933825179146469210566289898413752320 teraflops.

If we consider a boring old Pentium 4, 3ghz desktop computer, then
2248 would see it running a little faster, at
35076039295941670036888435985190792471742381031424ghz.

Unfortunately, Moore's law will probably only hold for a little while
longer, so we're unlikely to see these rather fast speeds - Intel
suggests that it can keep going until the end of this decade:
http://www.intel.com/research/silicon/mooreslaw.htm

I hope this answers your question! If anything's unclear, then please
don't hesitate to request clarification before rating this answer.

All the best,

--seizer

Search strategy: 

Moore's Law
fastest computer

---

Request for Answer Clarification by spacebloom-ga on 07 May 2003 20:28 PDT

Hi seizer,
I am a bit confused as to why you use teraflops in the first example
and gigahertz in the second. Are those interchangeable? (ie 1gHz = 1
teraflop) Otherwise great answer.
spacebloom

---

Clarification of Answer by seizer-ga on 08 May 2003 03:34 PDT

I'm sorry to confuse! 

Teraflops is a measurement which indicates trillions of floating point
operations per second. It's a solid measurement, and you can say
something meaningful about a computer's speed, using that measurement.

Gigahertz merely indicates how many times per second the computer's
internal clock ticks - in the Intel P4's case, 3 billion times per
second. Despite this, the P4 can't reach 3 billion floating point
operations per second (which would be 3 gigaflops, by the way) because
it takes more than one clock tick to do a floating point operation. We
don't really know what the P4's flops rating is, because Intel prefer
not to use it (see this Usenet posting for a technical explanation
why: http://groups.google.co.uk/groups?selm=37CED591.CA341215%40mailbox.intel.com
)

Nevertheless, I thought I'd quickly indicate what Moore's law would
take the gigahertz rating to, because it's a measurement that more
people are familiar with!

Hope this helps!

Regards,

--seizer

I do not think that the moore's law will hold good for the
computer that you refer. I would rather say the speed of
the computer at the time you mention would be few seconds(or
several hundred milliseconds).  But at that time the arcitecture
will not be dependent on the conventional digital logic(or)
electron propogation.  The intelligence will be far superior
equalling human brain, and the memory capacity almost unlimited.
There will not be any input devices like we have now, it will
be just a natural conversation!.  If you are interested in
more details let me know:)

Regards
ldavinci-ga

Idavinci-ga I am interested in what you have to say on this subject. I
am particularly curious about your statement that the speed will be
measured in seconds or milliseconds. Can you elaborate? What will the
formula for comparing computers be? Assuming they will be produced or
at least designed by different entities and people will want to
compare the different models.
Looking forward to your answer.
Spacebloom

Hi spacebloom-ga,

Actually what I meant as seconds or milliseconds will be the
information
processing cycle time and will not compare with the raw clock cycles
(which now-a-days are related to speed of instruction processing in a
cpu).
The future computer will evolve more closer to the architecture of
human brain(and its thinking capability).
Basically, the future computer WILL NOT evolve out of a planar silicon
technology(whose function if at all exists will be limited to
sensors/interfaces to older digital computers).
Now coming to the formulas for comparison for relative performance
estimation.
How do you judge yourself against someone else with almost same level
of knowledge and learned skills in accomplishing a particular task?
Even if a common architecture exists for the future computers
it WILL NOT be possible to use a low level entity such as clock
frequency,
the main reason being those computers essentially will utilize fine
grained distributed processing, where the connection efficiency and
the topology
of the overall circuit(which would be evolved than a created one) will
dominate over the individual processing speeds of the atomic computing
elements. Also the method of evolution of individual processing units
(similar to self wiring that happens in the human brain) will
determine
the quality of the evolved circuits. This should give an idea of the
factors
that will be involved in the formula to compare relative performance.

Let me know, if you are interested in any other specific details.

Regards
ldavinci-ga