Practical quantum computers will be like dedicated GPUs, Neural Engines, and Crytpo Engines. It will be hardware dedicated to a particular task which it is exceptionally good at and a peripheral to a CPU to off-load special class problems for faster execution.

The CPU will still be there, and most computing will still use it.

And why not, quickly breaking the passwords of some thousand of bank accounts in few minutes to get the appropriate amounts transferred in an additional few more.

 

Just imagine what will happen if quantum computers are involved with artificial intelligence, and then what happens when some fool is able to make the computer system self-aware. Suddenly there will be computers convinced that they are much wiser than humans and they will have some levels of intelligence that allow them to access areas to take over. Then the worst case situations of science fiction may become reality. Robbing bank accounts will just be the start.
What will happen to fully electrified California when the computers decide that the humans are wasting electricity??

 

Matrix

 

IBM president , 1943.
" I think there is a world market for may be 5 computers ."
DEC founder , 1977 ,
"There is no reason anyone would want a computer in their house"
Founder of 3Com , 1995
"I predict ... By 1996 the internet will collapse " ..
Founder Microsoft , 2004
"I predict that by 2006 spam will be solved"

Were great at predicting. .... !

 

And why not, quickly breaking the passwords of some thousand of bank accounts in few minutes to get the appropriate amounts transferred in an additional few more.

Sorry but all of that is based on BS from the POP-Sci Quantum Hype Matrix.

Practical quantum computers (and classical machines) won't be breaking the passwords of thousands of bank accounts in a few minutes, it fact they likely won't be used for breaking passwords at all because there is really only a narrow class of cryptographic functions that will fall due to quantum computer factoring speedups. Those that will be affected are being updated and the ones that won't be affected will carry on.

https://spectrum.ieee.org/post-quantum-cryptography-nist

The vulnerability in public-key encryption is because of quantum computers’ ultimate ability to have almost exponentially faster factoring of large prime numbers. What types of strategies are being used to create these new public keys? Can you run through a few of them?

Moody: In crypto, we like to use ideas that have been around for a while. Since [Peter] Shor’s algorithm was discovered back in the ’90s, researchers have been looking into this. There are three big families where a lot of the solutions are coming from. The most popular one involves what are called lattices. This is a mathematical structure. You take basis vectors; you take integral linear combinations of them. And you can do some pretty interesting things cryptographically. There are no known quantum algorithms that do better than the classical attacks on cryptosystems based on lattices. Cryptosystems based on lattices seem to be the top contenders in terms of key size and efficiency. So it wasn’t surprising that we received a lot of lattice submissions. The second family is based on what are called error-correcting codes or code-based cryptography. These codes have been used in information security for a long time, because data gets sent on noisy channels. If you use error-correcting codes, you can account for the error and recover the message that was originally meant to be sent. We use ideas similar to what’s used in lattices with these codes. And there, they seem to be just a half-step behind the lattices in terms of key sizes and efficiency, but they’re pretty good as well. So we saw a number of code-based submissions. The third biggest family includes some signature schemes based on what’s called multivariate cryptography. You’ll use a multitude of variables, x1 to xn, and create a system of quadratic equations. And it’s very easy to define and evaluate one of these systems, but very hard to solve. So it turns out that works well for digital-signature schemes.

https://www.nsa.gov/Cybersecurity/Post-Quantum-Cybersecurity-Resources/

Quantum key distribution utilizes the unique properties of quantum mechanical systems to generate and distribute cryptographic keying material using special purpose technology. Quantum cryptography uses the same physics principles and similar technology to communicate over a dedicated communications link.NSA continues to evaluate the usage of cryptography solutions to secure the transmission of data in National Security Systems. NSA does not recommend the usage of quantum key distribution and quantum cryptography for securing the transmission of data in National Security Systems (NSS) unless the limitations below are overcome.

 

My point is really that there is no need fo greater sjngle-package processing power except the reduction of package size, which is all a marketing goal.
Making the package smaller raises the power density and increases the internal temperature, reducing the reliability. It also makes any system more difficult to service, thus increasing the amount of electronic waste materials.
So really, the results are simply not worth the effort.
A much greater gain in performance would come from more efficient code. But creating more efficient code will require levels of both skill and talent presently not available.

