So who the hell is still supporting BU?

[jbreher]

Moore's Law is no longer tenable, as microchip manufacturing has (ironically) hit it's the technological limits.

aww..... that's just precious.

Indeed... After decades of educated people making the same hurried conclusions that the human race has at various intervals and industries hit its technological limits only to be proven premature not much later in the future, we'll continue to have people we can rely on to make the same (misguided) conclusions.

I always find that precious=)

what you 2 lithographic etching engineers are failing to appreciate is that while Moore's law could be continued with a different paradigm, that paradigm does not exist yet. Until it does, Moore's law as described stopped working last year, it took ~24 months to get 14nm transistors, and will take longer than that to reach 10nm.

5nm transistors have been demonstrated in the lab fifteen years ago. 7nm fab pilot production has been achieved two years ago. Volume production this year. By at least three independent fab companies with different processes. And as pointed out earlier, yesterday's planar processes are giving way to the third dimension - which opens vast new frontiers for scaling. Chips built with 64 layers of cells (far more layers of lithography) are shipping -- in volume -- for regularly-ordered device technologies. Which will lead to a quantum leap in scaling, before settling back down to something approximating Moore's 'law'.

Incidentally, while I am not a lithographic engineer, my day gig has me exchanging research info with lithographic engineers as a regularly occurring interlock activity. You?

[1714039562]

1714039562

[1714039562]

1714039562

[1714039562]

1714039562

[1714039562]

1714039562

[1714039562]

1714039562

[1714039562]

1714039562

[classicsucks]

No; unlimited was created because there shouldnt be a maximum blocksize that is part of the consensus rules.

The idea/argument that larger blocks will centralize hashing power in areas that have faster propagation
times is silly and does not hold water since A) such areas already have that 'advantage'  , B) miners can
always choose to solve smaller blocks, at least for several decades, and C) the bitcoin relay network already
purports to mitigate propagation issues, claiming global latencies as low as 100ms, which compared to an
average block interval of 10 minutes (which is 0.016%, or about a hundredth of a percent) is insignificant.

Unlimited was created with only one purpose - to boost an ego of a single person who's silly idea was not accepted by core. That person is neither a dev nor understands anything about economics, that is why he keeps pushing this most outrageous idea of on-chain scaling to infinity - and now that he has spent significant amount of money on pushing this silly idea, it has become a matter of principle for him.

The fact that you monkeys jump around him does not change the fact that exactly zero people (of the ones that matter in bitcoin world) do support that fork. It will never ever be accepted, just deal with it and move on do something useful. What he might succeed in is blocking real thought-through proposals indefinitely; in that case bitcoin remains immutable and all innovation goes into additional layers. Works this way, too.

Actually, Peter Rizun seems to understand the economics of bitcoin far better than your average core fanboi... His idea is not the only alternative to core's Segwit flop. It's merely coming to the fore because of the vocal majority's opposition to Segwit and their exhaustion with waiting for core to finally raise the blocksize. The excuses and FUD have worn thin after 4 years. The time will come soon that a majority will compile and run code from someone other than core... I really dread that moment, because core is technically the most competent team. However, if they can't raise the blocksize because of Blockstream investors or whatever other covert reason, then we'll have no choice at some point. The latest mempool surge was an ominous indicator that this moment may come sooner than we expect...

As far as Moore's law, processing power is plenty fast right now, and more than able to compute the necessary hashes and signatures for bitcoin transactions. It's not Moore slowing bitcoin down, it's central planners in core's "Politburo".

[jbreher]

when there's a moore's law for bandwidth, that's when it'll become a 'law' that's worth paying attention to.

Nielsen's law. Nielsen's law of Internet bandwidth growth. It's a thing.

https://www.nngroup.com/articles/law-of-bandwidth/

eta: I see Lauda beat me to the post

[Carlton Banks]

Moore's Law is no longer tenable, as microchip manufacturing has (ironically) hit it's the technological limits.

aww..... that's just precious.

Indeed... After decades of educated people making the same hurried conclusions that the human race has at various intervals and industries hit its technological limits only to be proven premature not much later in the future, we'll continue to have people we can rely on to make the same (misguided) conclusions.

I always find that precious=)

what you 2 lithographic etching engineers are failing to appreciate is that while Moore's law could be continued with a different paradigm, that paradigm does not exist yet. Until it does, Moore's law as described stopped working last year, it took ~24 months to get 14nm transistors, and will take longer than that to reach 10nm.

