Google set to make its own energy efficient ARM Server chip - Page 2

Do you have a question? Post it now! No Registration Necessary

Translate This Thread From English to

Threaded View
Re: Google set to make its own energy efficient ARM Server chip
Quoted text here. Click to load it

Reference please.  
There is no "halfway house". And where you get that 10x to 100x
improvement I do not know.  
Entanglement is a harsh mistress.

Quoted text here. Click to load it

Re: Google set to make its own energy efficient ARM Server chip
On 12/17/2013 11:42 AM, unruh wrote:
Quoted text here. Click to load it

Why not look at a real quantum computer?  They seem to be
readily available now and there are several nearby me at NASA
and Google and another at Lockheed-Martin.  See:

https://en.wikipedia.org/wiki/Quantum_computer

http://www.wired.com/wiredenterprise/2012/02/dwave-quantum-cloud/all/

http://www.wired.com/wiredenterprise/2013/05/google-dwave/

A picture of one complete quantum computer is here -- it's a huge cube:
http://www.wired.com/wiredenterprise/2013/06/d-wave-quantum-computer-usc/


Re: Google set to make its own energy efficient ARM Server chip
Quoted text here. Click to load it

Sorry, those are not quantum computers. I know DWave well.  
Exactly what they are is not yet clear ( not least because DWave as part
of their "sales" puts on a restriction that the purchaser is not allowed
to look into the system to see what is there.)  

Quoted text here. Click to load it

Oooh. I can post pictures too. Mine looks like a beached whale.  


Quoted text here. Click to load it

Re: Google set to make its own energy efficient ARM Server chip
Quoted text here. Click to load it

Sorry, those are not quantum computers. I know DWave well.  
Exactly what they are is not yet clear ( not least because DWave as part
of their "sales" puts on a restriction that the purchaser is not allowed
to look into the system to see what is there.)  

Quoted text here. Click to load it

Oooh. I can post pictures too. Mine looks like a beached whale.  


Quoted text here. Click to load it

Re: Google set to make its own energy efficient ARM Server chip
On 17/12/13 21:23, Thad Floryan wrote:
Quoted text here. Click to load it

I suppose D-Wave is the nearest you get to a "halfway house" - it's an
unusual system that is not a quantum computer, but is sold as a quantum
computer.  It's a fantastic marketing ploy - they can sell this device
for vast sums of money, when it is merely a specialised logic system for
"solving" (I say "solving", since its answers are approximate) some
types of optimisation problems using simulated annealing.  It is not a
classical logic device - but it is not a quantum computer either.

/Real/ quantum computers have been built - and they have managed such
feats as factoring 21 quite quickly.

D-Wave have made and sold several systems, despite being slower than a
desktop machine that can be bought for perhaps a 10,000th of the price.
 (The famous "3600 times faster than a PC" claim has been thoroughly
debunked - and even it were true, D-Waves would still not offer value
for money.)


Personally, I don't think quantum computing will ever offer value for
money, or value for energy.  It is an interesting idea, and I am sure
that from all the time and money spent on it there will come new ideas
and understanding of quantum mechanics - and maybe /useful/
applications.  And I also think that it might inspire alternative
computing architectures using classical logic, that could solve some
problems more efficiently than normal cpu's (like using graphics cards
as computing cores).  But I think quantum computing itself will be about
as successful as neural networks (and how many people remember them?).




Re: Google set to make its own energy efficient ARM Server chip
On Wed, 18 Dec 2013 11:06:03 +0100, David Brown wrote:

Quoted text here. Click to load it

I don't know whether quantum computing will ever live up to the hype.
Neural networks may not have, but they're still an active area of
research (see, for example, the IEEE Transactions on Neural Networks
and Learning Systems) and - I have heard - have largely taken over pattern
recognition/classifier tasks.

Probably not too many embedded systems (these days) have such needs...


Re: Google set to make its own energy efficient ARM Server chip
Frank Miles wrote:

Quoted text here. Click to load it


The problem/answer is that everything around you is implemented with
quantum computers.

To borrow quantum computing elements from everything around you
and then make a computer that does useful things is where
half the problems lie. The other half of the problem
is how to feed everyday problems into a quantum computer
and get an every day compatible answer.

There is nothing certain about quantum reality, and that makes
the task of building quantum computers a very real uncertainty.

My idea was to do away with building quantum computers with quantum
computing elements and replace it with digital logic (like the way we do  
away analog computing elements with digital computing elements) and see if  
something useful can be salvaged.

