Debunking those who think they’ve invented ASIC resistant proof-of-work
Can you take a look at much newer solutions: snowblossom (proof of iops of storage with large snow fields
I explained the invalidity of his assumption on the author’s blog:
Consensus is PoW mining based on a random reads into a large snow field. This makes the mining more IO bandwidth intensive than CPU/GPU/ASIC. Sure, someone could build something that does more raw IOPS than a regularly available NVME drive or DDR4 stick, but it probably won’t be much more price effective. This means mining will be in the hands of a large number of people. Hopefully.
Incorrect assumption. I presume you understand that the economics of proof-of-work is fundamentally tied to the electrical power consumption, and not so much the hardware cost. A consumer memory or storage device is not power optimized, because it has to trade-off many different use case patterns. When the use case is narrowed by a specific proof-of-work algorithm, then there are further optimizations that can be made. Ethereum’s Ethash was designed to be memory I/O intensive and yet significantly more efficient ASICs were created:
A major motivation for memory hard PoWs is to limit performance by the memory IO bottleneck […] it behooves a memory hard PoW to also have increasing memory requirements. Ethereum’s ethash PoW recognized this in having a linearly growing dataset (growing by 8 MB every 125 hours […]
The predominance of electrical power cost in TCO would be especially true for commodity memory technologies which don’t advance as fast as Moore’s law, but less true for ASICs which employ large amounts of memory on die because of the high cost of non-commoditized silicon:
A typical highly optimized Bitcoin mining chip produces between 0.3 and 0.5 watts per square millimeter. Because of this heat profile, a typical $1200 miner may cost $1200 per year to operate, primarily due to the cost of electricity. For an American mining on typical consumer electricity rates, that annual cost is more like $2400 per year, meaning that consumers really cannot afford to mine at home.
Our Cuckatoo31 chip has a heat signature that’s less than 0.1 watts per square millimeter […] The same $1200 spent on a mining device results in electricity costs that are closer to $400 per year for a typical mining farm, and $800 per year for a typical consumer. The total cost of ownership gap between a consumer and a professional farm is substantially lower for cuckatoo miners […] When optimizing over total cost of ownership, wafer pricing becomes a lot more important […] 16nm silicon is cheaper and more accessible, which means that 16nm chips are more competitive [and increasing the cost of the silicon relative to electricity in the TCO may increase mining centralization due to not needing to seeking out lowest electrical power costs]
Additionally there’s an asymmetry because memory devices only consume a fraction of the power in a computer system, so a customized rig will have much better efficiencies, even if ASICs are not factored into the mix.
The Snowblossom design is going to be more power efficient running on GPUs than CPUs, because of the IOPS efficiencies on GPUs due to coalescing of memory accesses. This will make CPUs uneconomic. Yet compared to ASICs, GPU memory access coalescing and latency masking is optimized only up to the level where the GPU typically becomes computation-bounded, so as to make GPUs cost effective for the general purpose computation for which they’re designed. Whereas, in theory an ASIC can be designed achieve higher levels of such efficiencies, because essentially the ASIC will never be computation-bounded due to the ability to optimize the computation circuits for the specific Snowblossom algorithm.
Additionally, apparently the Snowblossom fields are read-only, so the ASIC designer may have the option of choosing among a wide range of non-volatile random-access memory technologies so as to further optimize the power consumption per “hash”. GPU-only proof-of-work doesn't scale equitably because the supply of GPUs can't scale, so is actually worse and more irresponsible design than ASIC friendly proof-of-work.
Additionally memory technologies have many factors (c.g. the first link in this comment post) which can be presumably optimized for given pattern of the repetitive use case, e.g. optimizing for prioritization of read (not including writes) power efficiency in the case of Snowblossom. There’s even a risk that competition in the ASIC space decreases due to complexities of segmenting the memory systems and/or some miraculous technological innovation in memory technology could become patented, enabling a monopoly on mining the proof-of-work!
It’s impossible to design any proof-of-work which depends on burning electricity which wouldn’t be orders-of-magnitude more efficient in a customized circuit (c.f. also and also), unless someone can discover a way to make “proof-of-storage” work wherein the best an ASIC can do it rely on commodity memory power consumption. In 2013, I proposed “proof-of-storage” in discussions at bitcointalk.org and unfortunately I wasn’t able to figure out how to make it work.
As I explained to @tromp in 2016, an ASIC most definitely can be made that is more efficient than consumer NVMe or DRAM. A consumer memory or storage device is not power optimized, because it has to trade-off many different use case patterns. When the use case is narrowed by a specific proof-of-work algorithm, then there are further optimizations that can be made.
This is just a variation of proof-of-stake. Same bad smell. Unless something is burned during mining the new tokens, then the token has no value. Tokens that were supposedly sold for fiat or BTC when the stake was created, is unprovable (e.g. ICOs do not tell you that they buy their ICO from themselves) and (even if initial supply was distributed with proof-of-work) there’s no ongoing marginal cost of production in order to maintain the valuation. Read what was written about proof-of-stake (which I refer to as proof-of-nothing) here and here.
Note: None of these were written by the coins developers or team members nyzo developer is anon
Most people who are writing about crypto don’t have the requisite knowledge to analyse it correctly.
I hope you can join their discord to talk with the developers of above projects. They are active on discord.
