Thursday, December 15, 2022

The Power Of Ethereum's Merge Revisited

Alex De Vries has published Cryptocurrencies on the road to sustainability: Ethereum paving the way for Bitcoin, a detailed review of the energy implicantions of Etereum's switch from Proof-of-Work to Proof-of-Stake. This analysis broadly concurs with mine from The Power Of Ethereum's Merge that while the power reduction of Ethereum's network is of the order of 99+%, the impact on the total energy consumption of cryptocurrencies is much less. Below the fold I discuss the details.

De Vries' abstract reads:
Amid the current climate emergency and global energy crisis, regulators have started to consider their options to limit the power demand of cryptocurrency networks. One specific way crypto-asset communities can limit their environmental impact is by avoiding or replacing the energy-intensive proof-of-work (PoW) mining mechanism. Ethereum, the second largest crypto-asset by market capitalization, had its PoW replaced with an alternative known as proof-of-stake during an event called The Merge on September 15, 2022. In this perspective, the likely range of electricity saved due to this change is estimated, while the limitations in assessing these figures are highlighted. Lastly, the challenges and opportunities in replicating The Merge on other cryptocurrencies such as Bitcoin are discussed.
De Vries is clear about the challenges of estimating network energy use. First the pre-Merge lower bound:
A major limitation in estimating the power demand of a PoW network is that even though it is possible to estimate the total computational power (known as the hashrate) in the network, the exact distribution of participating devices and their (overhead) energy costs is not known. The only figure that can be estimated with a high degree of certainty is the minimum power demand of the PoW network. ... At the time of the Ethereum merge, on September 15, 2022, [the most efficient] device was Jasminer’s X4 with a maximum hashrate of 2.5 gigahashes per second at a power draw of 1,200 W. At an estimated total network hashrate of 871 terahashes per second on the day before The Merge, assuming this hashrate is solely coming from Jasminer X4 devices, the total power demand of the network would be at least 418 MW.
It has been generally assumed that a large (though unknown) part of the Ethereum network was comprised of GPUs. Reperforming the previous calculation with a top-performing GPU, such as the Nvidia RTX 3090Ti, with an estimated hashrate of 132 megahashes per second at 346 W, would already yield a total power demand of 2.23 GW instead of 418 MW. A realistic estimate for the Ethereum network’s power demand before The Merge may be even higher, as a PoW mining network typically comprises a mix of device types, and additional energy costs may be incurred for cooling large batches of these mining devices. A tracker by Kyle McDonald estimated Ethereum’s power demand at 2.44 GW before The Merge.
Next, the pre-Merge upper bound:
Moreover, the Ethereum Energy Consumption Index put this figure at 8.88 GW. The latter is approximately the maximum power demand Ethereum miners may have been able to afford, as they earned roughly 13,000 coins per day from mining. With Ether trading at around $1,700 USD in the week before The Merge, this amount translates to an available income of $22.1 million USD. With an electricity rate of 10 cents per kWh (commonly used in mining profitability calculators), the power demand of miners should not exceed 9.21 GW to avoid operating at a loss (also assuming no further expenses other than electricity).
Then the post-Merge bounds:
At around 4,500 total nodes in the Ethereum network in the days after The Merge,14 assuming every node is a Raspberry Pi 4GB, this node count would translate to a power demand of just 36 kW. Using an enterprise server could increase the power requirement per individual device to 100–150 W, but it would still limit the total power demand of the Ethereum network to 675 kW, even if each node runs at 150 W.
Finally, the range of power reduction:
Thus, in light of the previously calculated power requirement ranges, it appears likely that the Ethereum network reduced its power demand by at least 99.84% ... In absolute terms, the reduction in power demand could be equivalent to the electrical power requirement of a country such as Ireland or even Austria.
De Vries and I agree that the reduction in the Ethereum network's power consumption is very large, and praiseworthy, especially in the light of the eight years of sustained effort needed to implement it. But, unless the network is implemented exclusively with the not-recommended Raspberry Pi 4s, it still leaves the network's per-transaction energy usage far above that of a credit-card network:
For example, with Ethereum handling roughly 1.1 million transactions per day after The Merge, the average electricity consumed per transaction ranges from 0.8 to 14.7 Wh. In comparison, a Mastercard transaction consumes only 0.7 Wh on average.
De Vries discusses one reason the decrease in Ethereum's usage of somewhere between 418MW and 9.21GW does not map to a similar decrease in the total usage of cryptocurrencies:
Moreover, it has been suggested that Bitcoin miners have been able to take advantage of the data center space that became available after the reduction in Ethereum mining activities. Between 250,000 and 500,000 new Bitcoin mining devices reportedly remained unused as only a limited amount of rack space was immediately available.
In A Huge Glut of Bitcoin Mining Rigs Is Sitting Unused in Boxes, Eliza Gkritsi writes:
Hundreds of thousands of brand new mining rigs that could be generating bitcoin (BTC) have never been used, further skewing the economics of cryptocurrency mining, a sector that has been hit hard by sinking prices for bitcoin and other tokens and by high energy costs.

