ASIC = The end of decentralized mining

[crazyates]

ASIC are like Buying a Car that can only turn Right.(SHA-256 ASIC)

What happens if someone finds a pot hole in the road? (SHA-256 Weakness)

Turn left 3 times successively?
Metaphor fails 

We're talking about Bitcoins, not Nascar!

[notme]

ASIC are like Buying a Car that can only turn Right.(SHA-256 ASIC)

What happens if someone finds a pot hole in the road? (SHA-256 Weakness)

Turn left 3 times successively?
Metaphor fails 

We're talking about Bitcoins, not Nascar!

ASIC is like buying a chip that can only do one thing that is not very useful outside bitcoin mining.

Okay, that's not really a metaphor, but I think it makes the point clear enough.

[nedbert9]

ASIC are like Buying a Car that can only turn Right.(SHA-256 ASIC)

What happens if someone finds a pot hole in the road? (SHA-256 Weakness)

Turn left 3 times successively?
Metaphor fails succeeds  

We're talking about Bitcoins, not Nascar!

ASIC is like buying a chip that can only do one thing that is not very useful outside bitcoin mining.

Okay, that's not really a metaphor, but I think it makes the point clear enough.

ASIC is like Raisin Bran without the bran.   And we all know that bran is the catalyst of decentralization.

[runeks]

Aren't we getting our figures a bit mixed together in this thread? I mean, it's obvious that the $/GH figure will go way down, that one's obvious. But I think a previous poster was right when he claimed that efficiency wouldn't increase more than around 10x.

The efficiency increases from CPU, to GPU, to FPGAs, and to ASIC are all comparable. Roughly a 10x increase every time:
40nm CPUs do up to about 0.1 to 0.5 Mhash/Joule.
40nm GPUs about 1 to 2 Mhash/Joule.
40nm FPGAs about 20 Mhash/Joule.
40nm ASICs will do about 200 Mhash/Joule.

We've always been using ASICs. A CPU is an ASIC, a GPU is and an FPGA is. The reason the great leap (much more than 10x) with regards to cost per hash takes place is because instead of us buying pre-manufactures ASICs (CPUs, GPUs, FPGAs) from someone else, we'll be making them ourselves. That's like the difference between buying apples in a store and planting your own tree in your garden, the costs are several order of magnitudes less in the latter case.

The efficiency of the previously mentioned SHA-3 (and SHA256) chip is about 14.1 W/GH as far as I can calculate. See here and here. That's around 71 MH/J for the unoptimized tiny SHA256 core running at 50 MHz. There would probably be a lot to gain from optimizing the design, but this figure would also need to increase a lot to surpass 10x as this is only ~3x the efficiency of this FPGA: http://fpgamining.com/documentation/hardware/power-measurements

With that being said, it's obvious that the efficiency figure will become increasingly unimportant as we reach these levels. No one is going to be spending a fraction of their investment on power with custom SHA256 ASICs, practically all the cost will be from the hardware.

[siggy]

First you say....

We've always been using ASICs. A CPU is an ASIC, a GPU is and an FPGA is. The reason the great leap (much more than 10x) with regards to cost per hash takes place is because instead of us buying pre-manufactures ASICs (CPUs, GPUs, FPGAs) from someone else, we'll be making them ourselves. That's like the difference between buying apples in a store and planting your own tree in your garden, the costs are several order of magnitudes less in the latter case.

Then you say....

With that being said, it's obvious that the efficiency figure will become increasingly unimportant as we reach these levels. No one is going to be spending a fraction of their investment on power with custom SHA256 ASICs, practically all the cost will be from the hardware.

Or to re-iterate..
your point one:  ASIC hardware is going to be dirt cheap. 
your point two:  ASIC hardware is going to be the majority of ROI when compared to electricity...

Does anyone except me see the disconnect here?
 
Sure, in the short term immediately after the introduction of ASICs, the cost of hardware will drown out the cost of electricity (ASIC hardware is NOT dirt cheap at this point)..
However, as more and more ASIC units come online, the difficulty will sky-rocket. As even more ASICs come up for sale, the ASIC manufacturers will need to start dropping the price or their market will dry up (assuming most miners can do simple ROI calculations). The manufacturers will need to keep droping prices as difficulty incresses.. This will continue till we get to the point where hardware really IS dirt cheap.  At this point difficulty will be so astronomically high, ROI will no longer be so much dependant on hardware but once again, will be majorly based upon the cost of electricity.

Historically (during the golden age of GPU's) the pain point has been around 20 cents per KWh.  If you pay more than that, it is generally a losing proposition to buy GPU's for the sole purpose of mining.

I predict that a year after ASICs go into full bore production, the pain point will be around 20 cents per KWh.  If you pay more than that, it will be generally a losing proposition to buy ASICs for the sole purpose of mining.  Since bitcoin ASICs will be Application Specific, this applies even more to ASICs than GPU's.

Sigg

[runeks]

First you say....

