A Simple Explanation of ASICs

An ELI5 explanation of ASICs, decrypting marketing speak from the tech.

Arjun Balaji
Crypto Simplified
6 min readFeb 26, 2018

--

Bitmain’s top-tier Bitcoin ASIC

“ASIC resistance” has been a popular buzzword that many alt-coins use to describe their relative advantages. As with much of the technical jargon used when describing different cryptocurrencies, it’s important to distill this to first principles. Is this marketing speak or a legitimate feature?

What are ASICs? 🤷

ASIC stands for “Application Specific Integrated Circuit.” While that’s a mouthful, you can understand it as “computer chip designed to complete a specific task.” ASICs are designed for all sorts of specialized computing tasks: while they are limited in their use, they are extremely effective when operating on the specific tasks they are designed for.

An ASIC can be thought of as a race car: very expensive to design and manufacture and very useful for competitive racing. However, it’s sub-optimal for highway driving or once it falls off peak performance.

Why ASICs? 🤔

Producers and consumers of ASICs must run a cost-benefit analysis to consider the trade-off with efficiency. Are the efficiency benefits of leveraging ASICs so much greater that the cost of producing a specialized processor? When developing a general processor, the cost is spread across many different products.

Bitcoin ASICs can only be used for mining bitcoin (or another alt-coin that uses the same hashing algorithm), so the arms race in development for bitcoin-focused ASICs began when bitcoin reached a high enough price for it to be worth it.

How did ASICs come to Bitcoin? 📚

First, Bitcoin was mined with ordinary central processing unit (CPUs), then specialized graphics cards (GPUs). CPUs and GPUs are found in regularly home PCs. Gradually, the first specialized hardware, field programmable gate arrays (FPGAs), were designed. Gradually, as competition for mining bitcoin increased, people started to design ASICs specifically for Bitcoin’s SHA-256 hashing algorithm.

Since mining bitcoin involves brute forcing solutions to capture the block reward, these ASICs could iterate over solutions faster (and were more power-efficient to boot). To put it in context: top-of the line ASICs can compute hashes for Bitcoin up to 100,000 times faster than CPUs.

The largest manufacturers of GPUs are AMD and NVIDIA, where the original use case was in graphics (and gaming!). As machine learning and crypto-mining becomes more prevalent, the value of these companies has soared, with hardware shortages everywhere.

Who makes & uses ASICs now? 🏭

Now, a number of Proof of Work based tokens have ASICs designed for them. The largest producer of [crypto] ASICs in the world, Bitmain, also controls more of the Bitcoin hash rate than any other party — a fact not glossed over by ASIC haters. As it stands, mining Bitcoin with non-specialized hardware is economically unfeasible given the massive cost and limited efficiency of using GPUs.

The amazing thing about Bitcoin ASICs is that, as hard as they were to design, analysts who have looked at this have said this may be the fastest turnaround time — essentially in the history of integrated circuits — for specifying a problem, which is mining Bitcoins, and turning it around to have a working chip in people’s hands

— Joseph Bonneau, Princeton University (via Princeton’s Bitcoin lectures)

What do the advocates say? 💪

ASICs offer security to these networks. ASICs are made by a limited number of startups and are not widely available. Given the difficulty of sourcing ASIC hardware, it would be very hard (err, impossible) for even a nation-state level actor to acquire enough hash rate to mount a credible 51% attack.

What are the drawbacks? 💩

ASIC production by a small number of companies is a double-edged sword. Many of these producers also operate massive vertically integrated mining operations. Mining in large-cap tokens has largely become industrialized and (relatively) centralized as a result of massive barriers to entry. People that disfavor ASICs often claim that the original vision for Bitcoin was that it will be mineable with commodity hardware.

Bitcoin ASICs are as yet useless for anything other than mining bitcoin — if they were more useful, producing them would be less efficient and more expensive. This, combined with heavy competition, creates waste when ASICs are deprecated for newer, better hardware.

Litecoin mining farms, from Quartz’s feature on industrial mining.

How do ASICs actually work? 🤓

[Warning: Nerd Alert] ASICs achieve efficiency through “pipelining” — by placing chips with essential bits physically closer to each other, they’re able to eke out parallelization gains.

What is “ASIC resistance”? 🚫

While any PoW-based coin that’s sufficiently large, some are more “resistant” to ASICs than others by design of their hashing algorithms. In general, if your form of mining is very memory-intensive, ASIC design tends to be much more infeasible.

Skeptics of ASIC resistance will claim that any sufficiently large mining reward will incentivize the creation of ASICs, at which point the network will need to “fork” (update the codebase to obsolete the specialized hardware developed). Who decides what algorithm to fork to? Is that centralized?

What are some ASIC alternatives? 📶

Other than “resistant” hashing algorithms, there are also projects looking at alternatives to Proof of Work. There are some alt-coins that believe by having multiple PoW algorithms, it makes them more ASIC resistant.

From the Ethereum white paper, on ASICs:

The current intent at Ethereum is to use a mining algorithm where miners are required to fetch random data from the state, compute some randomly selected transactions from the last N blocks in the blockchain, and return the hash of the result. This has two important benefits. First, Ethereum contracts can include any kind of computation, so an Ethereum ASIC would essentially be an ASIC for general computation — ie. a better CPU.

From Monero’s StackExchange site, on ASIC resistance and why Monero’s PoW and the associated memory requirement prevents the development of ASICs:

Cryptonight requires 2 MB of fast memory to work. This means that parallelizing hashes is limited by how much memory can be crammed in a chip while keeping cheap enough to be worth it. 2 MB of memory takes a lot more silicon than the SHA256 circuitry, for instance.

Vertcoin is a project that promotes “ASIC resistance” as a feature and have proposed enforcement from the community (e.g. hard forking the Proof of Work to render ASICs useless). From their project site:

Mining should be fair & easy! We do not allow ASICs on our network. The development team is wholly committed to keeping it that way. This ensures the long term security and fair distribution of the currency.

Other tokens, such as Bitcoin Gold, offer similar guarantees to retain GPU efficacy in Bitcoin mining.

Where can I learn more? 🙋‍

The story of Sia’s recent run-in with an ASIC maker provides a fantastic case-study of what happens when community principles conflict with crypto-based businesses.

“Bitcoin and The Age of Bespoke Silicon” by Michael Bedford Taylor at UCSD. A great academic overview of early Bitcoin ASIC production.

The Bitcoin Wiki’s overview of large ASIC producers.

If you enjoyed this post, please 👏 clap 👏 until your fingers hurt and tweet at us to stay updated on future writings (or holler if you have any questions).

--

--

Arjun Balaji
Crypto Simplified

professional risk taker. enjoy: markets, watches, meta-twitter, and documentaries. open-sourcing my thinking