Quantum Computing.

[Viper Frag]

You guys remember the supercomputer IBM made for the government for the sole purpose of nuclear missle simulations right? Well is that done through massive parallel processing? How many processors does it consist of? Lets say I wanted to build a supercomputer with thirty-two billion processors hooked together, could this be possibly done? Lets say one inch between motherboards, but then that will make a stack half a mile high wouldn't it? but then if it was reconfigured into a cube would it still be possible through simple parallel processing? Would it still be possible even though the processors are a bit too far apart? I would think that the only possible way to do it is to use the quantum characteristics of individual electrons right? What I understand about the normal processors is that they only make calculations using two electron states, which are designated 0 and 1 right? A couple of years of years ago, I don't know exactly when, Richard Feynman suggested that it might be possible to make an extremely powerful computer using all thirty-two quantum states of an electron. I want your input on this, and thanks for taking time to read this.

 

[pm]

Most massively parallel machines usually involve a processor with it's own memory interlinked with other processors using some common form of networking - like IEEE1394, etc. They do not run programs like a PC, the memory is usually shared across the entire computer (usually, but not always - it depends on the architecture). So, a machine with 2048 processors each with 16MB of RAM, essentially has the same amount of memory as one 16MB machine. This is noticeably different from the SMP multi-processor architecture of the PC.

While there are a lot of problems that are readily solved with massively parallel machines (finite element, matrix analysis, weather prediction, and problems that border on chaos theory like aerodynamics), there are plenty of other problems that are not adaptable to a massively parallel machine (like Unreal Tournament).



<< I would think that the only possible way to do it is to use the quantum characteristics of individual electrons right? >>



We are quite a ways away from this right now. Quantum computing - if you can even call it that - is still in it's infancy. They have barely even demonstrated &quot;proof of concept&quot; let alone prototype machines - deploying a massively parallel quantum computer is still well over a decade away... optimistically.



<< What I understand about the normal processors is that they only make calculations using two electron states, which are designated 0 and 1 right? >>



Conventional (non-quantum) computers are based entirely on binary arithmetic. Everything is based on zeros and ones, on and off, high voltage or low voltage. There are two levels and everything is based off of this. But they are not electron states. The charge holding data in even a 0.18um 6T memory cell involves millions of electrons (or probably more - I've never counted) - not just a single electron. I don't understand how, considering Heisenberg(sp?)'s theory and quantum statistics, even quantum computers are going to store data on one or two electrons.


One of the advantages of quantum computers is said to be the fact that they are quantized into more states than just two. So, rather than binary arithmetic, they might operate on hexadecimal arithmetic. Or perhaps even higher. This alone would result is vastly smaller and faster chips - even leaving aside the fact that these devices are built out of single atom devices. But still, this is a field that has barely even figured out how to define the devices, how to create circuits using them, and they still haven't figured out how to wire them up last time that I checked. They have a LONG way to go.

I don't mean to demean the work that is going on. It's fascinating, exciting and cutting edge work and I'm constantly amazed at what they have achieved. But make no mistake: they have a long way to go to even prove that this approach is practical (IMHO).

 

[Viper Frag]

Thanks pm thats just the answer I was looking for. I totally agree with you that it is fascinating and exciting.

 

[rc5]

If you put 32 billion processor together and every processor could last 1000 years before failure, you would get 100,000 processors failing every single day. Lot's of maintenance work to do.

 

[wviperw]

what? how do you get 100K processors failing daily if none of them fail for 1000 yrs?

 

[Viper Frag]

how do you get 100K processors failing daily if none of them fail for 1000 yrs?
My point exactly. How do you get 100K processors failing daily?

 

[rc5]

It should be read in this way: The average life span of a cpu is in 1000 years. That doesn't mean that every CPU will live for 1000 years and then die at same time. It's based on statictics. Some CPUs die in a few years and some might last several million years(Purely assumption).