D-Wave's 2,000-Qubit Quantum Annealing Computer Now 1,000x Faster Than Previous Generation

[Lucian Armasu]

D-Wave announced its next-generation quantum annealing computer that's powered by 2,000-qubits and is 1,000 faster than the previous 1,000-qubit generation.

D-Wave's 2,000-Qubit Quantum Annealing Computer Now 1,000x Faster Than Previous Generation : Read more

 

[targetdrone]

If D-Wave's 2,000-qubit computer is now 1,000 faster than the previous 1,000-qubit generation (D-Wave 2X), that would mean that, for the things Google tested last year, it should now be 100 billion times faster than a single-core CPU.


and yet it still can't run Crisis.

 

[texastim65]

I'm curious as to whether breaking encryption is something that the qubit computer is designed for or can do.

Being billions of times faster than conventional computers could mean breaking encryption becomes more practical for the NSA or anyone else able to afford it.

 

[leoscott]

Odds are NSA has something that will break encryption already. They generally don't talk about their good toys.

 

[Guest]

For encryption, quantum computing doesn't really help. Time for a 512-bit key is still infinite. A quantum computer saves at most a squared amount of tries. So a 256 bit key would take an average of 2^127 tries, 2^128 tries to be sure, rather than 2^256 bit tries on regular computing.

That's still about a billion billion years, or considerably longer than the age of the universe.

People that claim you can "break encryption" just don't understand the math. You can break it by implementation issues, or other means, but not mathematically.

 

[Adr2t]

Well one factor is that key it self wouldn't be random though. Most people don't pick random passwords but passwords that have some English words in them - in theory - that cuts a good chunk of the guess work out.

 

[Guest]

[quotemsg=18660930,0,2312744]Well one factor is that key it self wouldn't be random though. Most people don't pick random passwords but passwords that have some English words in them - in theory - that cuts a good chunk of the guess work out.[/quotemsg]

But you can't predict where it cuts the guess work. That's the beauty of modern encryption.

Unless you try directly the passwords. And that's all what modern hacking is, rotate the usernames using a fixed password. Not using various passwords for a same user. Someone in the group probably uses "Elv1s4ever", right?

 

[bit_user]

I noted the improvements in the description of these machines and what they're good at, Lucian. However:

their performance increases much more than just 2x, unlike with regular microprocessors. This is because qubits can hold a value of 0, 1, or a superposition of the two, making quantum systems able to deal with much more complex

No, it comes from the fact that each new qubit works in conjunction with all the others. So, the performance improvement should be exponential. Assuming they can still readily achieve and maintain entanglement.

One thing I find so exciting about quantum computers is the kinds of optimization problems we'll be able to solve in areas like system and even mechanical design.

 

[Guest]

I love how I'm being downvoted, by people that don't understand math. You're either O(n^(1/2)), or O(log(N)), which, in binary, is the same. More qubits don't help. Read a book.

 

[bit_user]

[quotemsg=18661375,0,723938]I love how I'm being downvoted, by people that don't understand math.[/quotemsg]I didn't down-vote you, because I'm no expert on this subject.

But, since you dinged me, I'd like you to explain why a quantum computer of at least 512 qubits wouldn't be able to simultaneously try all keys. That's my understanding of how they work.

Do you think Google is flat-out wrong, in their claims?

 

[Guest]

[quotemsg=18661395,0,328798][quotemsg=18661375,0,723938]I love how I'm being downvoted, by people that don't understand math.[/quotemsg]I didn't down-vote you, because I'm no expert on this subject.

But, since you dinged me, I'd like you to explain why a quantum computer of at least 512 qubits wouldn't be able to simultaneously try all keys. That's my understanding of how they work.

Do you think Google is flat-out wrong, in their claims?[/quotemsg]

What Google claim? Search "Shor's algorithm", or "Post-quantum cryptography". Basically, some calculations are extremely faster. Some, including prime numbers, not so much.

IMO, if it's number theory related, don't expect "exponentially faster" solutions. I wish they existed, but, it's so basic, it doesn't yield.

 

[bit_user]

[quotemsg=18661430,0,723938]What Google claim? Search "Shor's algorithm", or "Post-quantum cryptography". Basically, some calculations are extremely faster. Some, including prime numbers, not so much.

IMO, if it's number theory related, don't expect "exponentially faster" solutions. I wish they existed, but, it's so basic, it doesn't yield.[/quotemsg]Google's claim that the previous generation was 10,000x faster than a modern, single CPU core.

And your point about speed-up is that it's only exponential for certain classes of cost functions, correct?

 

[hoofhearted]

Wonder if Peter Shor plays Crysis?