5nm transistors have been demonstrated in the lab fifteen years ago. 7nm fab pilot production has been achieved two years ago. Volume production this year. By at least three independent fab companies with different processes. And as pointed out earlier, yesterday's planar processes are giving way to the third dimension - which opens vast new frontiers for scaling. Chips built with 64 layers of cells (far more layers of lithography) are shipping -- in volume -- for regularly-ordered device technologies. Which will lead to a quantum leap in scaling, before settling back down to something approximating Moore's 'law'.

Incidentally, while I am not a lithographic engineer, my day gig has me exchanging research info with lithographic engineers as a regularly occurring interlock activity. You?

That's all very fascinating. But where's the part that illustrates Moore's Law is still in effect?

Oh yes, that's right, every step since the 14nm process has been taking LONGER than the 18 months of R&D time Gordon Moore predicted, so you've proven precisely nothing

[jbreher]

Moore's Law is no longer tenable, as microchip manufacturing has (ironically) hit it's the technological limits.

aww..... that's just precious.

Indeed... After decades of educated people making the same hurried conclusions that the human race has at various intervals and industries hit its technological limits only to be proven premature not much later in the future, we'll continue to have people we can rely on to make the same (misguided) conclusions.

I always find that precious=)

what you 2 lithographic etching engineers are failing to appreciate is that while Moore's law could be continued with a different paradigm, that paradigm does not exist yet. Until it does, Moore's law as described stopped working last year, it took ~24 months to get 14nm transistors, and will take longer than that to reach 10nm.

5nm transistors have been demonstrated in the lab fifteen years ago. 7nm fab pilot production has been achieved two years ago. Volume production this year. By at least three independent fab companies with different processes. And as pointed out earlier, yesterday's planar processes are giving way to the third dimension - which opens vast new frontiers for scaling. Chips built with 64 layers of cells (far more layers of lithography) are shipping -- in volume -- for regularly-ordered device technologies. Which will lead to a quantum leap in scaling, before settling back down to something approximating Moore's 'law'.

Incidentally, while I am not a lithographic engineer, my day gig has me exchanging research info with lithographic engineers as a regularly occurring interlock activity. You?

That's all very fascinating. But where's the part that illustrates Moore's Law is still in effect?

Oh yes, that's right, every step since the 14nm process has been taking LONGER than the 18 months of R&D time Gordon Moore predicted, so you've proven precisely nothing

Your statement was that "microchip manufacturing has (ironically) hit it's the technological limits". Which, of course, I have demonstrated to be objectively false.

Of course, Moore's 'law' is not what you claim it to be. If anyone would be able to define 'Moore's law', one would expect that logically to be Gordon Moore himself. The quote attributed to him, that serves as his vision of any sort of 'Moore's law' is "The number of transistors and resistors on a chip doubles every 24 months". Which remains pretty well on-track.

But in the context of Bitcoin, such is largely irrelevant. The only question that matters in this space is 'Is semiconductor technology improving at a rate that exceeds the expected rate of increase of Bitcoin usage, would Bitcoin have an unrestricted maxblocksize?' The answer to which would be an unqualified 'Indubitably'.

[Carlton Banks]

Your statement was that "microchip manufacturing has (ironically) hit it's the technological limits". Which, of course, I have demonstrated to be objectively false.

It's very simple: 5nm is far too close to the actual size of individual silicon atoms. Presumably you now would like to argue with the atoms themselves about how they are exhibiting the wrong dimensions

Do you understand what individual means, lol

[franky1]

and while people pretend to think they know bitcoin by talking about moores law.

they need to realise that hashing a block solution(mining) and block size are not linked.
no matter how big a block becomes.. its hash will always be a sha256. and a mining pool will hash the hash the same.

dont confuse the cost of hashing a single sha256 piece of data.. with the blocksize debate.
mining and block size are 2 different things.

also bitcoins blocksize does not require 100% utility of the top end CPU in existance to validate transactions/blocks.. so there is still lots of capability in block propagation/validation.(no where near the end of the limit of moores law)

yes a raspberry Pi 3 can handle 8mb blocks, even core devs agree but devs think 4mb is safer. even at consensus 2015 the compromise went down to 2mb just so that core devs and their sheep cant play that doomsday card.

but it seems the blockstream sheep are trying to pull that doomsday back to life without actually thinking about the reality.

a raspberry Pi 3 is far far less than a standard desktop PC. so block validation is not at its limit so block size should not be limited due it such fake doomsdays.

