Monero’s hashrate has fallen (pictured) to just 144 Mhash per second from close to 1,000 Mhash/s after the network hardforked this Saturday.
The fork contained a Proof of Work (PoW) tweak to the CryptoNight-R algorithm with the aim of “curbing” the ASICs currently present on the network and to further preserve ASIC resistance according to the release notes.
From the hashrate movements, some 85% seems to be asics, with suggestions that as much as 95% is botnets, FPGAs (programmable GPUs) or asics based on the fact that more than 95% of the hashrate was unknown prior to the fork.
The algorithmic change wouldn’t affect botnets, which are basically “enslaved’ CPUs/computers that knowingly or unknowingly answer to a virus owner.
The focus instead was to get rid of asics, with it unclear whether that would affect FPGAs. A reasonable analysis however would be that botnets make about 10%, with that now increased to 50% or more following the asics fork.
This algorithmic change is just months after Monero forked in May last year to remove asics. Hashrate plunged back then by 85% too. A 51% attack followed.
A Failing Strategy?
The idea was that if asics manufacturers know they will be forked out, they will not develop asics. However, going from zero asics after the last fork, to 85% asics in just ten months, appears to invalidate that suggestion.
The reason might be because developers simply can’t keep up with manufacturers. Devs have a trillion things to do, while the manufacturers are only concerned with how to speed up the algo.
We’ll quote at some length a recent conversation between a representative of an asics manufacture, Linzhi, and Monero devs:
“[08:11:38] <linzhi-sonia> after initial skimming, I would think it’s possible to make a 10x asic for RandomX. But at least for us, we will only make an ASIC if there is not a total ASIC hostility there in the first place. That’s better for the secret miners then.
[08:13:12] <linzhi-sonia> What I propose is this: we are working on an Ethash ASIC right now, and once we have that working, we would invite tevador or whoever wants to come to HK/Shenzhen and we walk you guys through how we would make a RandomX ASIC. You can then process this input in any way you like. Something like that.
[08:13:49] <linzhi-sonia> unless asics (or other accelerators) re-emerge on XMR faster than expected, it looks like there is a little bit of time before RandomX rollout.
[08:14:22] <sech1> 10x in what measure? $/hash or watt/hash?
[08:14:46] <linzhi-sonia> watt/hash…
[08:16:37] <sech1> I’m trying to understand how it’s possible at all. Why AMD/Intel are so inefficient at running FP calculations?
[08:18:17] <linzhi-sonia> hardware development works the other way round. We start with 1) math then 2) optimization priority 3) hw/sw boundary 4) IP selection 5) physical implementation
[08:22:32] <sech1> This still doesn’t explain at which point you get 10x
[08:23:07] <sech1> Weren’t you the ones claiming “We can accelerate ProgPoW by a factor of 3x to 8x.” ? I find it hard to believe too.
[08:30:20] <linzhi-sonia> sure
[08:30:26] <linzhi-sonia> so my idea: first we finish our current chip
[08:30:35] <linzhi-sonia> from simulation to silicon 🙂
[08:30:40] <linzhi-sonia> we love this stuff… we do it anyway.”
Monero finds itself in a fairly difficult position because they differentiate by being mineable on CPUs. The problem, however, appears to be that manufacturers can produce asics far quicker than devs can change the algo.
Defending Decentralization or Gov Intervention?
Zooming out slightly, any coin is mined because it generates profit for the miner. The current hash of a coin, thus, and taking it to a conceptual abstraction, is say 1x profit. If one can speed up the algo by 10x, then there’s a fairly instant 10x profit.
In the abstract, price is irrelevant especially if it concerns a top 13 coin. If price goes down, hash goes down, so whoever remains is still at say a perfect 1x, with that 1x being the profit.
Thus, 10x-ing something can be hugely profitable, even if it is for a short time. Logically, therefore, there would have been specialists looking at the code development and by the time the code is out, such specialists may have well produced the new asics that works on the new algo.
Making it a developers’ waste of time and potentially far worse. Worse because very few have the resources to keep up with the minute details of algo changes, but those few have huge incentive to do so.
Thus if we look at Monero’s hash, and you can’t see if very well but there’s a big dive in April 2018. That’s not the real dive. Within days, what are probably secret miners turn on their asics. Then we have the usual hash following price dynamics, with a jump recently and another dive.
Looking at this, if you have two brain cells and the talent of algo speeding up, you’d think next time you should try and turn your machines far more slowly. If you’re a bit cunning, you might even put out some fake PR to claim CPUs or botnets mining monero are on the rise etc.
The one that misses out here is the honest miner because he is forced to be “dishonest” in as far as he has to operate in secret even though he might have well liked to create a dynamic, competitive, mining environment, and thus a decentralized space.
Plenty of miners obviously are cryptonians. Some don’t care, some are inclined towards cheating, but you’d think the majority would rather do the right thing while profiting.
That means they might want to sell asics to the public so that the bedroom CPU guy who competes with botnets (zero chance) or the bedroom GPU guy who competes with vast industrial farm (zero chance there too), can instead be the bedroom asics guy who competes with vast asics industrial miners.
There too our little guy has zero chance, but practically speaking, it is almost zero rather than literally zero. If asics producers are instead forced to operate in secret, then the bedroom asics guy wouldn’t exist at all.
In effect, an algo change is sort of government intervention in the free market. It distorts incentives and it forces good guys to be bad.
If you look at bitcoin, they have so many asics producers all competing with any of them at risk of going under. Collectively, if any of them misbehaves, then the entire asics industry is effectively put out of business by a defensive algo change which so far has, and probably for long will, remain just a threat.
The threat itself is probably sufficient to add more incentives for any and all miners to well-behave. So we get a system where $5 billion is moved a day without any problem.
While with these algo changes, imagine you’re buying a house – as now some do with bitcoin – and the hash just plunged 85%. Is any amount of confirmations enough?
If an algorithm that can not be asics-ed is found, then of course we’d love to mine on our laptop. With that however being impossible as far as current knowledge is concerned, buying a day or two of CPU mining while significantly distorting incentives appears to be not very wise.