Topic: Ethereum Anti-ASIC fork, is it the right time for bitcoin too? (Read 2475 times)

I'm all for increasing decentralization and involving as many parties as possible to securing the network, but I also despise the feuds that follow forks and the brand confusion that always seems to follow. Knowing how the bitoin developers think, their top priority will be security. As long as this can feasibly be done with widespread user and industry consensus, I'm all for it.

Anything computers can do (to solve PoW) will be casted into ASICs or how ever u want to call These. 

This is a dead end.

today I have received this link about the cost of Progpow and ASIC design and could give us more useful ideas for further discuss:

https://medium.com/@ifdefelse/the-cost-of-asic-design-a44f9a065b72

"Introduction

The speculation about hardware design and development costs as it pertains to both ProgPoW and Ethash are usually followed by a statement of authority: trust the author, because they have previous experience in . Sometimes, it is cryptocurrency-ASIC production, other times, it is integrated-circuit design.

For the audience who is more familiar with code rather than fan-out and rise-times, a level of perspective on why this statement of authority does not apply to ProgPoW may be helpful."

Who told you S9 is profitable? S9 doesn't pay off for electricity in most countries where it costs more than  $0.05/kwh. It is how you are trying to disqualify my points?  

Bitmain has delayed S11 and S15 for other reasons and now they are out and miners have no choice other than buying.

No comments!

I guess it's important to at least give something to back up what you are saying, rather than yet again making claims with no proof, that are also incorrect.

Just to pick the above quoted example:

Saying that ASICs don't last for more than 2 years is easy to disqualify.
https://bitcointalksearch.org/topic/bitmains-released-antminer-s9-worlds-first-16nm-miner-ready-to-order-1493601
S9 "May 31, 2016" - 2 years 9 months ago and still profitable to mine with them - and still a large part of the Bitcoin mining network.

Meanwhile I have a Canaan A741 that I use as a heater in winter since they can do the equivalent of a heater by slowing them down enough to make the noise sub 50dB but of course mine BTC as well as using similar power to heat a room ... they are about 2 years old

On the other hand, using bitmain miners to decide that all ASICs break more frequently than GPUs is also incorrect.
Simplest example to refute that is Canaan miners - that have nothing like that 30% failure rate. Well under 10%

GPU scamcoins do not get the equivalent growth that Bitcoin has - so with scamcoins it's the lack of growth that make GPUs last longer, not some inherent design that makes GPUs better.
If an algorithm has enough growth it will get ASIC hardware designed for it and then bye-bye to GPUs.
Most all altcoins are scams and thus they rarely invite any sort of growth to make this happen, so that discounts that argument of yours also.

Aside - as an example of not knowing ... S9s have been in the $5k price range around the time of the top of the BTC price spike a year ago

Political figure? Gmaxwell is not doing anything "political". He is viewed by the community, like some of the other Core developers, as a "leader" because of his competence, not politics.

Roger Ver is the one trying to put politics into everything. But when debated on more technical matters, he spins the argument, and makes it like he's right and everyone is wrong. He's a gaslighter, and an altcoin scammer.

@WIND_FURY,
I understand besides his undeniable expertise in the field, Greg Maxwell is a political figure and should be handled more carefully, but what I don't understand is how to manage the situations like this, when he shows up with "case dismissed" verdicts and accuses me of being part of a "constant influx of suckers" who have no clue about anything.

Anyway, I'm sorry about what happened in last few posts and I do my best for this not to happen again.  

I believe we should get gmaxwell's attention to read your latest post, and hopefully he replies to continue the debate. That's if he did not put you on ignore. There's so much to learn in my opinion, and it would be bad if the technical debate stopped because of personal attacks.

You are implying that I've not answered Gregory Maxwell's arguments and instead have attacked him personally, It is not true:

1- Gregory Maxwell inductive reasoning: "ASIC resistance has not happened and is very unlikely to happen in future" has been refuted strongly by me:

• Induction is a poor reasoning instument in a technological context
• Even with such a type of  reasoning, one of the premises: Ethash being a failed ASIC-resitant attempt does not hold, actually  Ethash should be considered a success, so,  the conclusion is false. Ethash is a success because it is not basically broken by Ethash Asics as they aren't more efficient than 2x and they have not managed to overtake the network within a year

2- I refuted his allegations against ProgPow and made it clear that it is an open-source and public domain project and such claims are absolutely void. It's really absurd, how in the hell an algorithm would be considered "corruption"?

