Very nice read https://download.wpsoftware.net/bitcoin/asic-faq.pdf
It’s a interesting point of view, but you also have to consider his association with Block Stream and Bitcoin Core…underlying bias in regards to are ASICs good/bad and are they inevitable. Not saying he’s wrong or right, just that you can’t call it subjective when you are involved with coin #1 which is pretty much owned by the number one ASIC maker. Give me a well written paper from someone completely divorced from ASIC and CPU/GPU on the subject…that would be one that carries some weight. A purely academic with no ties to Bitcoin, Block Stream, Bitmain, Nvidia or AMD…if such a person exists (no idea).
Yes I am aware of who he is and I strongly disagree in many ideas he has in other areas. Nonetheless the ideas presented are worth reading.
Not disagreeing, just wanted to point out potential biases for folks.
This paper is deeply flawed, and Poelstra should retract the paper. The premise of this paper is that POW problems such as SHA-256 will tend towards the thermodynamic limit known as Landauer’s principle. Unfortunately, Poelstra has failed to realize that cryptocurrency mining is not bound my Landauer’s limit. Let me explain.
Landauer’s principle states that every bit of information deleted costs kTln(2) energy. Computers are bound by Landauer’s principle since every single AND and OR gate deletes energy since the AND and OR gates each have two inputs but one output. However, there is a type of computation called reversible computation which should come out in the near future where one gets by the thermodynamic restriction from Landauer’s principle by not deleting any information. In particular, with reversible computation, one uses logic gates which do not delete any information such as NOT, CNOT, and Toffoli gates.
Poelstra makes the common yet flawed assumption that claims that since SHA-256 is an irreversible algorithm, reversible computers will not be able to mine Bitcoin. However, it is well known that any algorithm that runs in space S and time T can run on a completely reversible device using time O(T(T/S)^(epsilon)) and space O(S(1+Log(T/S))) ,. The space/time overhead incurred from partial reversibility is even milder. For Bitcoin mining the space/time overhead incurred from completely reversible computation should be within one order of magnitude and much less with partial reversibility since there are reversible adders  which can compute the modular addition used in SHA-256.
This means that the energy usage in Bitcoin mining will not approach Landauer’s limit since it will eventually fall under it. On the other hand, if we stick to conventional irreversible Bitcoin mining rigs, then the energy efficiency of these Bitcoin miners will not hit Landauer’s limit because it will have to hover above 20kT in order to overcome thermal noise, and this 20kT must be spent for ALL conventional Bitcoin mining rigs regardless of what types of materials the conventional Bitcoin mining devices are using.
Now, energy efficient reversible computers currently do not exist in the free market BUT researchers are now starting to take reversible computing seriously because the energy efficiency of conventional computation is approaching the point where overcoming thermal noise is a serious problem. Furthermore, these are several different kinds of prototypes for reversible computers. We should therefore expect to see the first free market reversible computers arise pretty soon.
-Joseph Van Name Ph.D.