[European Central Bank]

Nielsen's law. Nielsen's law of Internet bandwidth growth. It's a thing.

https://www.nngroup.com/articles/law-of-bandwidth/

eta: I see Lauda beat me to the post

it's not a thing where i live. my bandwidth hasn't changed for more than a decade. moore's law can be executed by anyone buying a new machine and plugging it in. nielsen's depends on some nice guys showing up and digging holes in your road.

[jbreher]

Your statement was that "microchip manufacturing has (ironically) hit it's the technological limits". Which, of course, I have demonstrated to be objectively false.

It's very simple: 5nm is far too close to the actual size of individual silicon atoms. Presumably you now would like to argue with the atoms themselves about how they are exhibiting the wrong dimensions

Do you understand what individual means, lol

"5nm is far too close" ... for ... what exactly? What is it about 5nm that makes it "too close"? What occurs, or stops occurring at 5nm?

Of course I understand what individual means. A 5nm cube of silicon contains on the order of 12,500 silicon atoms. Still quite far from 'individual'.

But this is irrelevant. Again. Despite being irrelevant to Bitcoin, it neither brings an end to semiconductor scaling. As you keep missing, we are just starting to exploit the third dimension in semiconductor fabrication.

Sure, there will be problems at each process node. That will continue to be solved just like we've solved each process barrier for the last half-century. The only measure that matters -- the amount of computation one can do for a certain amount of money -- will continue to climb unabated.

[Meuh6879]

the answer is simple : don't product electrical processor.


product laser processor.

[Carlton Banks]

But this is irrelevant. Again. Despite being irrelevant to Bitcoin, it neither brings an end to semiconductor scaling. As you keep missing, we are just starting to exploit the third dimension in semiconductor fabrication.

hmmmmm, jbreher thinks silicon chips can be fabricated at 5nm and less, knows that 14nm took 24+ months, yet also knows that 3D TSV chips is suddenly being looked at as the actual vector with which to continue Moore's law. Why aren't the engineers considering going into femtometer processes? (like you seem to think is possible lol)

You're right this is irrelevant, but it was your point to begin with, jbreher. So now we know your original post was irrelevant, spare yourself the embarrassment https://bitcointalk.org/index.php?topic=1771911.msg17957121#msg17957121

[jbreher]

But this is irrelevant. Again. Despite being irrelevant to Bitcoin, it neither brings an end to semiconductor scaling. As you keep missing, we are just starting to exploit the third dimension in semiconductor fabrication.

hmmmmm, jbreher thinks silicon chips can be fabricated at 5nm and less, knows that 14nm took 24+ months, yet also knows that 3D TSV chips is suddenly being looked at as the actual vector with which to continue Moore's law. Why aren't the engineers considering going into femtometer processes? (like you seem to think is possible lol)

You're right this is irrelevant, but it was your point to begin with, jbreher. So now we know your original post was irrelevant, spare yourself the embarrassment https://bitcointalk.org/index.php?topic=1771911.msg17957121#msg17957121

to wit:

You do realize Bitcoin is working exactly as intended (ie allocating scarce resources efficiently)

Well, no. To quote satoshi:

The existing Visa credit card network processes about 15 million Internet purchases per day worldwide.  Bitcoin can already scale much larger than that with existing hardware for a fraction of the cost.  It never really hits a scale ceiling.  If you're interested, I can go over the ways it would cope with extreme size.

By Moore's Law, we can expect hardware speed to be 10 times faster in 5 years and 100 times faster in 10.  Even if Bitcoin grows at crazy adoption rates, I think computer speeds will stay ahead of the number of transactions.

[iCEBREAKER]

"5nm is far too close" ... for ... what exactly? What is it about 5nm that makes it "too close"? What occurs, or stops occurring at 5nm?

Is this a trick question?  Casual checking in...  

At some point down there you get too much electrical leakage to carry a signal.

Classical mechanics break down as quantum effects begin to appear.

Registers begin to act too much like qubits for digital reliability, which requires static discrete states (1 or 0).

The easiest gains were at the beginning.  Moore was observing the harvest of the low hanging fruit, nothing more.

Now it takes more capital to build a modern fab than it took to go to the moon.

[IadixDev]

Just to post my two cents in the trolling  

Already as it as already been said, moore law has little to do with bitcoin transaction ceiling in itself.

Even if cpu power goes x100, there will still be the same blocks every same ten minutes with same number of transactions.

