Mining Difficult and Moore's Law

[CurbsideProphet]

Transistor density doubles roughly every two years, as Moore's law attests.  Information technology therefore advances exponentially.

Some are of the opinion that mining will become increasingly more difficult thus limiting the supply of BTC entering the market.  While this may be true, short-term, it does not account for the exponential increase in technology.  If Moore's law is correct, wouldn't it be rather easy for technology to keep up with the ever increasing mining difficulty?

One could possibly even argue that at some point, technology will surpass and SHA-256 may be cracked all at once resulting in a flood of the remaining 21 million BTC's in a very short period of time.  Mathematical weaknesses in SHA-1 were found back in 2005, who's to say another breach in the current encryption will not happen? The 2033 target date assumes perfect intervals.  Moore's law seem to contradict this expected curve.

Remember, we're talking about exponential growth.  If you haven't seen a graph of exponential growth, think of a hockey stick.  A long near horizontal line followed by a very rapid vertical increase.  Once we hit that vertical increase, the 10-minute interval assumption goes right out the window.  Thoughts?

[1714951159]

1714951159

[1714951159]

1714951159

[1714951159]

1714951159

[1714951159]

1714951159

[1714951159]

1714951159

[kokjo]

its not encryption, its called hashing. get your terms right before you post.

[Jaime Frontero]

Once we hit that vertical increase, the 10-minute interval assumption goes right out the window.  Thoughts?

the interval assumption is not really an assumption - it is the way the software is designed to work through difficulty increases.  over a long period of time the 10-minute/block value should hold up to pretty much any technology advance.

[CurbsideProphet]

its not encryption, its called hashing. get your terms right before you post.

Fine hashing.  Are you now going to contribute anything or are we going to continue to play semantics?  I'm pretty sure you understood the premise of my question.

[imperi]

Moore's law, in the sense of computing power, does not work as quickly as the increase in difficulty, even if both are exponentially based.

[CurbsideProphet]

Once we hit that vertical increase, the 10-minute interval assumption goes right out the window.  Thoughts?

the interval assumption is not really an assumption - it is the way the software is designed to work through difficulty increases.  over a long period of time the 10-minute/block value should hold up to pretty much any technology advance.

It's designed to increase difficulty but isn't it at a static rate?  What I'm trying to figure out is if mining difficulty increases at a static rate but technology increases at an exponential rate, wouldn't there be a point where technology surpasses?  Isn't that basically inevitable?

[[Coins!]]

Moore's law, in the sense of computing power, does not work as quickly as the increase in difficulty, even if both are exponentially based.

This.

Moore's law doubles computation speed (aka transistor count) every 2 years. (730 days)

BTC difficulty has been increasing by 50% every 9 days. Therefore doubling every ~16 days

Doubling every ~16 days >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Doubling every 730 days.

[Jaime Frontero]

Once we hit that vertical increase, the 10-minute interval assumption goes right out the window.  Thoughts?

the interval assumption is not really an assumption - it is the way the software is designed to work through difficulty increases.  over a long period of time the 10-minute/block value should hold up to pretty much any technology advance.

It's designed to increase difficulty but isn't it at a static rate?  What I'm trying to figure out is if mining difficulty increases at a static rate but technology increases at an exponential rate, wouldn't there be a point where technology surpasses?  Isn't that basically inevitable?

no.  it does not increase at a static rate, but rather at a rate calculated on what is happening now.  it's quite fluid.

[imperi]

Moore's law, in the sense of computing power, does not work as quickly as the increase in difficulty, even if both are exponentially based.

This.

Moore's law doubles computation speed (aka transistor count) every 2 years. (730 days)

BTC difficulty increases by 50% every 9 days. Therefore doubling every ~16 days

Doubling every ~16 days >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Doubling every 730 days.

The top super computer in 2011 is 500x or so more powerful than the top one in 2001. The data is at top500.org if you need me to quote it. It is faster than every two years, even taking into consideration the different types of computers.

Doubling every two years means 32x performance.

A modern CPU + GPU can do a lot more than 32x performance than a 2001 computer.

[[Coins!]]

Moore's law, in the sense of computing power, does not work as quickly as the increase in difficulty, even if both are exponentially based.

This.

Moore's law doubles computation speed (aka transistor count) every 2 years. (730 days)

BTC difficulty increases by 50% every 9 days. Therefore doubling every ~16 days

Doubling every ~16 days >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Doubling every 730 days.

The top super computer in 2011 is 500x or so more powerful than the top one in 2001. The data is at top500.org if you need me to quote it. It is faster than every two years, even taking into consideration the different types of computers.

A modern CPU + GPU can do a lot more than 32x (doubling every 2 years) computing than a 2001 computer.

My quoting of you was to agree with you.

; )

But even so, with 16 days being the doubling difficulty for BTC, that is still way faster than Moore's law, or the current technical advances.

[Epinnoia]

So according to Moore's law, computing power should double in two years.  

And how much has bitcoin's difficulty level increased in those same two years?

Yes, both computing power and the difficulty levels increase exponentially.  But just because they're both exponential doesn't mean that one will reach or otherwise catch up with the other. 

