First Von Neumann Architecture Quantum Computer 90
holy_calamity writes "The first computers with a von Neumann architecture, where a processor has access to RAM, appeared in the 1940s. Now the first quantum computing system with a von Neumann design has been made, at University of California Santa Barbara. Their quantum processor made up of two superconducting quantum bits can use a 2-bit "quantum RAM" to save entangled bit values into."
Overstuffed box (Score:0)
Re:Overstuffed box (Score:0)
Re:Overstuffed box (Score:5, Funny)
Re:Overstuffed box (Score:2)
*golf clap*
so... (Score:1)
Re:so... (Score:2)
I don't know, but someone will try to run Doom on it.
Re:so... (Score:2)
Re:so... (Score:2)
You would have every possible Quake, Doom level as well as all the cheat codes.
Comment removed (Score:2)
Re:so... (Score:2)
This stuff generally wouldn't find a use at home. Least not for a long time. It would be in the hands of researchers... you know, actually doing stuff.
Re:so... (Score:2)
Oh ya, I spit up my coffee when I read your sig, now I have a mess to clean up...
Re:so... (Score:2)
It still won't be able to run Crysis.
Re:so... (Score:1)
If it is like the PCs I work on most of it would be twiddling its thumbs waiting for something to do or be a "just in case' cache of pretty much everything ever run on the machine? I generally like to slam the crap out of my PC but since going to 8Gb I've found most of it is just cache because i simply don't have enough data being worked on at any one time to need it. Right now checking ResMon I'm looking at 6300Mb being used for cache and only 1800Mb in actual use.
Mine Says: In Use: 6981MB, Cached: 8149MB, Free: 1220MB... And I don't do CAD, 3D-Modelling, Video-Editing or something like that.
As for TFA while i'm sure it'll be useful for someone like the military to simulate nukes going off for most folks it would be a big old "meh" as they simply don't have enough work for even their duals and triples, much less something thousands of times faster. There just hasn't been a "killer app" that would really slam a CPU in so long it isn't even funny, that is why Nvidia and AMD are pushing 3D and Eyefinity because even games just aren't slamming the shit out of chips like the old days.
Not true either. NVIDIA pushes 3D because they can and because 3D currently is the hottest new shit there is. AMD on the other hand pushes Eyefinity not only for enthusiast gamers but for bussiness, too. If you've ever worked with SAP, Datev or other bussiness software that needs huge amounts of screen real estate you might imagine how people could possibly appreciate having 3 or more big screens hooked up to the same box. Also NVIDIA has Tesla. And it's getting used. I just ordered a server with 2 M2090 (6GB) cards in it for testing(!) because our GPGPU guy hits walls with his 1,5GB top-of-the-line Fermi GPU. Also there are quite some GPU clusters out there nowadays. And it's not true that they're only used for military purposes. A lot of them are used for things like climate modelling, genetic research, physics simulations and what not.
Everyone keeps saying "Oh pads are killing PCs!" because the PC sales are down but frankly its bullshit, it is simply folks don't know what to do with what they got power wise so adding MORE power is simply a waste of money for zero gain. While speed for speed's sake like in TFA is cool and all unless we have another killer app that royally slams machines most folks simply don't need it. Hell even my mom that only checks recipes and uses email has a 3.2Ghz P4 with HT and a laptop, simply because I didn't know what else to do with the things, what more do we need?
This expresses your ignorance not only redundantly but also explicitly in addition to the former two paragraphs. You know... just because you and your mom don't know what to do with a computer other than checking emails, surfing Youtube, printing recipies and hack up invitations in Word doesn't mean that increasing processing power goes to waste. You don't even realise how the increase in processing power, memory and storage capacity over the last 10 years changed how we interact with information. Most people just take those things for granted, when they can't actually see the technology that makes this possible. That's sad. But what really bothers me is that there are people like you who even think there's no use for more because they wouldn't have a use for it... how egocentric
Comment removed (Score:1)
Re:so... (Score:0)
Well if you are using entangled kittens you would certainly have PETA bite.
