Bitcoin Gold, the Latest Bitcoin Fork, Explained (arstechnica.com)

Timothy B. Lee via Ars Technica explains Bitcoin Gold: A new cryptocurrency called Bitcoin Gold is now live on the Internet. It aims to correct what its backers see as a serious flaw in the design of the original Bitcoin. There are hundreds of cryptocurrencies on the Internet, and many of them are derived from Bitcoin in one way or another. But Bitcoin Gold -- like Bitcoin Cash, another Bitcoin spinoff that was created in August -- is different in two important ways. Bitcoin Gold is branding itself as a version of Bitcoin rather than merely new platforms derived from Bitcoin's source code. It has also chosen to retain Bitcoin's transaction history, which means that, if you owned bitcoins before the fork, you now own an equal amount of "gold" bitcoins. While Bitcoin Cash was designed to resolve Bitcoin's capacity crunch with larger blocks, Bitcoin Gold aims to tackle another of Bitcoin's perceived flaws: the increasing centralization of the mining industry that verifies and secures Bitcoin transactions.

The original vision for Bitcoin was that anyone would be able to participate in Bitcoin mining with their personal PCs, earning a bit of extra cash as they helped to support the network. But as Bitcoin became more valuable, people discovered that Bitcoin mining could be done much more efficiently with custom-built application-specific integrated circuits (ASICs). As a result, Bitcoin mining became a specialized and highly concentrated industry. The leading companies in this new industry wield a disproportionate amount of power over the Bitcoin network. Bitcoin Gold aims to dethrone these mining companies by introducing an alternative mining algorithm that's much less susceptible to ASIC-based optimization. In theory, that will allow ordinary Bitcoin Gold users to earn extra cash with their spare computing cycles, just as people could do in the early days of Bitcoin.

Comments

8.5/10

[PopeRatzo]

I'm holding out for the Bitcoin Ultimate GOTY Edition with all the DLC.

Oh God, this again?

[Orgasmatron]

...but it uses an alternative proof-of-work algorithm called Equihash that supporters believe is impervious to being sped up with custom hardware

I laughed and laughed and laughed.

From the paper (PDF):

a reference implementation of a proof-of-work requiring 700 MB of RAM runs in 15 seconds on a 2.1 GHz CPU, increases the computations by the factor of 1000 if memory is halved, and presents a proof of just 120 bytes long.

Hmm... Needs less than 1 gigabyte. For external chips, that costs, let me see, $16 in modest quantities. Want it cheaper? Here is a magazine article from 4 years ago about people embedding memory in ASIC dies. In a modern chip process, 700 MB fits into what, a 5mm square?

They wave their hand over it in the paper, so it isn't like they ignored the cheapness-of-memory problem entirely. What if we assume that they are right and they have indeed found a problem with a critical dependency on memory throughput. Is there an obvious solution to that problem? What is the fastest memory in the world? (What word is entirely missing from their paper?) SRAM. In-die SRAM can be absurdly fast, like full core speed for arbitrary values of "core speed" and no wait cycles. It makes no sense to load a CPU up with piles of the stuff because caching has diminishing returns. But, what if your goal was to use ~6 billion cells of SRAM not as a cache, but as your main memory... How much faster would that ASIC be than a general purpose CPU?

Bottom line, if you imagine that you have a computation problem that can't be solved by building a single-purpose chip, you are almost certainly wrong. Either you don't understand your problem, or you don't understand the array of solutions available to solvers.