The answer seems to be yes - 5% extra silicon for digital version of quantum
computers in exchange for 10x to 100x performance improvement.
But software is a lot harder to build.
I would like to say how its done, but you would have patent trolls
on your back in no time. Particularly from USA.  

Re: Google set to make its own energy efficient ARM Server chip
On 18/12/13 20:19, 7 wrote:
Quoted text here. Click to load it

<snip the waffling nonsense>

Quoted text here. Click to load it

This is certainly possible - but it is far from being a new idea.
Almost everything in the "quantum computing" world is done like this -
it is all simulated in classical logic.  Most is simulated on normal
cpus and PC's, but some will use FPGA's for faster specialised logic.
Since the best /real/ quantum computers take months to set up, then run
for a few milliseconds at a time to answer problems like "what are the
prime factors of 21?", research and theory is all done with simulations
(or just pencil and paper).

Quoted text here. Click to load it

It is this statement that no one believes - it is completely fabricated.
 It is so far from any sort of reality that it barely warrants a reply.


What /is/ realistic is that thinking in terms of quantum computing can
give ideas to new algorithms or new ways of solving problems (using
classical logic or normal cpus) - sometimes simulating aspects of a
different system can be used for other purposes.  Examples include using
neural network ideas and simulations for pattern recognition, or using
simulated annealing (as done by the D-Wave machines) to give fast
approximate answers to some optimisation problems.


Quoted text here. Click to load it

Don't worry about patents or trolls.  If you think you've had a good
idea here, you can be confident that a lot of other people have thought
of it already.


Re: Google set to make its own energy efficient ARM Server chip
David Brown wrote:

Quoted text here. Click to load it

?
Wrong page?
We gone past ideas of simulations.

  
Quoted text here. Click to load it

That remains to be seen surely?

Quoted text here. Click to load it

If I see it, I let you know.


Re: Google set to make its own energy efficient ARM Server chip
On 19/12/13 21:45, 7 wrote:
Quoted text here. Click to load it

No, "we" have not gone past simulation.  "real" quantum machines are far
too small and far too expensive to be used for developing theories and
ideas.  They are used to demonstrate that things like Shor's algorithm
can work in reality - even though it is only for tiny numbers.  And
obviously there is plenty of research work done on how quantum computers
can be implemented.  But that's it.

If you are talking about /your/ ideas here, then clearly they have not
got as far as simulation, or even as far as vague plans about how
simulation might be possible.

Quoted text here. Click to load it

No, it is historical fact that considering new types of computing
architecture can lead to new ideas and algorithms on conventional
architectures.  Of course no one can predict that a particular
architecture /will/ lead to new algorithms - we can merely say that it
is likely.

Quoted text here. Click to load it

Perhaps I am being unreasonably cynical, but do you /really/ think /you/
have thought of an idea that will give 100x computing performance for 5%
extra silicon, and no one else has thought of it?

As I said, if you've had a /good/ idea here, others will have got there
first.  So if it is unique to you, perhaps it is not so good after all?

Of course, just because something is impossible to implement does not
mean you can't patent it - IBM has patents on faster than light travel,
just in case someone /does/ manage it.  You could certainly patent your
ideas, and maybe even make some money suing people or by selling it to a
patent troll - reality has never been much of a hinder in the American
patent courts.


Re: Google set to make its own energy efficient ARM Server chip
David Brown wrote:

Quoted text here. Click to load it

Do you mean "you" are still stuck?

Quoted text here. Click to load it

Eh? Seems to be stuck in some old page of an old book again.
Every day new reasearch results are made.
Do you ever review them?

Quoted text here. Click to load it

Fail again. If I see it, I will let you know.


Quoted text here. Click to load it

Whatever.


Re: Google set to make its own energy efficient ARM Server chip
On 18/12/13 17:18, Frank Miles wrote:
Quoted text here. Click to load it

Yes, there /are/ some neural networks in use.  In almost every case,
neural network systems are simulated on classical computers - the
structures and algorithms are used for (as you say) pattern recognition
systems such as optical character recognition software.  There have been
a number of "hardware" neural network systems made, but few have made it
out of laboratories.

In other works, neural networks did not come close to "living up to the
hype", but are still of interest in niche areas and in research.  I
believe the same will happen with quantum computing.


Re: Google set to make its own energy efficient ARM Server chip
Grant Edwards wrote:
Quoted text here. Click to load it

Shhhh! :)

--
Les Cargill


Site Timeline