The centralisation of hash power makes it much easier for them to attack the chain. It also makes it easier to use coercion to attack the chain since there are fewer people to coerce.
I basically agree with this framing, I would sharpen it with the following distinction:
Centralization of hash power makes it easier for existing miners or those who can influence existing miners to attack the chain if they want to. But if the chain is secured by ASICs then it makes existing miners want to attack the chain less, because by doing so they lose more.
However if the attack is coming from someone other than the existing miners (or from an entity that can control the existing miners), then this attack will be cheaper for GPU mined coins vs. ASIC mined coins.
So people who are more worried about state attacks or other big external attackers should definitely prefer ASIC mining. People who are more worried about a coin’s own miners “going rogue” should on the surface prefer GPU mining.
I say “on the surface” because the thing about ASIC mining is that the community can switch from ASICs → GPUs if ASIC miners ever start acting up. But switching from GPUs → ASICs if a government starts attacking you is much tougher.
So the downside of guessing wrong about which attack you should be worrying about is very different. ASIC mining gives you the option to defend at least pretty well against either attack.
This is the exact opposite conclusion to what I argued. Given we agree ASICs lead to centralisation of hash power then it is far easier for an external actor to compromise by coercing the miners because there are fewer of them. Hence, the centralisation of hash power that ASICs lead to is a problem we want to avoid.
It is secured by ASICs. Everything is an ASIC. Move forward with that, it doesn’t matter if it is ASIC or GPU or even CPU, in this statement because no one would want to attack the network.
Remember, that ProgPoW doesn’t get rid of ASICs, it removes the economic incentive to build an ASIC by making GPUs the most efficient ASIC out there.
No, you can’t just go back, without changing the PoW and that would be contentious.
I and others have provided you with an endless stream of concrete in multiple forums and you keep just claiming none of it is valid. It makes it very hard to have a conversation with you.
The cost to 51% attack is the cost of purchasing the majority of the hashrate. Electricity costs being the majority of what PoW actually “costs” means that for the sake of argument, the value of the devices can largely be ignored. This is reinforced by the fact that markets largely seem not to care about the effects (e.g. valuation) of such an attack, so I would argue yes indeed, it does not matter if you’re renting or buying.
If we’re talking about renting, it is largely impossible to rent anywhere close to the amount of capacity required to perform this attack on the highest difficulty chain that makes use of any given algorithm/device. Since we are the highest difficulty chain for GPU devices period, that means your concerns are simply not valid. You could say “wait, what if another higher market cap chain came along for GPUs?”, I would respond that this is all the more incentive for developers to execute and ship Eth1.x and Eth2.0 roadmaps.
Any line of questioning about how much easier it is to get one or the other, that’s my line of reasoning. The likelihood of coordinating an attack or disruption is what matters.
Bitmain forked Bcash from Bitcoin. I think we all agree on this. They tried to use their power to force the narrative of Bcash into the Bitcoin community. They also largely failed (because miners are not the network, full nodes run by the community are). But to say this did not cause a significant disruption would be will fully ignorant. This is what we can expect from a centralized entity: if they have a reason to attack or disrupt a chain, they are very well-organized to do so.
GPU mining farms are very disparate and georgraphically diverse. Because GPU mining is permissionless, it’s very hard to predict exactly where they’ll be, although places with cheaper power tend to win the economies of scale. To amass 51% of the hashrate, you would need to talk to something like 50-100 separate farms. That’s a lot of coordination, and attack would be seen coming from a mile away. (Hell Bcash fork was known for a month). Giving time to the community to prepare is what counts and allows us to survive such an attack through a social majority running nodes.
Now you might say “Hey, what about mining pools? Those are centralized!” Mining pools have some of the lowest overhead costs of any business I have every seen. They take in all your mining rewards, hold them until you’ve reached some payout threshold, and then redistribute after taking a couple percent cut for the cost of their “coordination” (really their social reputation among the mining community). If they aren’t working on what’s involved to do a PoS Validator pool, they’re really missing out, because that’s largely the same business.
All of that is to say, mining pools have very little incentive to disrupt the chain moving to PoS because they have very little overhead costs invested besides running a node with very high uptime, and some sort of automated database to keep track of who contributed what. That incentive might be slightly higher for Hybrid PoS/PoW but… Basically the same argument works there.
I refuse to engage with the “well funded state actor” arguments because their involvement would entirely change the game, and unlike Bitcoin, our “target” is traditional tech companies and banks, not sovereign government’s most valuable tool (fiat currencies).
That choice of language puts you into alignment with the censorious and bitterly hostile to Ethereum Bitcoin Core community.
Bitmain did not create Bitcoin Cash. Bitcoin Cash was created when large blockers decided they needed an insurance policy in case the S2F agreement unraveled. Which, of course, it did, because Bitcoin Core never had any intention of allowing any other groups to share their power to control Bitcoin’s future. They reached out to Jihan Wu for help, because he is also a large-block proponent and has significant hash power.