OP, you really don't have an argument here, unless there is a realistically realizable upper limit to the built-in difficulty increases.  

[CurbsideProphet]

The data is at top500.org if you need me to quote it.

Thanks for the link.  So if I'm reading the data correctly, in November 2010 the Chinese Tianhe-1A was the #1 supercomputer at 2.57 petaflop/s.  Roughly six months later the K Computer was over 3 times faster at 8 petaflop/s.  

With that type of growth rate, it's still hard for me to wrap my mind around how a technology created now (fluid or not) can account for the ever increasing power of technology.  In the short-term (which we are in) sure.  But out a couple of decades?  

[kokojie]

Nothing can grow exponentially forever, Moore's law is coming to an end soon.

[imperi]

The data is at top500.org if you need me to quote it.

Thanks for the link.  So if I'm reading the data correctly, in November 2010 the Chinese Tianhe-1A was the #1 supercomputer at 2.57 petaflop/s.  Roughly six months later the K Computer was over 3 times faster at 8 petaflop/s.  

That this jump was so close was coincidental.  There have been periods where the top computer remained as #1 for a few years, and the next wasn't substantially better.

With that type of growth rate, it's still hard for me to wrap my mind around how a technology created now (fluid or not) can account for the ever increasing power of technology.  In the short-term (which we are in) sure.  But out a couple of decades?  

People said the same thing a few decades ago. The scaling of nanolithography has never slowed. Intel plans to have 14 nm chips by 2014. Also, scaling transistors vertically is the next break-through paradigm shift and there has been great innovation here. This will further increase computing power per watt and dollar incredibly.

On another note, optimizations in the transistor layout also has not slowed, such as better pipelining, etc. Lastly, there's been improvements in C compilers.

Everything I've mentioned allows for the sustained exponential increase.

Carbon nanotube based technology is another future possibility, but isn't necessary for this trend to continue.

[Epinnoia]

With that type of growth rate, it's still hard for me to wrap my mind around how a technology created now (fluid or not) can account for the ever increasing power of technology.  In the short-term (which we are in) sure.  But out a couple of decades?  

Do you have any idea of the maximum limit to the difficulty increases?  If there was no limit on the difficulty levels, computing power improvements would be irrelevant.  It just gets extremely difficulty to solve blocks.  There is no problem here, unless the upper limit of the difficulty increases can be realistically attained.

Even if all the worlds super computers were focused on solving bitcoin blocks, it wouldn't kill bitcoin.  It would simply make it so difficult for us regular GPU miners to solve blocks that we'd probably all quit.  

[CurbsideProphet]

Do you have any idea of the maximum limit to the difficulty increases?

No idea.  Do you?
  

There is no problem here, unless the upper limit of the difficulty increases can be realistically attained.

Exactly, which is why I brought up the subject.  If the maximum limit cannot be obtained, then there is no problem.  However, IF the maximum limit can be obtained then we have a pretty big problem.

I think those taking offense are assuming that I'm making some sort of statement.  I'm not.  I'm just trying to think of potential scenarios/pitfalls and see if people can shoot holes through them. 

My basic concern is this, if computing power can grow exponentially, how can the developers of Bitcoin account for this over the long-term (read: decades).  Unless they have a crystal ball, I don't see how that's possible.  Remember, the BTC model is already in place, they can't go back and modify it at some point in the future to account for unforseen events. 

[Epinnoia]

My basic concern is this, if computing power can grow exponentially, how can the developers of Bitcoin account for this over the long-term (read: decades).  Unless they have a crystal ball, I don't see how that's possible.  Remember, the BTC model is already in place, they can't go back and modify it at some point in the future to account for unforseen events.  

Let's say that the code permitted the difficulty to raise by a factor of 2 raised to the power of some 32-bit integer.  I suspect that the Sun would die off before that happened.

[Epinnoia]

Remember, the BTC model is already in place, they can't go back and modify it at some point in the future to account for unforseen events.  

I am not so sure that is true.  They could modify the coding of the difficulty increases such that it grandfathered in all previous transactions at the old difficulty limit, and then have a new formula for all transactions which occur after a specified block number.

Using my previous example...  If the old limit was 2 raised to the power of a 32-bit integer, they could change that to 2 raised to the power of a 64-bit integer (which is a mind-boggling large number).  They could set the code so that it would grandfather all transactions under block number XXX,XXX at the old formula.

[DamienBlack]

The maximum difficulty with SHA-256 is 2^224.

If we continue to double every month, as we have been, (which is very unlikely), we would reach it in about 200 months from now or 18 years. At that point the hashing power of the network would be gigantic. The network would use magnitudes more electricity then the entire world uses now.

If we just continue at Moore's law (double every two years), it'll be about 400 years before we reach the maximum difficulty.

And, if that happens, we just switch to SHA-512 which will give us another 400 of exponential growth. This modification would be simple.

And by the way, the "weaknesses" in SHA1 just make it so that someone can hash a little faster with certain algorithms. It would still work for bitcoins, the difficulty would just be a little higher.