Will ASIC mining destroy Bitcoin?

[Gavin Andresen]

Several people have asked me privately via email what I think of Butterfly Labs and ASIC mining in general, so:

What do I think of ASICs hitting the "mining scene" ?

meh.

It shows that Bitcoin is considered stable enough for a company to invest a LOT of money on design and production on it, which is very good. ASICs were bound to happen if Bitcoin survived long enough.

The long-term outlook for mining is "mining will happen with very specialized hardware, in places where either electricity is free or generating lots of heat is a benefit and not a cost"  (and probably both; I still think Iceland will be a big mining hot-spot eventually).

I'm not worried about Butterfly Labs deciding to take over the blockchain with their superior hashpower. They'd be idiots to do something like try to mount a 51% attack-- they don't want to kill Bitcoin, they want it to get more successful so they sell more hardware.

And if they are successful they will very quickly have competitors.

Full disclosure:  I spent some of my bitcoins to pre-order their USB coffee-warmer doo-hickey.  It gets cold here in the winter.

[Steve]

I think I'm most excited about the $150 jalapeño.  There's been a lot of concern that GPUs changed mining from being something anyone can easily do, to something only for people willing to spend a lot of time and money.  While the jalapeño doesn't fully restore the accessibility of CPU mining, an efficient plug and play device for $150 comes about as close as you can get.  Anyone that has an interest in bitcoin and would like to help secure the network will soon have an option with a very low barrier to entry.

[SgtSpike]

I agree - I see ASIC as a sign that serious investors are taking Bitcoin seriously, which is a good thing!

[pekv2]

It's cool to see a programmer come in and pop his thoughts out. Good read indeed.

[InstantBTC]

Steve makes a valid point. I got into GPU mining back when GPU mining was first going. I pooled a few grand in with a buddy and we did pretty good. BFL's rigs are great because the cost per GH/s doesn't vary that much between models, and the biggest gap is between the Jalapeno and the Single, both affordable to most people.

Jalapeno : $42.85/GH

Single : $32.00/GH

Mini Rig : $29.90/GH

By upgrading to the Mini Rig, you only save ~$3,000 off compared to 25 Single's worth. This means that for somebody interested in mining, they can drop $150 in to give it a try, if they like what they see, or they enjoy being apart of it, they can pick up a single and the $/Ghash for them, doesn't differ too much from an investor who wants to drop $90,000 on Mini Rigs, after all, the investor who dropped $90,000 only saves $3.00/GH compared to the guy who bought the single. Compare this to how it was when I got into GPU mining...

I don't remember the exact numbers but, our $/MH (note: MH, not GH) was something like $0.50 less than somebody who could only spend $300-500. I really think this is good for bitcoin as a whole. Bad for anybody who recently got into GPU mining, but when I got out of GPU mining, cost of electricity with our rig was greater than amount of bitcoins we where mining.

[Peter Todd]

ASIC mining will make it much, much tougher for an entity to borrow a bunch of computing power to attack bitcoin.

For instance, lets suppose BFL sells just $5 million dollars worth of their "coffee warmers". (a highly conservative number that'd probably leave them bankrupt after they paid for the ASIC development costs) That's about 33 thousand coffee warmers, or 234TH/sec. Suppose the attacker decided to requisition a whole bunch of computers to attack Bitcoin, for instance by asking Amazon or Google "nicely" One 4-way Opteron CPU can do about 115MH/s, so for your 51% attack you'll need about 1 million CPU's. If you're renting from Amazon, that's costing you something like a million dollars an hour, assuming you could even get them to let you rent that much computing power. The capital cost of all that computing power is also in the range of hundreds of millions of dollars, heck, easily a billion dollars with server farm overhead.

Finding a whole bunch of GPU's is actually rather tough, as most GPU farms are for scientific computing and use floating-point optimized GPU's that aren't very good at computing hashes.

A final possibility is borrowing an FPGA farm. We could make the rough assumption that the value of the farm's FPGAs will have the same $/Hash ratio as BFL's currently shipping product. So that's 117TH/sec / 0.8GHash/Single * $600/Single = $87.7 Million dollars worth of FPGAs. Intel might have that kind of FPGA farm available - they're used for chip verification - but again, renting it won't be cheap. Also, it looks like BFL is getting it's FPGAs at pretty cheap prices - a $600 single has $2000 worth of FPGAs in it - so with wholesale discounts we still might need to triple or quadruple that $87 million.

With ASIC mining, the cheapest way to computationally attack Bitcoin is probably by doing a run of your own ASICs, and it's not something you can do quickly. All that effort and money just so you can find out the myriad ways that the devs can stop 51% attacks using techniques possible now that Bitcoin is widely established.


For instance, lets suppose the NSA decides to attack Bitcoin. They could probably round up the hundreds of millions of dollars worth of computing power to make it happen, although it'd be a big hit to their black budget. Chances are within a few hours to days the devs will respond with something like a "coin-age" rule and ask everyone to upgrade. Now blocks get rejected, and nodes blacklisted, if they try to pass blocks into the network that don't meet coin age requirements. Transactions start flowing again, although the price on Mt. Gox has dropped severely, lets say 50%. At the same time the "known-legit" mining pools are also taking steps to protect their investment, by temporarily centralizing a bit, and blocking connections to nodes that aren't on a whitelist; the "most-difficult-block-wins" rule has been temporarily suspended. Note that at this point it's still not possible for anyone to steal coins, and not much more possible to do double spends.

