An anonymous individual going by the name of Ikmyeong Na has stated Nvidia gear would perform better under ProgPoW, but that’s not the intended design. Speaking in broken English as he claims to be South Korean, Ikmyeong said:
“It looks like cause AMD gets a lower hash than Nvidia, but it doesn’t mean the algo is for Nvidia. Think of why AMD gets more hashrate than CPU. AMD still works fine however due to GPU core usage it don’t get any benefit from ProgPoW mine like we did for Ethash.”
Asked: “So it is true that ProgPoW favors Nvidia?” – Ikmyeong says: “No Nvidia is being too good for ProgPoW. That’s all. And if AMD releases their card with better GPU core usage it will overcome Nvidia cards. Meaning there is no reason to benefit specific brand of cards, we don’t optimise our algoritms to be better for specific GPU brands. But GPU manufactures should do the optimization for algorithms if they want more sales on their card.”
In other words, the algorithm just happens to favor Nvidia gear, but it isn’t designed to do so. Moreover, Ikmyeong is suggesting that AMD’s gear currently has an advantage. That presumably means miners have bought more AMD gpus.
If the algo switches, however, those current AMD miners will presumably be at a disadvantage. That might mean Nvidia basically gains a manufacturing monopoly over eth mining, at least to begin with.
AMD could compete, presumably, but it does appear the proposed algo wouldn’t just affect asics but would also reorganize current gpu miners, some of whom have been making quite a lot of noise in opposing ProgPoW.
Ikmyeong, a persona who seems to have appeared in October when he started contributing on Github, would not reveal his real name or position after repeated questioning. He says:
“I developed all the ProgPoW related things, miner, pool , all I need for our miners… and that’s how we sustain our testnet Gangnam and prepare for the mainnet. I also am charge of these developments, QA, bug fix, etc…
I get help from several developers including ifdefelse, ethminer maintainers, community, etc.”
The proposed algorithmic change has faced some resistance from other miners and developers. A new start-up that is developing an eth asics, Linzhi, says mining dominance can be achieved through a difference in the cost of production and the resale cost or the mining gain for the gear.
They say Nvidia and AMD currently do not mine themselves, but Linzhi claims they are open to working with mining farms directly. The latter would gain a cheaper price for the gpus than the public, thus giving them a competitive advantage.
The same, of course, can be said of Linzhi itself as they would be producing the asics at a far lower cost than they’d be selling them. They may also mine with those asics, so giving them a competitive advantage.
That’s in fact the primary argument for an algo change. Asics manufacturers would be able to dominate with the first to market having an accruing advantage until they miscalculate and go bankrupt as was the case for Bitfury and as may now be the case for Bitmain.
The difficulty there lays in the fact that there is no known algorithm for which asics can’t be designed. They can at best be delayed, but the problem there is whether such delaying wouldn’t be picking one dominant miner over another.
The ProgPoW team has been working on the algo since May last year. They won’t reveal who exactly funded this work, but the only known person from that team, Kristy-Leigh Minehan, works at Core Scientific. That being a vast mining farm with its own hydraulic power.
Minehan says Core Scientific has nothing to do with ProgPoW, but there are suggestions she previously developed eth FPGAs for only Nvidia gear.
Some suggest there’s a relationship between Nvidia and Minehan, which she denies. It now looks likely, however, that the algo favors Nvidia, at least based on the current gear out there.
An audit and an independent report, therefore, would probably be necessary for such algo change to be done in an objective manner. Otherwise the main point of contention might be that the ecosystem is picking winners and losers by interfering in the free market in a way that may increase incentives for others to demand a certain change or another.
That same argument, however, can be made if no action is taken, but there everyone had the opportunity of first advantage. While here the ecosystem might be giving it to someone, perhaps unfairly.
In an ideal world, we wouldn’t have needed to bother with any of this as we would have had Proof of Stake (PoS). That however has been delayed and suggestions are now that the Proof of Work (PoW) chain will keep running in parallel for a very long time.
What is meant by a long time isn’t clear, but first there will be the Beacon Chain. That’s a fully PoS new blockchain that links to the current PoW chain.
Initially the Beacon Chain won’t have any functionality but staking. That will provide finality in the PoW chain, so miners’ rewards there will decrease considerably to perhaps 0.6 eth per block.
Then the first phase of sharding will launch perhaps next year. State will be added by perhaps 2021. Then cross-shards and so on.
During all that time, the PoW chain will keep running. There will be a transition as smart contracts and transactions gradually move to the Beacon Chain. When that’s done to a reasonable extent, the PoW chain can effectively be folded into a smart contract and deprecated.
We’d probably therefore be looking at five years optimistically. So if the algo is changed, it wouldn’t be buying some time with PoS around the corner. Instead, if the ecosystem favors a change to the algo, there should be some sort of strategy regarding what happens when new asics appear for the new algo.
The ecosystem can’t really bind itself, so no one can really say we’ll just fork again. We might be far too busy with other things. Secret asics may have gained 90% of the hash, making a 51% attack a certainty, and so on.
Making all this quite a complicated matter as the mining algorithm lays at the foundations of ethereum’s security. Vitalik Buterin, however, has come up with a method of CBCifying the beacon chain. He says:
“Switching FFG to CBC is actually quite a small change to the spec; the purpose of this post is in large part to illustrate concretely what changes would be required.
Most of the hard things (all the non-consensus-related spec details, LMD GHOST implementation, signatures…) are already part of the FFG spec…
I actually think it would be nice if one of the major client developers could have a side project of implementing CBC and seeing what it looks like from an efficiency point of view.”
Meaning that all this ProgPoW debate might not quite matter as full PoS may come sooner than expected, but soonish in this space is a meme for a reason.