Ethereum DAG Size and Hash Rate Impact

Background

At block 9,840,000 (~April 9th) the popular Antminer E3 ASIC will be bricked due to a growing DAG size on Ethereum. A recent study by Kristy-Leigh Minehan estimates that up to 40% of Ethereum’s hash rate may come from mining equipment that will be bricked due to this issue. There is likely some variability in this number and we won’t know until the event occurs but for the sake of this article let’s go with it.

What will happen?

There are a few important things to understand before analyzing the impact of hash rate drop and I suggest people go read this EthHub page on Ethereum mining. The most essential is Ethereum has a difficulty measurement which determines how hard it is to mine a block on the network. As hash rises and falls, this adjustment goes with it and attempts to keep block times in a certain range.

In the scenario of 40% of hash rate dropping off the network, the difficulty adjustment will start to drop to get block times back towards the 13s we are used to. This will take a little bit of time and while I’m not exactly sure how long, let’s go with 30 minutes.

However, as ASIC miners drop off the network, GPU miners that previously were not profitable, will start to come onto the network as well. It’s hard to estimate how much of the lost hash will be filled but it will be a decent amount.

So to summarize the likely scenario:

1. Hash rate drops from 181.8 TH/s to 109.08 TH/s
2. Difficulty starts to adjust down
3. More GPU miners start to come online
4. Difficulty goes back up
5. Block times go back to ~13s

I think it’s important to note that this could all occur basically in the same instant. There is nothing stopping GPU miners from coming in right at block 9,840,000.

Is the network at risk?

Concern 1: Immediate drop in hash
One concern of an immediate drop in hash rate would be a malicious actor coming in for the small window (30 minutes?) before difficulty fully adjusts and attacking the network. The likely easiest path for this to occur would be renting hash rate. I won’t act to be an expert in renting hash rate but there is a handy website which shows us the current cost to 51% attack any network.

It currently costs $110,000 to attack the Ethereum network for 24 hours. If we assume a worst case scenario of: 40% hash drop, no immediate GPU miners filling in and an hour to adjust properly, then the cost to attack would fall to $66,000 for that hour.

For context, it would have cost ~$80,000 to attack the network in early 2019 when hash rate was lower.

A very important note here is that cost above assumes you can actually acquire the hash rate. Currently, it’s only possible to rent 4% of the entire network’s hash rate on nicehash.

Concern 2: Will other ASICs fill the gap
Another concern I’ve seen raised is will the bricking of the E3’s result in better/more efficient ASICs being created?

To me, this is a hard risk to quantify and understand.

Has this occurred before?

Ethereum has gone through large drops in hash rate in the past, although not at an immediate moment in time.

Should we do anything?

I’ve tried to keep the above unbiased and just a collection of different arguments I’ve seen. The following is now my opinion.

1. The fact that it’s an immediate drop is irrelevant to me. Actually, I find a slow, sustained drop to be more concerning. The reason is that it gives malicious actors time to acquire the hash over weeks of time and coordinate the attack. This will be a very quick window and while it’s known, I’m not sure what would motivate someone more to do it in that moment versus over the past 18 months when hash started to drop off the peak and rewards were cut. Even in the worst case of no other miners immediately filling in, we are talking just a $12,000 difference versus early 2019. To me, if someone wants to 51% attack Ethereum, this won’t be the difference.
2. I’m not sure where this hash rate would come from. One of the reasons we don’t see many 51% attacks is simple game theory. Does it make sense to “burn down” your investment (ETH) to attack a chain for an hour? So far the answers appears to be no. We’ve also just gone through lengthy ProgPoW debates where one side has consistently said that GPU miners would not attack the Ethereum chain. With the hash rate not being rentable and ASICs being bricked off the chain, where exactly is this attack hash coming from?
3. Some have the concern that better ASICs will now be made and come online. I fail to see how this relates to the DAG issue specifically. If ASIC manufactures were able to create a better chip, they already would have and they’d be pushing out the E3s. There would be no reason for them to wait until the DAG issue kicks in. I’m not saying that better ASICs cannot occur, I just don’t think it’s related to this specific issue and I’ve made clear in the past that I’m relatively indifferent on where Ethereum’s hash rate comes from. This point here is just rehashing the entire ProgPoW debate, not the impending hash rate drop.

In conclusion, this seems like a scenario where we let the current mechanisms in place do their thing. The market for mining will take care of the rest.

With that said, I welcome more analysis on the concerns raised above.