Re:so... (Score:2)
I suppose, but I don't think a garbage scow needs that much power~
Re:so... (Score:2)
Is this coming from an official "Garbage Scow Captain"?
Speaking of which I can't believe the modern "Atari" trashed my favorite childhood game recently with that horribly s***ty remake.
Just a thought... (Score:0)
A few months back I was thinking if we ever moved on from von Neumann or we kept using it. I can see were are we now, but I was thinking that after all those years there would be some improvements or a completely changed architecture.
Von Neumann arch = Executable code in ram (Score:4, Interesting)
To be considered a Von Neumann architecture, the program code needs to be stored in the same ram as the working data. That's the whole point of it. Otherwise, it's a Harvard architecture.
Given that the total ram capacity seems to be 2 bits (though, in all fairness, the bits are qu), it seems implausible that a useful program could fit in it.
Though I have not actually read TFA, I'd say to be skeptical about this. What they probably meant is "It has RAM." Unfortunately they used the completely wrong term for this.
Re:Von Neumann arch = Executable code in ram (Score:2)
I disagree completely with your use of moderately (Score:2)
Please clarify or your entire statement must be dubbed invalid and utterly unreliable. Work and personal references as well as research references would be greatly appreciated.
P.S. - please don't forget to include bank account numbers and personal identification so we can properly audit your accounts for inconsistencies that could theoretically be used to prove you can't manage your own business and therefore should NOT be allowed to interfere with others.
Re:Von Neumann arch = Executable code in ram (Score:2)
"Given that the total ram capacity seems to be 2 bits (though, in all fairness, the bits are qu), it seems implausible that a useful program could fit in it."
How do you know? It's not like you went trough all possible programs and checked which are useful and which are not...
Re:Von Neumann arch = Executable code in ram (Score:2)
00 - useful? No.
01 - useful? No.
10 - useful? No.
11 - useful? No.
Stick another possible value in there and it's still not useful.
Re:Von Neumann arch = Executable code in ram (Score:0)
X0
X1
0X
1X
XX
where X = maybe 0 maybe 1
Re:Von Neumann arch = Executable code in ram (Score:2)
In other words, exactly what I said when I said "Stick another possible value in there and it's still not useful."
Re:Von Neumann arch = Executable code in ram (Score:0)
Qubits can hold 0, 1, or both at the same time; 2qubits can store 3^2 combintions, not 2^2.
Re:Von Neumann arch = Executable code in ram (Score:2)
and once again, an AC didn't bother reading my whole fucking post before replying.
Re:Von Neumann arch = Executable code in ram (Score:2)
You misunderstand qubits. A qubit can hold 0, 1, or a linear combination of 0 and 1, and with complex parameters (as in, real and imaginary). These are not discrete like digital information, they are continuous. While two bits can hold two dimensions of discrete information (amounting to 2^2 = 4 combinations), two qubits can hold four dimensions (2 times 2, because the parameters are complex) of continuous information (amounting to an infinity of combinations).
Discretization of qubits only happens when they are measured.
2 bits? (Score:1)
Re:2 bits? (Score:4, Funny)
May be a lot? Try it yourself.. 2 times 2.
Re:2 bits? (Score:0)
Overflow
Re:2 bits? (Score:0)
And then you wrap around, back to '0'.
Re:2 bits? (Score:2)
Not "nibble" it's nybble cause it's byte not bite.
Re:2 bits? (Score:1)
Just like it's byts, not bits? The spelling of nibble/nybble never really got nailed down like byte did. Both spellings are common. Read on. [wikipedia.org]
Re:2 bits? (Score:2)
I'm going to nybble on some Doritos and write some code in M68K assembler!
Re:2 bits? (Score:3)
... which may be a lot but still it seems pretty useless at this point.
There was a time when people struggled to put 10 transistors on a chip. Valve-based computers of that era ran rings around the puny and useless transistorized systems. You have to start somewhere ...
Re:2 bits? (Score:2)
Re:2 bits? (Score:0)
Valve-based computers of that era ran rings around the puny and useless transistorized systems.