3- Greg Maxwell says gpu mining is investment-bound but ASIC mining is cost-bound. After my reply, he added an example to his post:

• The example is confused and misleading: What was the price of a S9 "last year"?  What is the electricity cost? What happens when Bitmain is out of stock? I've seen prices in 2000$ range "last year" on ebay for used S9  without PSU. ASIC prices are more sensitive to both manufacturer market manipulation and bitcoin mining profitability.
  
• It is admissible that a gpu rig is like 2-3 times more investment intensive compared to an ASIC with same power consumption figures but it is compensated by:
  
  • Maintenance/repair costs: ASICs break much more frequently unlike gpus that typically work for years without a single glitch. My experience: 30% failure, more than half irrecoverable.
  • Lifetime: ASICs do not last profitable typically for more than 2 years, the manufacturer releases new versions with much better efficiency and forces the miners to upgrade, it is why we've experienced exponential total hashrate surge without comparable increase in total power consumption.
    GPUs remain tolerably efficient for much longer times when they are mining a memory bound pow algorithm.

4- He goes further and argues (because of patents and first-mover advantages involved), gpu mining is more prone to central control and monopoly, while:

• Fresh cutting-edge gpus coming out from research labs would never disrupt gpu mining industry for many reasons:
  
  • They are highly expensive and do not justify their costs in mining.
  • ProgPoW is tunable to become less computation intensive (it is an open discussion topic right now, I'm arguing about it with ProgPoW community), once it is tuned properly, it would make advances in gpu core technology almost irrelevant to mining because memory remains the king as the bottleneck and it is not something you could improve every couple of years, we have GDDR5 technology on the edge for almost 10 years! It is why we can continue mining Ethash with a r9 270 (2011 tech AFAIR) with still tolerable costs and it is why NVIDIA luxury GPUs don't dominate Eth mining scene, AMDs do.
  • As a commodity device a gpu is available to public and average user, without such an availability it is hardly a gpu at all.
    Government control is absolutely irrelevant in this context, no government would/could ban manufacturing/trading gpus because GPU manufacturers have a much larger market than mining: gaming/rendering/machine-learning/AI/...
• Comparatively:
  • New generation of ASICs, appear every 2 years with 2-3 times efficiency. manufacturers have incentives to take advantage of their products in an obvious conflict of interests practice and decide about how and when enforce miners to replace their stock.
    
  • ASICs are highly vulnerable to first-mover advantage it has been proved historically and can be realised logically
  • ASICs definitively escalate government control because they are application specific and it would be easy to ban or restrict their manufacturing/use by law/regulation/...

Most of what I've presented here is thoroughly discussed in the post that you take it as a personal attack to Gregory Maxwell.

It wasn't my intention to "attack" him, I didn't insulted him personally, I just accused him of being biased and making up justifications for one of 2-3 mistakes in bitcoin history: not resisting ASICS.
People make mistakes, it just happens, people deny and try to justify their mistakes it happens too.

But there is no obligation for Maxwell or anybody else to make excuses about the situation with ASICs or pools.
Bitcoin is young and we are all learning and experimenting but Gregory Maxwell was just opening a frontline no matter how flimsy it would be, just vague enough to be used as a "there are some arguments and excuses for bitcoin being ASIC-ed " thing.

 

Can you look beyond that, and make a constructive debate for ASIC-resistance? It might end up in both of you "agreeing to disagree", but I believe it would be productive for everyone to pick something from it, and learn something.

Try again please, https://bitcointalksearch.org/topic/m.49844670

@gmaxwell, just step down from the horse already.
You made weird and harsh arguments and you got the proper answer. You are welcome to choose hills for residency but the battle is happening right here, in the watersheds.

As part of decentralization agenda, ASIC-resistance is not a joke to make fun out of it by forging terms like "commodity energy" and reversing all the established discours around it like what you did by claiming commodity gpus being vulnerable to government control and Chinese made ASICs not!

You are not an ordinary forum member, you are the most distinguished bitcoin figure, you should better take care of the political implications of what you say and what you weigh on instead of jeopardizing your reputation by weighing on such baseless arguments.

If it wasn't @gmaxwell saying weird things with the famous tone, I wouldn't go that far, I had no choice other than reminding you this simple fact that you are used to forget: You are Gregory maxwell!

