Durable, Low Energy Bitcoin Algorithm Proposal | Hacker News Reader

Durable, Low Energy Bitcoin Algorithm Proposal

github.com

33 pointsneopba12 comments

Until photonic processors become cheap, and then it’s a keccak-256 ASIC. Silicon wafers at 90nm quality are already quite cheap, but actually, a photonic waveguide can be made out of doped glass or grown on any substrate. Also, the 3D integration possibilities are nearly endless. If there was a real economic demand for these processors, the cost would go to nearly zero.

If you can figure out an actual hash with just the photonic part, then you have a strong case.

You should take a look at companies like https://lightmatter.co/

Photonic chips for analog matrix processing are already being commercialized.

The underlying platform, CMOS Silicon Photonics is commercial already (in every data center) and produced in high volume.

Heavyhash can be constructed such that the Keccak hashes comprise a minimal amount of the work.

Quoting from the BIP:

"Below is a conceptual representation of a 3D-packaged oPoW mining chip. Note that the majority of the real estate and cost comes from the photonic die and the laser, with only a small digital SHA3 die needed (as opposed to a conventional miner of the same cost, which would have many copies of this die running in parallel)."

Firstly I'm not convinced that the photonic version would be more efficient than the all-electric version. The conversion is very expensive. Secondly, they don't need to be on the same chip. The Sha3 chip would be very dense, and the multiplex in time and frequency into the photonic circuit, which would have much larger features, probably printed on silica.

Are you familiar with silicon photonics? (though indeed they probably would not be on the same chip as we point out in the BIP)

We have a short primer on Si photonics in our paper actually https://assets.pubpub.org/xi9h9rps/01581688887859.pdf

This proposal, if implemented, puts Bitcoin on an equal footing with Chia(also capex bottlenecked). It's probably the "next best thing" for an efficient securing function without going towards staking.

Chia has the nice property that it leverages the existing 1.2 zettabyte per year storage market, though. Meaning that no specialized hardware for mining is needed — existing storage in the world can be used to mine, lessening the environmental impact of buying new drives. And if miners no longer want to use their storage space for Chia they can sell it to anyone who might want the storage space for whatever purpose, not just Chia mining.

It also means that any innovations in storage media that might happen to optimize Chia earnings would also very likely benefit the whole world since it would mean cheaper ways to store lots of data.

But immediately i have seen companies announce specialized PCIe 4 SSDs - in the beginning you could also just mine bitcoin with existing CPUs or GPUs.

SSDs are not required for farming Chia or even really advantageous at this point. Using them to write the plot data is faster than HDDs certainly, but that only happens once per plot and will happen a lot less as the network growth slows down. The main costs of Chia in the long-term are using the cheapest, most dense storage media to farm with minimal electricity.

We'll certainly see empirically what the long-term outcome of Chia is given that the network is at 7 exabytes and showing no signs of slowing down.

Chia is great (and the one real alternative to oPoW for low energy consumption that I am aware of), however the security assumptions are very different from PoW (it uses VDFs etc).

Bitcoin is conservative by nature, so the likelihood of adoption for a PoW with the same fundamental game theory and security assumptions is much higher.

Regarding targeting commodity hardware:

This has a good ring to it but introduces some 51% attack vectors. A lot of the security in BTC comes from the fact that miners will make their hardware worthless if they collude to double spend in a 51% attack. That's not true if the hardware is a real commodity and has resale/reuse value outside the system. Probably there will be specialized hw for Chia though as people often point out.

I don't think that specialized hardware is a security feature of Bitcoin, and definitely not one that was addressed or foreseen in the Bitcoin whitepaper (one IP one vote or one CPU one vote).

In fact, it's quite easy to argue the opposite, that commodity hardware makes the decentralization/security of the network higher because anyone with extra space can farm. This is true today since Chia already has more full nodes than Bitcoin — you can run a full node off a Raspberry Pi :)

So instead of wasting energy we waste resources building HDDs. And the market will be flooded with worn out HDDs. And instead of GPU prices it's HDD prices. Little gained.

Maybe! One of the bets the Chia team is making is that there's a lot of over-provisioned storage already out in the market that organizations and individuals will start to turn toward farming. This is obviously a lot more profitable than buying new drives just to farm since they're existing assets. And if in the long-run it continues to be profitable to farm Chia by buying new drives then it will likely help lower storage prices for the entire market.

Reply on news.ycombinator.com