Ethereum Anti-ASIC fork, is it the right time for bitcoin too?

[Wind_FURY]

@ٌWind_FURY, the diagrams you've included have nothing to do with hashing power and its distribution, pools are intermediary entities and belong to another class of centralization threats and I've read something about Bitmain hiding its dominance behind fake names.

But it doesn't change the debate that Bitmain's technological advantage is not as dominant as what it was before. Making of ASICs is more distributed today, and hashing power is also more distributed.

A hard fork to change the mining algorithm would be an opportunity for a highly capitalized company to be a dominant force in making ASICs for the new algorithm. It would be back to the same problem again, and with one company having the technological advantage.

Your claim about the end of Bitmain dominance is not supported by facts. The most crucial problem with such companies is not the lack of competition tho, it is more about the lack of transparency and accountability and conflict of interests.

But they're not merely claims. Ebang International Holdings and Canaan have been catching up on Bimain technically, and they are also planning their own IPOs.

But again, we can ask the people who know. The miners. I will open the topic later.

As of your argument regarding the algorithm change having bad impacts on competition, it is based on a false assumption, a hype about some magical rule that says every algorithm is breakable by an ASIC. It is discussed and refuted above thread already:

Nothing has been refuted.

Why don't you refute "ASIC-resistance is a myth".

ProgPow for instance is provably safe against such an attack because it is designed for a commodity ASIC from the first place: a GPU. Attempting to break ProgPoW by a hypothetical non-gpu ASIC is highly de-incentivized because the feasible efficiency gains would be lower than 20% which comes with very high R&D and manufacturing costs that users should compensate for by paying very high prices. Nobody invests on such a project ever.

It's not. Sorry but that's a fact.

[1714719752]

1714719752

[1714719752]

1714719752

[1714719752]

1714719752

[figmentofmyass]

@ٌWind_FURY, the diagrams you've included have nothing to do with hashing power and its distribution, pools are intermediary entities and belong to another class of centralization threats and I've read something about Bitmain hiding its dominance behind fake names.

But it doesn't change the debate that Bitmain's technological advantage is not as dominant as what it was before. Making of ASICs is more distributed today, and hashing power is also more distributed.

do we have any hard numbers to support that? there were some new entrants to the market over the last couple years (including some very disappointing flops) but is anyone really competing with bitmain's scale at this point? is anyone even close?

i'd echo the attitude that hashrate distribution is pretty opaque. it's hard to know how much bitmain really controls.

A hard fork to change the mining algorithm would be an opportunity for a highly capitalized company to be a dominant force in making ASICs for the new algorithm. It would be back to the same problem again, and with one company having the technological advantage.

i sympathize with that position myself. however, there is a distinct possibility that implementing such a hard fork would act as a deterrent against bad actors like bitmain in the future.

there's no precedent for this stuff. is there?

[aliashraf]

@WIND_FURY

We have been thru this for a while, your assumption about ASICs-eventually-win is false and an advertised hype. There is no way to make FETCH_MEM operations to be ASIC friendly, no scientific way, not speaking of magical and mysterious powers of Bitmain  

Above-thread I've extensively discussed why and how ProgPow is provably safe against ASIC optimizations higher than 1.2x and I've provided supplementary resources as well, why don't you bother researching a bit more instead of rehashing Bitmain ads ever and ever?

[Wind_FURY]

@ٌWind_FURY, the diagrams you've included have nothing to do with hashing power and its distribution, pools are intermediary entities and belong to another class of centralization threats and I've read something about Bitmain hiding its dominance behind fake names.

But it doesn't change the debate that Bitmain's technological advantage is not as dominant as what it was before. Making of ASICs is more distributed today, and hashing power is also more distributed.

do we have any hard numbers to support that? there were some new entrants to the market over the last couple years (including some very disappointing flops) but is anyone really competing with bitmain's scale at this point? is anyone even close?

i'd echo the attitude that hashrate distribution is pretty opaque. it's hard to know how much bitmain really controls.

If it's opaque, then it's hard to know if Bitmain truly controls most of those pools. But I'll open a topic in mining speculation and gather information from the people who might have an idea what's going on.

A hard fork to change the mining algorithm would be an opportunity for a highly capitalized company to be a dominant force in making ASICs for the new algorithm. It would be back to the same problem again, and with one company having the technological advantage.

i sympathize with that position myself. however, there is a distinct possibility that implementing such a hard fork would act as a deterrent against bad actors like bitmain in the future.

there's no precedent for this stuff. is there?