If anyone needed yet another example why the world of cryptocurrency is a cesspool... and why thoughtful people usually run for the hills-- and leave tech forums looking like ghost-towns with little interesting traffic left, well then I guess one has been provided.  I thought I gave a brief but reasoned point by point argument that made the case both from first principles and then again drawing from actual results. I might have expected a response in kind but instead, the response is a litany of abusive invective and personal attacks making it clear that my time-- and everyone elses here-- is just being wasted.

You dare not challenge the strongly held prejudice of the forum warrior. You just don't.

I originally thought this thread was likely to be a train wreak and didn't bother commenting. But I was in a good mood this evening... some day I'll learn.

He calls me and people who are in favor of ASIC-resistance "a flux of suckers" who are like newcomers that do not understand anything and buy hypes and fake stories and are so stupid to fight for it and they don't know that Greg the great and his paws have deliberately put this agenda ASIC resistance aside and refuted it, in ancient history of glorious bitcoin  

Gregory Maxwell is one of the best things happened to bitcoin, I like him more than you, but he made a LOT of arguments in that post which are very dangerous and misleading, I'm ready to fight, it is an official declaration of war and he is probably ready for the consequences.

 

Right from the start of your reply, there are already personal attacks. It might make everyone question your motives in the forum.


Can you please attack the debate, and avoid attacking the person? Thanks. There some people in the forum who wants to learn more from this topic.

Here we are again: saint Gregory Maxwell never made a mistake and our situation with ASICs was destiny, as the situation with pools and the synchronization nightmare.

But it is not true Greg and you know it already, don't you? you are a scientist after all. I wonder what kind of a scientist argues like that: "because Scrypt, Equihash and Ethash failed then it is unachievable"! Firstly,  this is a technological problem, in such a context induction is the poorest reasoning choice and secondly, Ethash is not failed. It has been somehow damaged but far from being a failure. Bitmain spent $$ to beat it but all they could do was a 2x efficiency gain over low-end gpus and a truce with high-end ones and falling behind of advanced gpus.

So, you are using a poor reasoning technique: It happened the other days, SO, it will happen forever (or not happened and won't happen respectively) and yet you fail to use it fairly, because you are biased. It is by no means a useful contribution, are you sure your account is not hacked by @WIND_FURY, it is what he is rehashing all over this thread and it really sucks.

It is open-source. Bitmain and ASIC manufacturers and their army of shills say things about Nvidia class of gpus fitting much better to ProgPoW which is not true, I expected more from you, a political propaganda against an open-source project. On the surface, we have money and power that speaks, in the depths we have science and reasoning.


It is the most unacceptable and misleading argument I've ever heard regarding ASICs. Congrats Jihan, you got a genius on your side!

The lifetime cost of a SHA256 ASIC is not electricity, who told you that? ASICs become obsolete by their manufacturer regularly. Miners need to renew their farms everytime the manufacturer releases a new version.  I have a RX295-x2 in my site, I bought it second-hand and it is mining Eth from the first day (2015), it has made ROI multiple times and is mining Eth in-spite of everything (ASICS, prices, multiple new generations of gpus, ... ) and competes in profitability with a brand new gpu. I have rx470s with same history. When I bought my rx295 it was still S5 erra in bitcoin, now my S9 is to be stocked and replaced by S15.

It is sad. As a prominent figure you shouldn't do this. You are using the classical arguments against ASICs in favor of them. Don't do this, stop weighing on ridiculous ideas like:

It is absolutely nonsense. Pure garbage. No government in the world can do anything about commodity devices like gpus simply because they are commodity devices for the god sake!
And there is no such thing, commodity energy, you have just made it up to use in the context of this debate for making everything foggy and uncertain.

With all due respects sir, you are spreading FUD against ASIC-resistance.

Heavy cost R&D projects end to highly priced gpus that are not a threat to average gpu mining rigs because their supply is short and their efficiency gains does not justify the price. It has always been the case in this industry and will get even better because gpu-friendly algorithms are typically memory bound rather than computation bound. Governments have nothing to do with commodity devices like CPUs and GPUs, advanced, top secret technologies are expensive and do not ROI even close to what commercial (commodity) devices do and are not part of the picture in mining industry.

That's everything that I have been struggling to say done in one post, but in a very articulate way of presenting it. Thank you, gmaxwell.

On the surface at least progpow looks like straight up corruption. If ethereum users won't fall for it, there isn't a chance in hell bitcoin users would.

It appears very unlikely that "asic resistance" is even possible, but putting that aside: there is also little evidence that it would be desirable.  Absolutely asic mining has created a multitude of problems but something having problems doesn't suggest that any particular alternative is better... the alternatives to common practices with known issues are often much worse.