They seem to possibly be good developers, although Java sucks and anyone coding in Java immediately is cast into doubt by me (being I am expert on programming language design). But they appear to lack requisite depth in economics and other technological research. They’re probably better off partnering with someone who is more expert, though the problem is these young guys (these days, unlike in my generation when we appreciated apprenticeship) typically don’t want to admit that they need to do that and unwilling to make the necessary sacrifices for that. C’est la vie.
I wrote a follow-up reply on Medium:
I’m sorry their Snowblossom design can’t resist ASICs. I edited my comment you replied to and added:
In short, it’s impossible to design any proof-of-work which depends on burning electricity which wouldn’t be orders-of-magnitude more efficient in a customized circuit, unless someone can discover a way to make “proof-of-storage” work wherein the best an ASIC can do it rely on commodity memory power consumption. In 2013, I proposed “proof-of-storage” in discussions at
bitcointalk.organd unfortunately I wasn’t able to figure out how to make it work.
I don’t have enough free time to attempt to teach the readers electrical engineering, game theory, and economics. They’ll learn the hard way, as the Monero developers and community also did.
I warned @fluffypony and the Monero community, so they banned me from their Github. Now they see I was correct.
The creators of Nyzo are lying, disingenuous, or not very astute. And I presume they’re astute, so it’s probably the former. Nyzo is a proof-of-stake system. The only distinction being that the stake resource is the set of all IPv4 addresses, instead of tokens. It will suffer from all the game theoretic and economics problems of proof-of-stake. To be blunt, another worthless nonsense PoS design. Please don’t waste more of my time by insisting there must be some chance that they've done something worthwhile and unique, which I then I have to expend more time to refute so that naive readers aren’t swayed by the FOMO hope.
I provided to you my opinion already on Steemit. I complimented them by writing that they appear to be reasonably good developers, although the use of Java is a red flag for me. But they ostensibly lack the requisite expertise in research, game theory, economics, etc. to be designing new consensus technologies. I have no interest in joining their Discord and try to convince Millennials that they should be more circumspect. They’re making money doing what they think values themselves highly. I should stand aside and let them take money from technologically-ignorant, greater fools. I have more important things to do. In the end, true valuations will be assigned. And they won’t be up there in the top echelon.
And wrote another:
1. you assume the creator of snowblossom may , 'take money from fool' which isn't fact. it has not premine no ico, and creator is already pretty well off in the past.
Premine and ICOs aren’t the only way to take money from greater fools by launching a proof-of-work cryptocurrency.
Here are some other ways:
Mine it heavily before it becomes well known, thus amassing a considerable share of the initial supply. Given you have pointed our that FireDuck was the creator of Satoshi Dice, he presumably has access to a lot of capital to mine a nascent, recently launched altcoin before it’s on the radar of many people.
Secretly develop a better hardware configuration and/or algorithm for the mining than the client and hardware presumed to be run by everyone else.
Use developer pronouncements to influence price moves, front running shorting and going long on exchanges. Which also has the risk that it ostensibly violates securities regulations in many countries can potentially cause them to get in legal trouble.
2. you assume that if asic can be made, than it's not good design.
That is not what I wrote. I was refuting the official claim that an ASIC is unlikely to be “price effective”.
fact, even in this article the snow creator did not say something more efficient can't be made
In fact, he claimed it was unlikely.
Asic can be made for bitcoin too in the long term, but bitcoin did not die from it.
I didn’t claim ASICs would cause Snowblossom based altcoins to die. I claimed that the author was making a bogus claim, and the Snowblossom doesn’t have the claimed uniqueness (i.e. not ASIC resist). If greater fools know it’s not any more unique than other 1000s of copycat, shitcoins in that respect, they might be more hesitant to doubt the hints the creators are making about a 5% chance of world domination.
what he said is it’s harder than my question is :even if asic can be made , economically, is anyone motivated to do that?
ASICs are built when the cryptocurrency becomes valuable enough, as was the case for Monero.
because when the snowfields doubled in size., people may need to build new asics again that is double the size…. and the cost doubled. suppose someone is doing it at scale profitablly (Idoubt, but not impossible maybe,) ,
That is what the Monero community thought and they learned it will not stop the ASIC development.
does it still save more energy than bitcoin?
This is an incoherent question.
And wrote another:
but will asic be able to do random assess on bulk data more cost effective than consumer ssd/nvme?
Yes an ASIC will be orders-of-magnitude more power consumption efficient. I edited my initial comment on their blog. You should re-read it. I provided more technological detail.
Plus nvme ssd that’s more than 1TB, is super expensive,
Irrelevant. Electric power consumption is all that matters in proof-of-work, especially more so with memory devices which are not improving as fast Moore’s law.
And wrote another:
the article was written by a community member, so you can’t just say nyzo developer lied, he didn’t even say anything
About nyzo , it doesnt’ really matter what one person says what it is, even if he ‘s someone famous or know/writes a lot.
If you’re not valuing my effort to analyze it, then please do not go on my Steem again and loudly ask me to come analyze something. Either you respect and value my contribution or you don’t. If you have some substantive technological rebuttal then make it. Btw, this butthurt, irrational stuff is why I have stopped responding to most Millennials. And I’m not famous, just diligent in my research and effort.
It does matter what one person says if they’re correct. Because those who follow incorrect information end up losers. If I’m incorrect, then someone knowledgeable needs to point out why I am. Also as you know I put some additional technological elaboration on my Steemit reply to you.
I think it’s much more secure than traditional proof of work altcoins, and POS coins.
Irrelevant. Do you have a technological rebuttal?