 

I don't think that pure performance are the practical limits today stopping us from some 'great, new thing' in computing. its not coding or programming better to get more efficient code. The root cause is not a lack of skill and talent in a massive pool of skills and talents today. We have/or IMO, hitting a fundamental wall in the way we understand the intelligence of living beings. I don't know what's beyond that wall, that seems to be, a limit to expression of what we need in computing to even things like self-driving cars that prevents, what and how, we 'really' want things do act in response to human actions.

 

Here's a hint... Computers are already much faster than the internet... So, if you really wanna speed something up...

 

I just looked it up: TSMC’s most complex processor has 59.4 billion transistors.

Back in the ‘70’s I was blown away when I read that the MOS6502 had 3,500 transistors!

Is that 59 billion transistors on a single die, or is it the total count for all of the die in a multi-chip module?

 

And why not, quickly breaking the passwords of some thousand of bank accounts in few minutes to get the appropriate amounts transferred in an additional few more.

You don't need to worry about that -- well before quantum computing gets to that level of capability, we will move to post-quantum crypto and do so very quickly. We don't have to wait for anything to be discovered or invented, we already know how to do crypto that is as resistant to quantum-based attacks as current crypto is to current attacks.

 

Here's a hint... Computers are already much faster than the internet... So, if you really wanna speed something up...

I have 2GB fiber internet up/down speeds. All it does to download spam emails faster.

 

Is that 59 billion transistors on a single die, or is it the total count for all of the die in a multi-chip module?

I believe that TSMC has the ability to stack chips inside one package.

 

My point is really that there is no need fo greater sjngle-package processing power except the reduction of package size, which is all a marketing goal.
Making the package smaller raises the power density and increases the internal temperature, reducing the reliability. It also makes any system more difficult to service, thus increasing the amount of electronic waste materials.
So really, the results are simply not worth the effort.
A much greater gain in performance would come from more efficient code. But creating more efficient code will require levels of both skill and talent presently not available.

Wonder if AI will be able to make existing code more efficient , now there a thought !

 

As a former IC layout engineer of 15 years, the biggest problem with a larger die is yield. There is more that can go wrong with a larger die and the cost of the real-estate goes up as well because you are limited to how many IC's you can get from a single wafer. Which in turn raises the cost of the IC to the consumer and with larger IC's this cost can be prohibitive.

As far as increased substrate leakage the smaller you go, that is true with most 2D transistors that were confined to Moore's law, but now there are 3D transistors that are built vertically like FinFET and CFET that minimize substrate leakage.

Reference:
https://www.embedded.com/integrating-cfet-into-the-logic-technology-roadmap-beyond-1-nm/

 

Wonder if AI will be able to make existing code more efficient , now there a thought !

We already have these things called optimizing compilers that generate code at the level of a good (not the best) human assembler specialist. Why is it that people think that some AI will be able to do much better until they have actual intelligence that understands problems instead of just understanding code (things like regenerative AI understand nothing) ? I'll say it again for the 1000th time, we don't have a problem with coding, we have an issue with problems (details, specification, operations, interactions, planning, forecasting, etc ...) being translated into human thoughts and then into computer languages as code. Code <-- that part is relatively easy.

 

We already have these things called optimizing compilers that generate code at the level of a good (not the best) human assembler specialist. Why is it that people think that some AI will be able to do much better until they have actual intelligence that understands problems instead of just understanding code (things like regenerative AI understand nothing) ? I'll say it again for the 1000th time, we don't have a problem with coding, we have an issue with problems (details, specification, operations, interactions, planning, forecasting, etc ...) being translated into human thoughts and then into computer languages as code. Code <-- that part is relatively easy.

To be clear .
Tong well and truelly in cheek when said ai ..
But thinking about it more.
Compilers are great ,but code still seems to bloat over time .
Wonder what tool could help fix that and could it be ai ?

 

Once there were components which are very Large and Heavy...(Like Hardisks and other components) Now there are so many are put inside a tiny block. But why only a few processors can be put inside a block now. Why not thousands of it inside? What's stopping us?

Hi there,


If you don't want to read this whole reply, skip to the very last paragraph where there is a summary.

Last i read, some months ago i think, AMD had a CPU with 64 processors on chip, but it may be more by now.

The main point is that to increase the computing power you either have to increase the frequency or increase the number of transistors or both. There are a number of problems that come up when both increasing the frequency and when adding more transistors to a CPU.