Increasing block size come with certain number of problems, like the amount of processing in a block, if you have 8Mo blocks contain a single transaction with only seg ops, and then another block of 8Mo containing a single transaction with only inputs to such transactions, it will clearly mean a huge a mount of time to validate the block, and the amount of computation it involve is bound to grow at much higher rate than the amount of transaction to be processed.

Basically the amount of transaction a block can process grow linearly with the block size, the potential amount of computation grow quadratically, much faster. So it's not a good permanent solution to rely on eternal block size' increase to increase tx rate. Even with the
Moore's Law.

Moore law doesn't apply to blockchain because the main limit is not computing power. Maybe bandwith a bit more, but that i think can still be solved with the increase of bandwidth as optical fiber become more available, there is still lot of progress to be expected in this area.

If it was really about more law, it would be on the scale of initial block download, where 8 year of blockchain can be processed and checked in 3 days. The problem is not the processing power to validate blocks. A good client can validate 10-20 blocks / sec in average .. And this could be easily improved with exploiting more the computing power, using good multi core & opencl, you could probably multiply this speed without changing a thing to the protocol.


To me what it look like is LN/Segwit is definately the way forward, or something like it, who can allow a certain number of operation to happen outside of the main block chain conscensus. But realistically, i don't think LN will be reach mass adoption before 2-5 years.

And users driven development is always a mess. You never get anything done by asking users what they want .. Already to explain the inside and out both on the theory, the protocol, the usage and the implementation to average user is just impossible.

It's just starting to get there with the basic bitcoin protocol after years and years. And yet you could easily say large number of users just don't really know how bitcoin work, or really care more than simple curiosity, don't want to take responsability for the development, and if you start to implement anything that user might think is good, even if it make them happy, in the end, it rarely lead to well made software.

And now coming with something completely new like LN/segwit, extremely technical, who change completely the way bitcoin work, to people who already have bit hard with trusting the basics 100%, and the conflict of interest with miners, it's clear it's not going to be fully accepted in mass before some time. At least blockstream seem have good funding so they can hold some time and keep developing and explaining it for some time, but it doesn't look like it's going to be adopted in mass really just now.

And just to say, the way developers represent themselves, after it's not my buisness, but i see only in this thread, when the person who represent it speaks, they always give a sense of elitism and not really listening, which i don't think help a lot to make it accepted and gain trust that the developers are thinking for everyone and not just to push their own thing in their own interest.

But that said there is still the need to increase TX rate, and ln is not going to be accepted soon i think, it's the problem that everyone see, and maybe it would not be worst to think of a way to keep bitcoin going one or two more year the time a true permanent solution like segwit can be found and accepted by everyone, even if that involve just tweaking the block size up a bit, and taking all the step necessary regarding boundaries with computation power, bandwidth, potential issue that come up on all the level with this increase, even if it just raise to 2Mo in a first step.

But something that is not that radical that segwit for the moment, who doesn't change the way bitcoin work and doesn't disturb the eco system that much like LN, and can give a bit of oxygen to the bitcoin network processing rate.

Well again just my 2 satoshis

[Carlton Banks]

"5nm is far too close" ... for ... what exactly? What is it about 5nm that makes it "too close"? What occurs, or stops occurring at 5nm?

Is this a trick question?  Casual checking in... 

At some point down there you get too much electrical leakage to carry a signal.

Classical mechanics break down as quantum effects begin to appear.

Registers begin to act too much like qubits for digital reliability, which requires static discreet states (1 or 0).

The easiest gains were at the beginning.  Moore was observing the harvest of the low hanging fruit, nothing more.

Now it takes more capital to build a modern fab than it took to go to the moon.

jbreher knows all this anyway, he claims to work with the semi-conductor industry after all. Presumably he's in marketing and not anything technical, there is no need or demand for talented liars in technical fields (they have a tendency to screw things up lol)

Tell us again jbreher about the imminent feasibility of femtometer transistors for CPUs, I'm in the mood for a laugh

[jbreher]

"5nm is far too close" ... for ... what exactly? What is it about 5nm that makes it "too close"? What occurs, or stops occurring at 5nm?

Is this a trick question?  

Not a trick question. Carlton made an assertion that 5nm was "too close", but did not specify what it was too close for. I was merely asking him to clarify his assertion.

[jbreher]

Tell us again jbreher about the imminent feasibility of femtometer transistors for CPUs, I'm in the mood for a laugh

Now see there you go simply lying to try to  ... I dunno ... win Innertubes points? I never said anything about femtometer transistors, accordingly, there is no way I could "tell us again".