Now, one thing the NSA could do is buy a bunch of coins so their blocks get accepted again. The problem is, now they're basically giving people a way to get out of Bitcoin, and boosting the price on the exchanges, restoring confidence. Exactly what they don't want! If they do nothing, they're still burning at least hundreds of thousands of dollars an hour, while the network figures out ways to mitigate the damage.

Honestly, ordering some assassinations on the guys running major exchanges sounds a lot cheaper...

[Hexadecibel]

Honestly, ordering some assassinations on the guys running major exchanges sounds a lot cheaper...

I've felt recently that it would be easier (cheaper) for Gavin and other developers to simply suffer sudden "heart attacks".

We need to round up all our developers and put them in a "safety pit". For their safety.


But seriously, if we suddenly lost our core developers, where would bitcoin be? Do you think others would step up if it could mean assassination?

And I realize our developers being assassinated is not realistic, but then again I really really really don't want them all on the same plane on any business trips...  

edit: I just realized I'm derailing the thread. Too soon... so....

YAY ASIC MINING

[Peter Todd]

But seriously, if we suddenly lost our core developers, where would bitcoin be? Do you think others would step up if it could mean assassination?

Sure, but this time, using tor. And, no offence to Gavin and the rest of the Bitcoin devs, but the development being done right is something that a large number of developers could do. If they had to step down for some reason, other people could be found to fill their place.

[Gabi]

ASICs are a good thing for Bitcoin, it's just the evolution of mining, from generalized hardware (gpu) to specialized hardware. Much much much more hashing power, much higher security.

[AmazonStuff]

If ASICs are going to be owned only by a small group of people, then it won't be good for BTC.

[Vladimir]

as soon as one realises that:

1. Good for CPU  miners != Good for Bitcoin
and
2. Good for GPU  miners != Good for Bitcoin
and
3. Good for FPGA miners != Good for Bitcoin
and
4. The higher the Bitcoin nework difficulty the more secure Bitcoin network is and therefore
Migration to ASIC mining and combined Bitcoin hashing power measured in ExoHashes as opposed to present TeraHashes  = Good for Bitcoin.

So if you want to whine about ASIC's try to be honest and not bring  "good of Bitcoin" into it. Go and whine like "ohh by October my FPGA will not pay for itself, my investment went baad, gotta go to work for Wallmart again!"

[cbeast]

I'm holding out for cheap quantum miners! 

[Turbor]

So if you want to whine about ASIC's try to be honest and not bring  "good of Bitcoin" into it. Go and whine like "ohh by October my FPGA will not pay for itself, my investment went baad, gotta go to work for Wallmart again!"

This. Most of the whiners just fear that they wont get rich anymore with asic around. Now they can prove that they're in for the project and not only to profit. This 51% bla bla just sucks. My fpga will be paid off when asic hits the market I'm ready for the future.

[lame.duck]

So if you want to whine about ASIC's try to be honest and not bring  "good of Bitcoin" into it. Go and whine like "ohh by October my FPGA will not pay for itself, my investment went baad, gotta go to work for Wallmart again!"

This. Most of the whiners just fear that they wont get rich anymore with asic around. Now they can prove that they're in for the project and not only to profit. This 51% bla bla just sucks. My fpga will be paid off when asic hits the market I'm ready for the future.

The problem is, there is no other use for a mining ASIC besides mining, so after the sales to the community goes down, BFL could only make more profit by selling more chips. What if now the evil whoever will ask them for 'The network hash rate + x%' in Chips with usual discounts?

[SgtSpike]

So if you want to whine about ASIC's try to be honest and not bring  "good of Bitcoin" into it. Go and whine like "ohh by October my FPGA will not pay for itself, my investment went baad, gotta go to work for Wallmart again!"

This. Most of the whiners just fear that they wont get rich anymore with asic around. Now they can prove that they're in for the project and not only to profit. This 51% bla bla just sucks. My fpga will be paid off when asic hits the market I'm ready for the future.

The problem is, there is no other use for a mining ASIC besides mining, so after the sales to the community goes down, BFL could only make more profit by selling more chips. What if now the evil whoever will ask them for 'The network hash rate + x%' in Chips with usual discounts?

Then don't buy it.

[lame.duck]

Then don't buy it.

Sure, but this doesn't solve the problem that a bad guy will get the mining ASIC at a much lower $ - GHash/s ratio.

[crazyates]

I'm holding out for cheap quantum miners! 

Quantum miners aren't really that great for mining. We deal almost exclusively with int numbers.

[SgtSpike]

Then don't buy it.

Sure, but this doesn't solve the problem that a bad guy will get the mining ASIC at a much lower $ - GHash/s ratio.

So can everyone else at that point.

[Comepradz]

Seriously, what if the developers decided to change the block hashing algorithm from double SHA-256 to something else (WHIRLPOOL or SHA 3 maybe)? doesn't it renders ASIC useless?

[Vladimir]

Then don't buy it.

Sure, but this doesn't solve the problem that a bad guy will get the mining ASIC at a much lower $ - GHash/s ratio.

But this solves the problem that bad guys could get large number of the mining CPU's GPUP's and FPGA's at whatever price.