The failure of Bitcoin Cash in the market, which you correctly noted, is the death-knell of the relevancy of Bitcoin to the future of crypto.
I don’t think it’s fair to lay the Bitcoin civil war at the feet of Bitmain and Jihan Wu.
I may have vastly oversimplified that discussion (and Bcash is just a shorthand definition to me, I don’t really care about it either way). I am using it as a historical example, I have no intention of drawing boundary lines with Bitcoin because they aim to do largely different things than us. If I mention Bitcoin at all, it’s as a rich source of historical precedent to draw from, and I am making no value judgements on the community itself. I think trying to make such value judgements of the Bitcoin community is largely a distraction to this discussion.
This was my main point. The fact that he could be reached to organize a support for the new chain, and for that to have largely worked, is entirely my point. That would not be nearly as easy to do with our current GPU mining community.
He is definitely not fully to blame, but played a large part in helping it succeed, and it continues being a distraction to the Bitcoin community to the present day. We should avoid creating central power figures like this.
The number of ASIC manufacturers will always be strictly less than the number of mining farms due to the economies of scale at play.
This is a widespread misconception. Mining pool it’s a for-profit business at least for pools with top-10 hashrate. Why should mining pool attack chain and ruin own business?
You can’t slap shit together and deploy new pool and gain top hashrate overnight. Development, cost of running, marketing costs - I would say it’s more than expected profit from 51% attack.
Top-10 pools are usually multi-coins pools, with more than one coin available for mining. Attacking just one of the coins would cost mining pool a reputation.
My argument was a little sloppy. They’ve built a business overtime based on reputation, marketing, and running some solid infrastructure. I meant their capital cost overhead of investing in hardware is really low. They don’t have a strong incentive to combat a change to the mining algorithm or consensus algorithm because of this. The fact that the do multiple coins just makes this even more true. They’re truly just service providers.
This is the exact opposite conclusion to what I argued. Given we agree ASICs lead to centralisation of hash power then it is far easier for an external actor to compromise by coercing the miners because there are fewer of them.
I specifically said my statement applied “if the attack is coming from someone other than existing miners (or an entity that can control the existing miners).” Your argument was about the situation where the existing miners can be controlled by an entity who is behind the attack. That is not the only situation to be concerned about.
Electricity costs being the majority of what PoW actually “costs” means that for the sake of argument, the value of the devices can largely be ignored.
I strongly disagree that rounding hardware costs to $0 can lead to a remotely plausible analysis of a build/buy attack.
Even if the majority of the cost of hashing is electricity in some cases, this is only true when the hardware cost is amortized over an extremely long period of time.
I’d really like to see you to provide some detailed back of the envelope calculations to support your claims.
Here’s my back of envelope calculations:
I searched around to find a tier 1 GPU for ethash, and it looks like the NVIDIA GTX 1660 Ti 6GB is pretty good. 29.5 MH/s at 90w. If you use the default electricity cost of $0.12 per KWh on cryptocompare you’ll see that it will take about 3 years until the electricity cost overtakes the cost of this thing on Amazon. Let’s say you’re a big GPU farm and you can get a good deal on these cards. It’ll still be years before the electricity price is higher than hardware price. And this is using an electricity price that is higher than what I just happen to get in Seattle, so I’m sure pro miners are paying less for electricity.
If the ETH hashrate is 150,000 GH/s you need about ~5 million of these cards to overpower the existing network. Let’s say you get a good deal and can get them for $200 each. That’s still a billion dollars for only the cards. Even assuming that your facilities/land/cooling costs are zero, this will dwarf your electricity costs even if you’re attacking Ethereum for many months continuously.
Suppose your electricity costs are what crypto51.app reports for the ETH hash rate rental cost. Then we scale the reported hashrate to 150 TH/s which we used above, and round up to $100k per hour. So $2.4 million per day, $72 million per month. You could 51% attack ETH for 13.8 months continuously before your electricity cost would overtake your hardware cost. And this is with very generous hardware/infrastructure cost assumptions, and generous assumptions about the existing miners continuing to mine for zero revenue for those 13.8 months (otherwise the attacker could scale back their electricity expenditure).
I’m very curious to what extent others in the Ethereum community believe that Ethereum should not have a goal of being resistant to state-level attackers (aka the goal should not be be “sovereign grade censorship resistance”).
This is largely considered not possible if you are the largest destination of hashrate for whatever algorithm/hardware you optimize for. So we just have to continue being that!
Most of the GPU hash is in the hands of big farms not independent small minners. It is foolish to remove the smaller hash power of ASICs just to give it to GPU farms. This is centralization and will in essence cause an easier resistance to the transition to Eth2
