About three months ago, Butterfly Labs was the first to announce a specialized Bitcoin mining device based on a technology with the potential to massively increase the computing power of the Bitcoin network per unit power and cost: the ASIC. ASIC, or application specific integrated circuit, is a technology taking specialization to a whole new level. Although FPGAs, or “field-programmable gate arrays”, are already used for Bitcoin mining, ASICs offer even greater improvements because they come specialized for one specific task right out of the factory. The process to create ASICs is hardly new, as they have been used for applications like digital voice recording for many years, but it has taken time to adapt the process to Bitcoin mining.
The statistics of Butterfly’s ASIC-based Bitforce SC lineup are impressive; dollar for dollar, the SCs are about thirty times as powerful as anything currently available, and, given the state of the Bitcoin network at the time of the announcement, each one would pay for itself in eight days. More recent specifications from Butterfly show that the power consumption is as low as the price; the middle-of-the-line $1299 SC Single would provide 60 gigahashes per second (up from 40 as per the original announcement) for only 60 watts, less than the company’s current FPGA offering.
The customer service offerings are as impressive as the product itself. Butterfly Labs now intends to offer a lifetime warranty on all of its ASIC products. As CEO and lead engineer Nasser Ghoseiri put it, “We’re making an infrastructure. Hardware is only part of it – our mission is to support the value of their investment by responding to market changes with things like the buy-back program, and making sure that their investment is safe.” Will the units actually last a lifetime? Butterfly’s Josh Zerlan says yes. The Bitforce SC ASICs are “not like the FPGA units at all… they are going to be far simpler and much more robust.”
Of course, many are incredulous, and ever since the original announcement, there have been concerns that the SCs would not actually be created and the whole thing was a scam – a worry that was heightened when it became public that one of the company’s employees had a criminal history. Some were also worried that if Butterfly’s offering was legitimate, it would be unhealthy for one company to have so much influence in a market, the whole point of which was to be decentralized.
Now, however, several other companies have joined the fray. The first is BTFPGA’s bASIC, which offered 27 GH/s for $1069 (a figure which has since been increased to 54 GH/s). The bASIC is intended to be the ASIC “successor” to BTCFPGA’s existing FPGA-based ModMiner Quad. Just like Butterfly Labs, BTCFPGA is offering a trade-in program by which existing ModMiner Quad orders that have been placed but not shipped can be replaced with orders for a bASIC, and any ModMiner Quads that are already in use can be traded in for a rebate. Also just like the ModMiner, the bASIC is intended to be completely open source.
Another option available to Bitcoin miners is the Avalon ASIC. Avalon’s ASIC, a project created under the umbrella of the upstart Bitcoin business incubator BitSyn.com, offers 60 GH/s for $1299. However, its power consumption is more worrying. An initial announcement suggested that the machine would consume 600 W, although a later post on the BitSyn.com forums stated that 600 W was a conservative estimate, and the actual power consumption would likely be 120-160 W. Given that Butterfly themselves took almost three months to come up with a power consumption estimate, and bASIC does not have any figures at all, the lack of precise specifications at this point is understandable; the ASICs are still under development, and shipping for the first 300 units will take place in January and February 2013.
Finally, DeepBit has released a line of ASICs of their own. DeepBit is well known for having been, at one point, the largest Bitcoin mining pool. DeepBit even exceeded 50% of the network’s total hash power for a few brief moments in 2011, although the mining pool has shrunk considerably in recent months, and is now not even close to first place. DeepBit has been making preparations for this ASIC offering since January, and is seeking to reclaim some of its former glory with the ASIC-based Reclaimer line. However, the results are not particularly impressive; the largest and most efficient of the line offers 80 GH/s for $2800. DeepBit may readjust its specifications in the future as Butterfly and bASIC have done, but at this point, its offering appears to be the least attractive of the four. Deepbit operator Tycho expects the Reclaimers to become available in March.
The sudden diaspora of competition on the Bitcoin mining market provides several positive signs. First, the existence of offerings from established players like Deepbit, and the similarity of the statistics across all of the companies, suggests that the ASIC offerings are indeed legitimate, and will provide the hashing power and efficiency at least near what they are offering. Second, it alleviates fears of market centralization. Finally, the development of ASICs in general is a very positive sign for the security of the Bitcoin network as a whole.
Bitcoin’s hashing power is likely to go up by at least a factor of 10 to about 200 terahashes per second, or 3 exaflops – likely the first use of the prefix “exa” (meaning 1018) ever to be employed in the computing world. Even billion-dollar government supercomputers are hard-pressed to match just one percent of that. From here, Essentially, this implies that the condition for overpowering the security of the Bitcoin network is changing. Formerly, all that one needed to do to overpower the rest of the network and gain the ability to pull off double spending frauds was to gain more hashing power than the rest of the Bitcoin network combined for a short amount of time. Now, of course, the same still holds, but doing so requires buying up large quantities of ASICs – in short, throwing in more capital investment than the rest of the Bitcoin network combined at least over the past few months. Right now, even with the ASIC revolution increasing the power of the Bitcoin network by a factor of ten, that is still only a few million dollars – although even now any attack will take months to prepare. From here, however, further improvements to mining technology will only come with iterative applications of Moore’s law, and so the capital barrier to pulling off a double spend will only go up as times goes on. If the Bitcoin network was precarious and vulnerable to any significant company, government or that may have wanted to bring it down before, it is now well on the path to lasting stability.