We've always been using ASICs. A CPU is an ASIC, a GPU is and an FPGA is. The reason the great leap (much more than 10x) with regards to cost per hash takes place is because instead of us buying pre-manufactures ASICs (CPUs, GPUs, FPGAs) from someone else, we'll be making them ourselves. That's like the difference between buying apples in a store and planting your own tree in your garden, the costs are several order of magnitudes less in the latter case.

Then you say....

With that being said, it's obvious that the efficiency figure will become increasingly unimportant as we reach these levels. No one is going to be spending a fraction of their investment on power with custom SHA256 ASICs, practically all the cost will be from the hardware.

Or to re-iterate..
your point one:  ASIC hardware is going to be dirt cheap. 
your point two:  ASIC hardware is going to be the majority of ROI when compared to electricity...

Does anyone except me see the disconnect here?
 
Sure, in the short term immediately after the introduction of ASICs, the cost of hardware will drown out the cost of electricity (ASIC hardware is NOT dirt cheap at this point)..
However, as more and more ASIC units come online, the difficulty will sky-rocket. As even more ASICs come up for sale, the ASIC manufacturers will need to start dropping the price or their market will dry up (assuming most miners can do simple ROI calculations). The manufacturers will need to keep droping prices as difficulty incresses.. This will continue till we get to the point where hardware really IS dirt cheap.  At this point difficulty will be so astronomically high, ROI will no longer be so much dependant on hardware but once again, will be majorly based upon the cost of electricity.

Historically (during the golden age of GPU's) the pain point has been around 20 cents per KWh.  If you pay more than that, it is generally a losing proposition to buy GPU's for the sole purpose of mining.

I predict that a year after ASICs go into full bore production, the pain point will be around 20 cents per KWh.  If you pay more than that, it will be generally a losing proposition to buy ASICs for the sole purpose of mining.  Since bitcoin ASICs will be Application Specific, this applies even more to ASICs than GPU's.

Sigg

You make a good point. I guess, as you say, it depends on what impact custom ASICs will have on the difficulty. As far as I can see, if we replace all current mining devices with devices that are 10x as efficient and the difficulty increases by a factor of 10, then there would be no difference - at all. Except BFL will have made a fortune .

I guess that's why a ROI of a year or less is preferable with these devices. Some might say that one year is too little to expect, but with a market that moves this fast it might not be. Ie., as I said, if difficulty increases by 10 times during the first year of operation with custom ASICs, then power efficiency-wise - kWh spent per $ in mining income - the miner will be in the exact same situation as before, only with added hardware costs.

I can understand why BFL is entering this market and not mining themselves. They must have made these calculations. I imagine not all the people on the BFL ASIC waiting list have.

[Ascholten]

THose who want to get in quick will pay, just like with any other processor or spiffy new video card.
Within a year, the price for that card is in the basement, the manufacturer made their money 100 times over and your card is just so Jan 2012.
The price of bitcoins is in the toilet because there is such a flood on the market, but that will quell soon as the difficulty increases to the moon.

In the end, if you want any you will need an asic, because otherwise it's just sheer luck, like cpu mining is now.

Aaron

[runeks]

THose who want to get in quick will pay, just like with any other processor or spiffy new video card.
Within a year, the price for that card is in the basement, the manufacturer made their money 100 times over and your card is just so Jan 2012.
The price of bitcoins is in the toilet because there is such a flood on the market, but that will quell soon as the difficulty increases to the moon.

In the end, if you want any you will need an asic, because otherwise it's just sheer luck, like cpu mining is now.

Aaron

What do you mean by this? The price of bitcoins doesn't seem to be declining, to put it mildly.

[bitarrow]

I assume he is speculating the bitcoin price will tank once the ASICS arrive and all the miners are dumping coins.
But hopefully the ASIC miners can not sell at market value and be smart with the selling of their coins. I will.

[runeks]

I assume he is speculating the bitcoin price will tank once the ASICS arrive and all the miners are dumping coins.
But hopefully the ASIC miners can not sell at market value and be smart with the selling of their coins. I will.

But why would miner's be selling a larger share of their coins compared to now, just because mining is powered by a custom ASIC? I don't get the logic. Also, coin reward will drop to 25 BTC per block at the most two months after the devices are released. I doubt an increase in miner's mining for profitability (who isn't mining for profitability already, anyway?) will more than offset the halving of new coins entering the market.

[bitarrow]

I assume he is speculating the bitcoin price will tank once the ASICS arrive and all the miners are dumping coins.
But hopefully the ASIC miners can not sell at market value and be smart with the selling of their coins. I will.

But why would miner's be selling a larger share of their coins compared to now, just because mining is powered by a custom ASIC? I don't get the logic. Also, coin reward will drop to 25 BTC per block at the most two months after the devices are released. I doubt an increase in miner's mining for profitability (who isn't mining for profitability already, anyway?) will more than offset the halving of new coins entering the market.

If you compare todays bitcoin price to a 1-2 months ago, we are double the value of bitcoin. So with reward cutting in half, as long as the bitcoin price is doubled then its a wash. I don't see the price being below $10 for any length of time so reward halving shouldnt be a factor.

[fabiomiguel]

Hello guys. I'm following this topic very close and some people don't know what are saying.

This is some relevant quotes from https://en.bitcoin.it/wiki
Anyone with enough computing power can take over the network?

CONFIRMED, see Weaknesses.
That said, as the network grows, it becomes harder and harder for a single entity to do so. Already the Bitcoin network's computing power is quite ahead of the world's fastest supercomputers, together.
What an attacker can do once the network is taken over is quite limited. Under no circumstances could an attacker take anybody else's money. An attacker's capabilities are limited to taking back their own money that they very recently spent, and preventing other people's transactions from receiving confirmations. Such an attack would be very costly in resources, and for such meager benefits there is little rational economic incentive to do such a thing.
Furthermore, this attack scenario would only be feasible for as long as it was actively underway. As soon as the attack stopped, the network would resume normal operation.

The value of bitcoins are based on how much electricity and computing power it takes to mine them

This statement is an attempt to apply to Bitcoin the labor theory of value, which is generally accepted as false. Just because something takes X resources to create does not mean that the resulting product will be worth X. It can be worth more, or less, depending on the utility thereof to its users.
In fact the causality is the reverse of that (this applies to the labor theory of value in general). The cost to mine Bitcoins is based on how much they are worth. If Bitcoins go up in value, more people will mine (because mining is profitable), thus difficulty will go up, thus the cost of mining will go up. The inverse happens if bitcoins go down in value. These effects balance out to cause mining to always cost the amount of bitcoins it produces.

Oh btw:
http://img193.imageshack.us/img193/8462/totalbitcoinsovertimegr.png

No matter what happens this will be allways true.

This and alot more at: https://en.bitcoin.it/wiki

[-ck]

This and alot more at: https://en.bitcoin.it/wiki

http://hyperboleandahalf.blogspot.com.au/2010/04/alot-is-better-than-you-at-everything.html

[runeks]

This and alot more at: https://en.bitcoin.it/wiki

http://hyperboleandahalf.blogspot.com.au/2010/04/alot-is-better-than-you-at-everything.html

LOL!

[candoo]

Well,

there is at least one thing we're not considering right now: BFL could start selling ASICs which are underclocked so that while being better than the average FPGA they don't give out the maximum available hashing power.

Doing so they can, next year, sell their ASIC/2 board, with higher clock and same price just like GPU vendors do all the time.

I mean what makes you think that the first board will be 100x when a 10-20x underclocked or otherwise "crippled" board can be sold as well?

[Korbman]

I've been following this topic for a bit as well. Though there is a lot of speculation, Fabiomiguel makes a good point:

Oh btw:


No matter what happens this will be allways true.

This and alot more at: https://en.bitcoin.it/wiki

No matter what happens, the creation of bitcoins will ALWAYS maintain it's curve to 21 million. Difficulty keeps it in line. The more ASICs that come to market, the higher the difficulty. It will be the same as the transition from CPU to GPU and GPU to FPGA...from MH/s to TH/s. Most of the mining that occurs right now is done by people dedicated to the cause.

Now, I do see where 'centralization' comes into play here. The "average joe" that bought a $200 card to get 300MH/s is also probably the person who will buy a BFL SC Jalapeno for $150 to get 3.5GH/s. Only the extremely dedicated and financially capable will go into the TH/s range...which is a handful of people so far.

NOTE: I'm actually writing up a sort of analysis about this, taking into account a fair amount of math and speculation that we've touched on during this thread so far. I'll link it here when it's done and edited.
Figured I should at least get some thoughts out into this thread first

EDIT: The first part of my analysis: https://bitcointalk.org/index.php?topic=102268.0

[Gabi]

Under no circumstances could an attacker take anybody else's money. An attacker's capabilities are limited to taking back their own money that they very recently spent, and preventing other people's transactions from receiving confirmations.

If the attacker make a blockchain starting from like a year ago he will have all the coins mined from a year ago until today so if you have coins that were mined six months ago, the attacker will have them. So yes, the attacker can take anybody else's money.

[amgomez]

...
Now, I do see where 'centralization' comes into play here. The "average joe" that bought a $200 card to get 300MH/s is also probably the person who will buy a BFL SC Jalapeno for $150 to get 3.5GH/s. Only the extremely dedicated and financially capable will go into the TH/s range...which is a handful of people so far.
...

Also, now you can buy a card from a bunch of different builders (Asus, sapphire, ...) at your closest store. Jalapeños can only be bought to BFL, and only in the states, which causes not only delays but also troubles at customs.

[Garr255]

Jalapeños can only be bought to BFL, and only in the states, which causes not only delays but also troubles at customs.

I offer a proxy service in the US for FPGAs and ASICs for 10% of the mining revenue.
</advertisement>

[TheDom]

I don't think it is the end of decentralized mining at all. I just started mining a few weeks ago on a 5770 I just happened to have when I rediscovered Bitcoin, and now plan on ordering a Single SC. The way I see it if it is profitable, great, if not, it will cost so little in power compared to the activism credits I get for supporting Bitcoin.