After we detailed accusations of corporate favoritism by a team behind a proposed ethereum Proof of Work (PoW) change, the semi-anonymous ProgPow team has now effectively admitted they’re working with both Nvidia and AMD. Someone called “IfDefElse” says:
“We were lucky enough to have an email review that included engineers from the Ethereum Foundation, Ethereum Core Devs, Nvidia and AMD. The Nvidia and AMD engineers gave the algorithm a generally positive review.”
They do not state what the feedback was from the Ethereum Foundation or “core devs.” Nor do they say who those core devs were. Of AMD, they say:
“AMD’s response addressed two major concerns:
- If the Ethash PoW algorithm was replaced with ProgPoW, wouldn’t ASIC manufacturers be able to quickly look at the open source code, and create a specialized ASIC to mimic it?
- Would ProgPoW make it more difficult for GPU miners to mine on Ethereum?
An AMD engineer stated that, yes, in theory you could create an ASIC for ProgPoW, but they would need GPU-specific knowledge, specifically around memory controllers.”
A miner called Linzhi claims they “can accelerate ProgPOW by 3x to 8x.” That’s disputed by Kristy-Leigh Minehan, the only known person from the ProgPoW team. While AMD is stating you can develop a ProgPoW asics, but you’d need specialized GPU knowledge. Alexander Levin Jr, president of gpuShack and founder of ethosdistro, says:
“In the past, using algorithms specifically for ‘ASIC-resistance’ has created unfair competitive advantages for first-movers because it increases the requirement for Research&Development. In the case of ProgPoW, the author is the first mover.”
ProgPow has been in development since circa the second half of 2018. It appears this was done in collaboration with Nvidia and AMD as far as feedback is concerned.
Soon after a very small group of eth devs tentatively gave the go ahead to ProgPoW, Minehan’s employer Core Scientific announced they have joined the NVIDIA Inception program.
Levin claims Minehan has a relationship with Nvidia. “This means that NVIDIA may be privy to future optimizations sooner, and more often,” he says. Of Nvidia’s feedback, the ProgPoW team says:
“The Nvidia engineer generally agreed with our approach. They said that the algorithm ‘fills in the holes’ between memory accesses with compute instead of the ‘GPU sitting idle as a glorified memory controller’.”
The ProgPoW team says the current eth asics are just 2x more efficient than GPUs, meaning you need two GPUs for one asics. The very low efficiency gain compares to GPUs effectively being unusable in bitcoin, while in eth, gpu industrial miners remain the dominant method for eth mining.
There are apparently new asics in development by Linzhi for the current ethereum PoW algo, ethhash. That too will provide a small efficiency gain of 2x or a maximum of 4x according to the ProgPoW team.
If the algorithm is changed, with the process taking five months under optimistic estimates, the main question is whether there won’t be a ProgPoW asics to await its deployment. So making the exercise completely futile.
If, as AMD says, you can develop an asics and at the same time GPU mining is more difficult in eth, then it isn’t very clear whether this change is actually against asics or in favor of them.
“[ETHlargement] is some kind of NVIDIA GPU tweak that allows for +60% hashrate boost on some NVIDIA models, developed by the same people behind ProgPoW. Now, we have ETHlargement, and both, GPUShackle and Linzhi claiming that ProgPoW is funded by NVIDIA, isn’t that worth at least an investigation,” Afri Schoedon of Parity says before suggesting there could be other algorithms deployed.
The question of asics in general hasn’t had much debate in ethereum. They say the original intention was for eth mining to be asics resistant, but there is no known algorithm for which asics can not be developed as far as we are aware.
The only method to really be asics resistant, therefore, would be to change the algo very often. That’s Monero’s approach, but they were 51% attacked after they changed Proof of Work.
Another approach taken by some coins is to mix different algos, but that complexity can create known and unknown vulnerability vectors.
Conceptually, one can ask why should AMD or Nvidia be favored? GPUs are quite expensive, so instead of buying that, a miner can buy asics.
At the beginning, a gamer with a laptop could mine eth, so increasing its distribution and thus decentralization. Eth mining now, however, is concentrated in huge industrial GPU mining farms. The laptop gamer has no chance of getting anywhere near a fraction of an eth.
Except for miners, why should anyone care whether those huge farms are GPU based or asics? As it concerns small or hobby miners, they now need a basement with GPU rigs. Well, they can have asics in the basement instead, why does it matter?
As stated it did matter initially as GPUs are widespread, but now at the industrial scale what might matter is who produces the hardware. Two mega corps, or some small start-up like Linzhi, or some bigger once start-up Bitmain, or Innosilicon, or whoever wants presumably.
The other point of consideration is the game theoretical aspect of asics where it comes to 51% attacks. Unlike General Processing Units (GPUs), asics are Application Specific Integrated Circuits. That means they’re good for only one thing, mining eth and nothing else.
As such, if an eth miner misbehaves and actually engages in a 51% attack, you just brick all the asics by changing PoW. The threat by itself should be sufficient to ensure it doesn’t happen.
Honest miners in this case would also be affected, but if there is a 51% attack, then they clearly have failed at their job of securing the network. So they should take the punishment.
The other far more complex aspect of asics is this revolving cycle of miners. Mining is a brutal industry where small mis-calculations can mean bankruptcy. In bitcoin, thus, once king miners tend to eventually vanish.
Whether that’s a good thing or a bad thing is a debatable point. The proposed algo change in ethereum, however, is basically an attempt to keep current miners.
We’re not sure what current miners have done for ethereum despite being rewarded by the billions. They do not fund any eth development as far as we are aware or any other aspect of the ecosystem. They do not even optimize their own gear to reduce uncles, eth devs had to do that. Why then should ethereans intervene to favor them instead of seeing whether other miners might have learned their lesson.
Finally, with GPUs being so widespread, wouldn’t an adversary find it far easier to attack ethereum as it stands rather than having to first engage in a massive operation of covertly manufacturing at least as many eth asics as there are in existence?
Now the above might sound like we’re against an algo change or in favor of asics. We’re in the business of providing information, not mining, so we don’t care one bit. We’re just raising points to consider before taking what could be pretty big steps.
Whether there should or should not be an algo change is obviously up to the entire ecosystem. If this is to really be proposed there should at the minimum be a token holders vote. Then we’ll take the side of whatever is the outcome of such vote.
Such decisions obviously can’t be made by a tiny number of devs prior to public discussions of all the foreseeable considerations as this is not really a technical question but more of a conceptual question of how do you maintain decentralization and security when mining reaches the stage of industrial farms.
The medium to long term answer in eth is by switching to Proof of Stake (PoS). Would asics before hand make that more difficult? Or the opposite, would current miners showing they have so much sway indicate that they are and will do whatever they can to delay PoS. Including perhaps by what may be an algo change distraction which may delay much else.
We haven’t put many arguments in favor of GPUs and that’s primarily because we haven’t heard many of them. The rhetoric so far has been GPUs because asics bad as they lead to centralization. What’s the difference between a GPU mining farm and an asics mining farm, however, has never been explained.
Finally, just to make this article even more one sided to court arguments in favor of maintaining GPU mining, where it concerns the move to PoS, it is probable that the ecosystem can stand up far more easily to some start-up than the two corporate giants.
The argument there would be easy: Bitmain bad (whether rightly or wrongly such argument is made). If – and we’re not saying they are – but if Nvidia started some shill army campaign near the time, then dismissing such suggestions would be as easy as shouting conspiracy theory.
Ethereum therefore should carefully consider the many complexities of an algorithmic change, including the potential loss of its primary defense against a 51% attack: bricking asics.