 

[hoofhearted]

Shor and Grover would be the obvious goals. Anyone happen to know the order of magnitude of these algorithms on a quantum annealer vs classical vs true quantum computer?

 

[Daekar3]

At billions of times faster, it seems like it would be worth doing the work to figure out a way to quantum-ify some high-value workloads. Going to be a very interesting next few years.

 

[SockPuppet]

The thing is, there are only a small handful of humans on this entire Earth that actually understand even the basics of quantum mechanics. I guarantee that none of them post on Tomshardware.com

 

[Saga Lout]

[color=0005ff]I will have you know, young Sir, that there are many people on this Forum who think they know everything about anything you care to name.

Stick around; you'll meet 'em! [/color]

 

[bit_user]

[quotemsg=18663379,0,74126]The thing is, there are only a small handful of humans on this entire Earth that actually understand even the basics of quantum mechanics. I guarantee that none of them post on Tomshardware.com[/quotemsg]Well, do you want to talk about science or technology? If you want to talk science, then I suggest you find a good forum on particle physics.

But, if you want to talk technology, then this is a pretty good place. On that point, I'm sure that effective application of D-Wave's computers doesn't require a doctorate in theoretical physics. Otherwise, they really haven't done their job.

On the flip side, I've seen some pretty awful code written by some very smart physicists. Understanding how a thing works doesn't necessarily mean you're good at using it. It's all about specialization, you see. That's what makes it all possible.

 

[Guest]

[quotemsg=18661584,0,328798][quotemsg=18661430,0,723938]What Google claim? Search "Shor's algorithm", or "Post-quantum cryptography". Basically, some calculations are extremely faster. Some, including prime numbers, not so much.

IMO, if it's number theory related, don't expect "exponentially faster" solutions. I wish they existed, but, it's so basic, it doesn't yield.[/quotemsg]Google's claim that the previous generation was 10,000x faster than a modern, single CPU core.

And your point about speed-up is that it's only exponential for certain classes of cost functions, correct?[/quotemsg]

I don't know what a cost function is, but basically, there are different ways to encrypt things. Some ways are vulnerable to quantum computers, but others (modern), aren't.

For modern encryption, we are talking about squared reduction. But encryption is about "disappearing", if you will.

Many other types of calculations are reduced by orders of magnitude. Like going from exponential to polynomial time. So it's not that clear. Their figures are total guesses, IMO.

Quantum computing is awesome, powerful, and has potential. But it's not "THE HAMMER OF GOD". Some things will remain hard to do. And modern encryption is hard.

 

[bit_user]

[quotemsg=18666141,0,723938]I don't know what a cost function is, but basically, there are different ways to encrypt things. Some ways are vulnerable to quantum computers, but others (modern), aren't.[/quotemsg]In simulated annealing, you define a cost function that you're trying to minimize. It's essentially a multi-dimensional search problem to find the set of parameters that represent the lowest cost or lowest energy state.

I don't know how D-Wave's computers represent the cost function, but there must be some way of influencing the qubits, in order to find the optimal solution to a specific problem. However they do it, I'm sure there must be constraints on its complexity. So, you probably have to re-factor any problems with more sophisticated cost functions and iterate. I hadn't really thought too much about that aspect.

Update: here's the answer... http://www.dwavesys.com/tutorials/background-reading-series/introduction-d-wave-quantum-hardware The cost function is represented as blue dots!
; )

I'll probably do a deep dive into programming these things, if/when it seems likely that I'd have access to one. I'm sure it'll soon be possible to rent time on them through the likes of AWS, Google, or Azure. I'm also sure it won't be cheap, as they need to cool 10 kg of material to an operating temperature of 0.015 degrees Kelvin, which they say is 1/180th the temperature of interstellar space. So, on top of the considerable purchase price, the operating costs must be substantial.

 

[Jan Kok]

You can learn how to use a quantum quantum computer, run simulations, and run programs on a real, 5-qubit quantum computer here: http://www.research.ibm.com/quantum/

 

[bit_user]

[quotemsg=18668216,0,2337974]You can learn how to use a quantum quantum computer, run simulations, and run programs on a real, 5-qubit quantum computer here: http://www.research.ibm.com/quantum/[/quotemsg]That's cool, but a 5-qubit machine seems like it'd be slower than using a simulator, on your PC.

Think about it like this: with only 5 cubits, there are only 32 possible answers for any iteration. On a conventional PC, you could naively try every single one. So, it'd be vastly cheaper for IBM to do this. They could even run the simulation right in your web browser.

Putting things in perspective, this new 2000-qubit machine can search a solution space of more than 10^602 possible answers.