The first is heat. Due to non-ideal components, there is wasted heat, and heat must be removed from a volume because if it is not removed the temperature will build and build and just keep increasing over time, and once it gets too hot the CPU will either shut down or get destroyed.
To remove the heat, there has to be a heat conducting path to the outside world where there is ambient air temperature that is lower than the core temperature. This turns the main problem into a heat versus distance problem, and as i always like to say, distance is probably the most kind of important measurement in the universe.
The heat has to travel a distance to get out of the package, and this requires materials that can conduct heat well. Just like an electronic circuit that has conductors that have to conduct current and suffer from resistive losses, the heat has to be conducted through the materials that conduct the heat and these conductors are not ideal either. This means that these non-ideal conductors will not be able to conduct all of the heat generated out of the package fast enough, only some of the heat. This causes heat buildup, which then causes more heat to be conducted out of the package. There comes a point where this process reaches an equilibrium state, where a certain amount of heat is conducted out of the package and the temperature reaches a final level that is lower than the temperature that would cause the core to shut down.
The problem is the heat path has a certain distance to travel, and the path is made up out of a certain type of material with a certain thermal resistance. There comes a point in the physical dimensions where the distance cannot be decreases and the thermal resistance cannot be decreased, and that's where we reach the limit of heat dissipation capability.
Now since the size of the transistor determines the distance the heat has to travel, making the transistors smaller helps to decrease this distance and also decreases the heat each transistor generates. This combination allows the frequency to increase, and the number of cores on a chip to increase. Thus, decreasing the transistor size allows more transistors in the same area and thus allows more cores to reside on the same chip. Since distance is a three dimensional measurement we have to consider the distance along at least three different paths. To the right and left, we dont see much improvement because if we decrease the size of each transistor by 1/2 but increase the number alone a line by 2 times, we end up with the same heat dissipation horizontally in four directions, so not much improvement there. Since the bulk of the vertical distance is not the transistors size itself, we dont get much improvement there either. This means we get two times as many transistors but not much more heat dissipation capability. That means the frequency cannot increase much.

In the end we end up with more transistors when we go to a smaller size, but not much higher frequency.
Since there is a limit to the nature of the lithography involved, we can only go down so small with the transistor size too, and this limits the number of transistors in a given area. This is why most improvements of late have been to go down in transistor size, and surprisingly it took Intel a while longer than AMD to go down to a smaller transistor size.
There is also the vertical stacking of transistors which has become an interesting topic.
The latest i think is the use of a modular approach, with adjoining substrate material. Different functions can be on each module, and the modules are then mounted on a substrate that is both electrically and thermally conductive. I don't know if this is adopted for regular CPUs yet or not.

The above also gives you some idea why some CPUs require water cooling. That's to get heat out of the package faster by cooling the surface area more than with just forced air cooling.

The simpler explanation comes from the way surface area and volume play out in the race to dissipate heat. The larger the volume and lower the surface area, the harder it is to get the heat out of the package. Surface area plays a big role, because that is the only way out for the heat. Volume plays a role because we try to pack as much as we can into one volume, and the more we pack the more heat we generate internally, and it could also mean a longer distance for the heat to travel to the surface. The ideal shape I think would be a sphere.
Everything else is about ingenuity, but there is no way to beat the surface area problem. Even if we use liquid nitrogen, we still reach a limit although it would be much higher than what we see with water cooling.

 

It would be very interesting to read a detailed technical description of just what the mechanism is of "quantum computing" that is going to deliver the touted wonderfulness. And who is going to write that amazing code that will run a hundred orders of magnitude faster.

Hi,

One way to think about this is to think in terms of the frequency domain vs the time domain.

In the time domain we see:
y=A*sin(w*t)

and to generate all the solutions to that we have to run t from 0 to some upper limit.
In the frequency domain, we have:
Y=A

and note there is no time involved in this, and this suggests an infinite number of (time domain) solutions without actually having to run time from 0 to some max.
The only time required then is to both program the frequency and pick out the solution(s). The reward though is an infinite number of solutions become available.

 

Responding to post #38: if there are suddenly a lot of solutions then the new problem is selecting the LEAST INCORRECT solution. And that explanation is either wrong or so generalized as to be useless.
Certainly it does not come close to answering what I was asking about. Consider that not all answers will have adequate value, now there is the challenge of somehow determining which is correct. A bit like using a checksum error to find which bit is incorrect.

 

And why not, quickly breaking the passwords of some thousand of bank accounts in few minutes to get the appropriate amounts transferred in an additional few more.

Quantum-proof encryption is actually already a solved problem, so this isn't really a concern unless the banks act stupid.