In particular, essentially by definition every "asic resistant" approach I've seen shifts costs to from highly commodity energy (the overwhelming majority of the lifetime cost of a sha256 mining device is power) to highly illiquid upfront manufacturing. (e.g. bitmain s9 last year hit power cost = purchase cost at about 3mo of operation) The economic arguments for mining potentially producing a secure and decentralized system depend on an ongoing cost-- if you reject those models you end up with something more like a quorum of trusted signers (e.g. "bobchain").

Manufacturing centric costs also seem highly prone to even stronger monopoly pressures both due to  government granted artificial monopolies on manufacturing techniques as well as natural monopolies arising from first mover advantage for a highly amortized R&D heavy cost model compared to a commodity energy cost model.

This speculation about the trade-offs also seems to be supported by prior efforts at "asic resistance"  (including litecoin, ethereum's work function, zcash's work function...):  In spite of many hypes and claims they have not actually turned out to be asic resistant at all and have in actuality resulted in a less diverse ecosystem than we've seen for Bitcoin. And enormous windfalls for secret operations...

Perhaps even more fundamentally: every commodity computing device is itself already an "asic"-- just one with extraneous parts (and a lot of patents required to build it, the "commodity" is something of a misnomer as a result...).  Simply taking an existing design and stripping off the extraneous area and power wasting I/O components alone is usually enough to make standard parts uncompetitive. After all, mining by its very design seeks to result in near zero returns for the _most_ efficient participants-- you don't have to be much less efficient to be hardly earning at all (or even operating at a loss). 2x the efficiency can mean 20x the profits, if the market is actually competitive. This is especially true because mining devices operate in a very different reliability region than other computing devices: it's typical for a mining device to get the best good-put at a percent or even more incorrect computations because the speedup/power-reduction more than pays for the wasted work. This sort of operation wouldn't be acceptable for virtually any other application, so a lot can be gained simply by designing for that alone.  In practice, every attempt so far (though much hyped, lauded, backed by academic research, and what not) has actually had orders of magnitude more efficiency gains than you'd get from stripping down a generic part. ... but even just the stripped part is likely enough to fatally wound the idea on its own... Before considering the IPR monopoly, amortization-centralization, risk of accidental trapdoor optimizations, or the speedups from customized architectures.

To me it seems that the idea doesn't work in theory and it doesn't work in practice.  It keeps re-occuring because the constant influx of suckers who will buy into things that merely claim to solve things which intuitively sound problematic, even if the solutions are barely credible (or even if they were tried previously and already failed!). There is little interest in actually reasoning out the tradeoffs, e.g. what makes for a good work function and a vested interest to not bother and keep churning out new flavours of snake oil-- because no matter how you want to cut it the picture is not that rosy for these efforts-- at best you might be able to say that with a lot of careful work you got a DIFFERENT set of tradeoffs that were potentially useful something with its own benefits but own serious downsides, far from the panacea that people want to sell-- and that isn't good for shilling the latest scamcoin.

But Bitmain is already eliminating itself. Jihan Wu has already left the company, and they are deep down in seas of bags of the BCH fork (the Ver side).

So as a result, they are now closer to bankruptcy than ever before. Perhaps this is part of the incentive-game theory mechanisms of Bitcoin at work. They got involved in the wrong end of the fork and paid for it.

As far as more exotic approaches, it would need to be implemented in some altcoin first but I guess that would still be useless since only when things are running on Bitcoin you can tell if they are fit or not for survival in the wild, and it may be too late for this now. How do moons align and a hard fork gets all the network effect from the legacy chain switching? think about all the incentives and game theory behind it.

PS: Im not against research of ideas that could be added in an hypothetical hard fork, I am just realistic and interested on the game theory behind and hard forks and how unlikely it is that the new fork gets enough consensus.

Which is why  making a coin to have an asic and a gpu linked via an acorn type controller for max profit. Would make the asic company think twice, instead they have an incentive to cut a deal with  amd or nvidia and with an fpga builder.

that holy trinity concept  fuses  companies together.  You can then use the mining tech as a financial tool to expand renewable power.

Of course it is, if taken in the right context. Let's pretend that Ethereum made the change to ProgPow, and the coin has rallied to $10,000 per ETH, a huge success for the cryptocurrency. Would the ASIC companies not develop any specialized hardware for it? Impossible. They would risk millions in investment to be the first in the race to build one.