Last year, miners struggled to buy enough rigs. Manufacturers couldn’t fulfill orders fast enough. Now, Matt Schultz, executive chairman of bitcoin miner CleanSpark (CLSK), figures 250,000 to 500,000 mining rigs are still sealed up in boxes in the U.S. alone, based on his conversations with analysts. Ethan Vera, chief operating officer of mining services firm Luxor Technologies, put the number at 276,000 worldwide in September.
BTC hash rate
But the evidence for increased Bitcoin mining consumption is far stronger than merely "suggested". The post-Merge spike in the Bitcoin hash rate is clearly visible, from around 222EH/s to around 273EH/s, or an increase of about 23% in hash rate. This doesn't map directly to a 23% increase in power consumption; it could be less if the added rigs were all brand new with leading-edge efficiency, or it could be more if the miners powered back up less efficient rigs.

Before The Merge, the central estimate of the Bitcoin network's consumption was around 11GW, with a range of abou 5-18GW. A 23% increase would be about an additional 2.5GW. Compare this to De Vries' range for the Ethereum network's pre-Merge consumption of about 0.4-9.2GW. If the Bitcoin spike was entirely reusing rack space that was previously mining ETH, the power usage of those racks would be roughly unchanged. Pre-Merge estimates were that Ethereum used about half the power of Bitcoin, implying an estimate of 5.5GW. If that were the case, about half the reduction in Ethereum usage was used to mine Bitcoin, roughly matching my earlier estimate.

De Vries discusses the possibility of eliminating most of Bitcoin's energy consumption by switching it to Proof-of-Stake:
the community behind Bitcoin has not worked on preparing a change to PoS, nor is there any substantial willingness in the community to do so. A campaign called Change the Code, launched by Greenpeace in March 2022, aimed at getting the Bitcoin community to replace its PoW mechanism. However, the campaign was met with hostility from the community as the underlying software’s lack of change (immutability) is seen as a key feature. The Bitcoin community also has a history of resisting substantial changes to the Bitcoin software, with one notable example being a past attempt to upgrade Bitcoin’s transaction-processing capacity. ... Any future attempt to replace PoW with PoS in Bitcoin might meet a similar fate, as the underlying network is decentralized and, therefore, does not have a central authority to enforce such a change.
This analysis omits a very important reason why such a change isn't going to happen. Ethereum miners, except the few using ASICs, could repurpose their rigs, but Bitcoin miners cannot. Thus a switch would instantly obsolete all the miners' investment in hardware and, since Proof-of-Stake requires much less rack space, most of their investment in data center infrastructure. The Bitcooin miners are already in severe financial difficulties; they are defaulting on loans and exploring bankruptcy. It is never going to be possible to persuade them to switch voluntarily.

No comments: