Let’s see… how much of a game changer is the Bitmain Antminer Z9?
Bitcoin GPU mining topped out at about 3 MH/s/W. The Bitmain Antminer S9 does about 10 MH/s/W.
So it took only about 3.3X MH/s/W to drive GPUs out of the business of SHA256d mining altogether. Is that right?
Could someone who has studied the tech and the history here check me? Are those numbers right? Is the Bitmain Antminer S9 a good example of the kind of thing that drove GPU miners out of the SHA256d business?
The Z9 specs say that it does 10,000 Sol/s at 300 W. That’s about 33 Sol/s/W, right? And the best modern GPU miners appear to do about 4.1 Sol/s/W, according to Hashrate/power usage observations and Best Sols to Watt Thread . So the Z9 is probably about 8 times as efficient in Sols/s/W than the current GPUs? Is that right?
I find this stunning, if correct! I would have assumed that the Equihash ASICs were more efficient than the Equihash GPUs, but that Equihash’s memory-oriented architecture would have made the advantage of ASIC implementation of Equihash much less efficient compared to ASIC implementation of SHA256d, but it appears the opposite is true!
SHA256d should be basically optimal for ASIC implementation, so the only explanation that I can currently grope toward, in my shocked state, is that all of our clever design work that went into Equihash (starting with the scientists behind Equihash, who designed it before they had ever heard of the Zcash project), made Equihash mining harder for GPUs more than it made it harder for ASICs!
This illustrates something that I think is critical to the debate here and that I think a lot of members of the community need to stop and think carefully about and stop just brushing it off:
If we rush to deploy a new “ASIC-resistant” algorithm, it could, for all I know, be even more friendly to ASICs than our first attempt appeared to be. We have no idea how the Bitmain Antminer Z9 works. Unless my back-of-the-envelope numbers above are wrong (maybe I’m comparing apples to oranges at some point), then the way it works is very different than the way we thought ASIC miners worked when we designed the first Zcash PoW.