Actually, no they didn't. E.g. the Manchester Atlas and the Cambridge Titan, which used transistors, were far more powerful than the valve EDSAC (the world's first von Neumann computer) - indeed they were among the most powerful computers of their day.
Atlas and Titan used germanium transistors (and Schottky diodes!) - the planar silicon silicon transistor, which led to integrated circuits and then the first microprocessor in 1971 - was still in the future. Which in a way does make your point, even though you're wrong on the details.
A few things to note (Score:4, Informative)
The term Von Neumann architecture has a variety of different meanings. One common meaning of the term is one in which instructions and data retrieval share a common bus. The original meaning was a bit more specific referring to a system that had a CPU, a separate memory for data and instructions, and input/output capability. Originally the real step forward was storing data and instructions together and treating them in some sense the same way which in many ways allowed a lot more flexibility in programming. Treating data and instructions the same way is something that still creates issues; SQL injection attacks are essentially just this: adding data that is formatted to look like instructions. But the upshot is that this use of the term- to use Von Neumann architecture to mean just having a working memory is a less common use of the term.
Moving on from there, the system in question uses superconductors to control qubits. This is one of a variety of different systems being proposed. For example, the most recent quantum computing article on Slashdot ahref=http://hardware.slashdot.org/story/11/08/31/1844252/Record-Low-Error-Rate-For-Qubit-Processorrel=url2html-5998 [slashdot.org]http://hardware.slashdot.org/story/11/08/31/1844252/Record-Low-Error-Rate-For-Qubit-Processor> used ion traps. It is important to realize that different systems cannot be used together in any meaningful way. This means that improvements on any one type don't really carry over to the others. This is important if one is thinking in terms of when all this research will come together. A really good example of this is how early quantum computing used NMR systems http://en.wikipedia.org/wiki/Nuclear_magnetic_resonance_quantum_computer [wikipedia.org] which was then abandoned due to scaling and other issues. A lot of what was learned with NMR systems could not be applied to later quantum computers.
Re:A few things to note (Score:2)
Correct me if I'm wrong, but Von Neumann architecture quantum computer is not very usable. There is not any way to program it in a usable way. Nor there is way to get results out.
Re:A few things to note (Score:2)
99,9% (did I miss any 9?) of the programmers out threre don't seem to have many problems doing it.
Re:A few things to note (Score:1)
The processor architecture has no relevance at the level of SQL. Even if you were running a Harvard architecture with all program memory in read-only mode, SQL injection would work just like it does right now: you introduce unexpected characters in a text string and the SQL interpreter ends up doing something that the programmer did not want.
However, buffer overflows are an intrusion technique that actually does sometimes depend on Von Neumann architecture, but even they can be performed by modifying only data. In particular the return address in stack is considered data, and rewriting it allows you to force a jump to any piece of code you can find in executable memory, which is all a clever intruder needs.
How can you store a qubit? (Score:2)
If you can't copy a quantum state ...
Re:How can you store a qubit? (Score:2)
Re:How can you store a qubit? (Score:2)
But you can copy it's results once they have been rendered readable.
Re:How can you store a qubit? (Score:3)
You can copy it, but it copies both a one and a zero. You don't know which it is until you look.
Re:How can you store a qubit? (Score:2)
Please do not make up quantum mechanics and pass it off as accurate.
Re:How can you store a qubit? (Score:2)
Would 50% chance of being accurate be acceptable?
Re:How can you store a qubit? (Score:2)
Only if you're in the state 1/sqrt(2) * ( |accurate> + |inaccurate> )
Re:How can you store a qubit? (Score:0)
You can copy it, but it copies both a one and a zero. You don't know which it is until you look.
Absolutely not. The No-Cloning Theorem states very specifically that what you are saying is impossible. CLASSICAL information (traditional bits) can be copied; arbitrary quantum information (qubits) cannot. There are clever work-arounds to this problem... if you're really interested, I suggest studying the Shor error-correcting code. Instead of repeating a message to prevent against errors (again, impossible), the information is encoded in an entangled state of several qubits.
As for the original question of storing a qubit, Nature does it all the time. The polarization of a single particle of light is a legitimate single-qubit storage device, and is used in most quantum information implementations.
Re:How can you store a qubit? (Score:2)
Quantum teleportation is a much more boring operating in quantum computing: it's the exchange operator, where the state of qbit A and the state of qbit B are swapped. This is a valid and boring procedure.
Further, you can make it destructive, since you can set either of the two qbits to the zero state. To set qbit A to zero, you measure qbit A, which puts it in either the state 0 or the state 1 (with probability depending on its original state). You then apply the controlled-not operator so that it's bit-flipped if and only if it is in the 1 state -- ensuring that it ends in state zero. Thus, you can move qbit A's state to qbit B and then destroy the value in qbit A.
You technically *can* copy the state of a qbit, but it has a nasty side effect. This is, essentially, the "no-cloning" theorem. If you copy the state of qbit A to qbit B, you end up with a A-B entangled system. As a result, if you measure qbit A and it collapses to a coherent state, qbit B always collapses to the same coherent state. "Cloning" (what a normal person would think of copying) is producing a system where qbits A and B have the same state but are not entangled. That turns out to be impossible.
Re:How can you store a qubit? (Score:3)
is that because copying is stealing?
Re:How can you store a qubit? (Score:1)
For sure he cannot copy links into Slashdot.
Re:How can you store a qubit? (Score:1)
I can't copy a bank note (well, with enough effort I might be able to, but it would be illegal anyway). Still, I can store one in my wallet.
2-bit is not enough (Score:0)
Just can't wait for someone to figure out that introducing anti-mater into the device would give us one more bit: 0, 1 and -1.
Re:2-bit is not enough (Score:0)
I just can't wait for you to figure out that introducing anti-matter into the device would give a small burst of energy and no bits whatsoever.
Re:2-bit is not enough (Score:1)
Actually, the result would be bits and pieces
Re:2-bit is not enough (Score:1)
As there are an infinite number of superpositions of the 0 and 1 state, a -1 wouldn't actually matter.
Not "has access to RAM", silly. (Score:0)
A von Neumann architecture is characterized by having a single bus and address space for data and code -- not by whether data and/or code is stored in drum, delay-line, core/RAM, EEPROM (including flash), or something else entirely.
This implies several interesting things, such as the ability to write self-modifying code and the ability to exploit buffer overruns to execute arbitrary code.
Compare a typical Harvard-architecture microcontroller, where the code space is usually backed by flash or ROM, and the data space by SRAM, so the "processor has access to RAM" -- but it's not a von Neumann architecture, because they're separate address spaces on separate buses.
Re:Not "has access to RAM", silly. (Score:2)
I thought a Von Neumann machine was supposed to be a self-replicating machine ( able to make copies of itself) ?
Damn - I was hoping for "Grey Goo" (Score:1)
When I first saw the article, I thought they had created a quantum version of THIS:-
http://en.wikipedia.org/wiki/Von_Neumann_probe#Von_Neumann_probes [wikipedia.org]
Oh well, we'll have to destroy the universe in using "classical" replicating machines instead.
Two bits of RAM should be enough for anybody... (Score:0)
Someone had to say it.
What is this?... (Score:1)
Everything I know about quantum computing leads me to believe this is a silly exercise.
- There's no benefit to having memory on the same chip as it's easier and more reliable to frame the problem and process the results with a non-quantum computer.
- Having anything that close to the qubits makes it that much harder to handle decoherence which remains an unsolved problem on large scales.
- "Conventional electrical circuits" aren't going to scale and if your quantum computing model can't scale, it's trash.
Publicity stunt?
Re:What is this?... (Score:2)
I'm not sure I understand your criticism completely but please not that it is not a normal memory they put on the chip but a "quantum memory" that can actually store the state of a qubit.
Re:What is this?... (Score:3)
Re:What is this?... (Score:1)
Quantum computing has been a hot research topic for 3 decades now. Do you really think no one considered "Conventional electrical circuits" before moving on to the elaborate qubit registers that current mainstream models use? Issues included error rate and scalability.
There's nothing new about this model except that they're trying to make it programmable. Quantum computers don't need to be programmable; they're best suited for solving a small set of specific problems.
Re:What is this?... (Score:2)
Re:What is this?... (Score:2)
Quantum computers don't need to be programmable; they're best suited for solving a small set of specific problems.
The same things were said about the first electromechanical and electronic computers due to the extreme costs involved and limited initial interest commercially. I'm sure good old Konrad Zuse and Turing had to do an awful lot of convincing to sell people on the idea of funding such pipedreams. Remember, ENIAC was only programmable in an extremely limited fashion and didn't become truly "programmable" until much later and that program could only be read-only.
Quantum computing now looks like the stone age of early electronic computers and this argument between the need for or merits of various quantum memories is similar to old debates about the merits of drum memory, williams tube and mercury delay line memory systems back in the tube days.
Eventually a handful of companies or government agencies will decide they want a couple big experimental quantum computers and the tech will trickle down to the rest of us over the next 50 years. Whether we want it or not. Personally, I think the government has far more computing power than they ever should be trusted with as it is.
Ha. (Score:2)
Their quantum processor made up of two superconducting quantum bits can use a 2-bit "quantum RAM" to save entangled bit values into.
And you thought your 32-bit system needed upgrading.
Re:Ha. (Score:2)
Two bits ought to be enough for anybody.
Re:Ha. (Score:0)
Laugh all you want, but there's no denying that all possible values of any encryption keys of length two bits or less that are protecting your data are now known.
Zuse gets no respect (Score:1)
-- http://en.wikipedia.org/wiki/Z3_(computer)#Relation_to_other_work [wikipedia.org]
Re:Zuse gets no respect (Score:1)
Proving once again that stuff gets named after the guy with the cool name.
Re:Zuse gets no respect (Score:2)
That's because Konrad Zuse sounds a lot like Keyser Soze.
Re:Zuse gets no respect (Score:0)
"I work for Konrad Zuse."
Nah, it doesn't sound right...
Shave and a haircut (Score:4, Funny)
That's what this computer is good for.
Re:QuShave and a Quhaircut (Score:0)
That's what this computer is good for.
FTFY.
You may laugh at the measly two bits. (Score:0)
You may laugh at the measly two bits, but don't you see that now all they have to do is build a Beowulf cluster of these and all your secrets are history, the singularity will be our benevolent dictator and nano-bots will eat the planet.
which one is named Neil? (Score:2)
That's one small step forwards, backwards, both or neither for electrons and one indeterminate state for quantumkind.
I have needs (Score:1)
Re:I have needs (Score:0)
Are you calling Linux a two bit operating system? Them's fight'n' words.
Re:I have needs (Score:2)
Heh I hope you're joking..... that's as bad as all the monkeys who wouldn't buy the Amiga or Atari ST because Lotus 1-2-3 and DOS wouldn't run without add-on hardware.
Cheap, Low Power Spin Setters (Score:2)
Where has all this quantum computing development brought the writing devices to now? Is there cheap (<$500) tech that can set the quantum spin value of electrons, photons, or of any other particle? Is there any spin-setting device that consumes less than the amount of energy differential between spin states it sets in order to set the state (in addition to the states' energy differential)?
Please stop naming things after that Nazi! (Score:1)
Re:Please stop naming things after that Nazi! (Score:2)
Re:Please stop naming things after that Nazi! (Score:1)
Re:Please stop naming things after that Nazi! (Score:2)
von Neumann != von Braun. Not every "von" is the same.
One billion quantum computers (Score:2)
It's not the one billion entangled bits so much. Its the terabyte of entangled kittens...
von Neumann? (Score:0)
The thing that really makes a processor "von Neumann" is the use of a common memory for data and instructions, PLUS the interpretation of a fetched word as data or an instruction depending on the current state of a sequential machine conditioned by the previous instruction - a system comparable with a Turing machine. It's not actually an optimal architecture for either robustness or speed, so I'm not sure why it's considered such a goal.
Quote (Score:0)
"No one will need more than 2 bits of memory for a personal quantum computer."