This is what aliasharf wants. He wants to risk amputating the network of full nodes and hashing power by having the network hard fork to a new mining algorithm, ProgPow. What is his assurance that there won't be an ASIC developed for it?

Best example I can give, Monero. Their network hard forked to stop ASICs. Guess what, there are ASICs again mining their network. But their network and community is small, and they can arrive to consensus easier. I believe it will be harder to get consensus for something like that in Bitcoin. It would split the chain again.

@WIND_FURY

We have been thru this for a while, your assumption about ASICs-eventually-win is false and an advertised hype. There is no way to make FETCH_MEM operations to be ASIC friendly, no scientific way, not speaking of magical and mysterious powers of Bitmain  

Above-thread I've extensively discussed why and how ProgPow is provably safe against ASIC optimizations higher than 1.2x and I've provided supplementary resources as well, why don't you bother researching a bit more instead of rehashing Bitmain ads ever and ever?

In theory, it's "provably safe", not actually proven in practice. I believe the Core developers would agree. They will never listen to your proposal.

[buwaytress]

My non-technical logic says: we apparently have historical evidence that, given enough motivation and resources (shall we say, Bitmains appetite and technician workforce as a prime iteration of this), there will always be ability to develop a piece of specialised hardware that will deliver what an ASIC does.

So I see anti-ASIC as ASIC-resistant, not ASIC proof. Fully cognisant that there can be no ASIC proof.

The question to ask is not can we make it ASIC proof, but can we make it so continued ASICs development (or whatever tech will replace ASICs) are not easily profitable? Should we even?

I see the same argument with quantum computing. Bitcoin cryptography is highly computing resistant, but never proof, theoretically. I think we can safely assume Bitcoin will upgrade security before quantum computing threat comes close to reality. I'd be afraid if it didn't!

Also food for thought. Bitmain didn't really get to where it was because it had superior tech or because Bitcoin was easy to game... it did because it had a great business model.

It sold older tech and pre-sold hardware that wasn't even ready yet, to fund further research. Great model if you think about it. Research Version 3, using funds collected to order and make Version 2, while operating with income from sales of Version 1. It ensured Bitmain was not only doing corporate mining, but that almost any other independent miner somehow fattened Bitmain.

[DooMAD]

Also food for thought. Bitmain didn't really get to where it was because it had superior tech or because Bitcoin was easy to game... it did because it had a great business model.

It didn't even have to be a "great" business model, it just had to be better than all the sketchy hardware vendors out there who defrauded their customers.  Companies like BFL and GAW set the bar pretty low, in fairness.  If Bitmain have a dominant position in the market, it's because the position wasn't really contested back in the day.  Things are improving now, though.  

The importance of strong competition can't be underestimated.  We could use sport as an analogy.  I got a silver medal in a Judo tournament once.  That's not a brag, since there were a total of two people in that particular weight division   .  Suffice to say, I wasn't very good at Judo, so it wasn't difficult for the other guy to take 1st place.  But if there had been a greater number of skilled competitors, it might not have been as easy for them to win.  And obviously, I didn't earn that second place.  Someone more deserving would have taken it if there had been more participants.

Bitmain, similarly, have had a pretty easy ride until recently.  They didn't have any real competitors to go up against.  Hopefully, as time goes by, there will be a greater number of successful companies fighting for market share.

[aliashraf]

My non-technical logic says: we apparently have historical evidence that, given enough motivation and resources (shall we say, Bitmains appetite and technician workforce as a prime iteration of this), there will always be ability to develop a piece of specialised hardware that will deliver what an ASIC does.

And my technical logic says: it is not the case, imho.  

First of all 'historical evidence' shows otherwise: Ethash was fully implemented in 2015 and used to mine billions of dollars worth of crypto in 2016-2018 and ASICs never managed to take over Eth mining scene in spite of high incentives and Bitmain's almost unlimited resources. All they could manage to gain was less than 2x advantage against rigs built of mid-range commodity gpus.

Most importantly, ASIC design is no magic, you need to break down an algorithm to giant instructions and implement them on die besides conventional pipelining and Single Instruction Multiple Data techniques to accelerate parallelization. When an algorithm happens to be truly memory hard, no matter how smart you are, you never manage to break it down to large enough macro instructions because they need to halt and fetch memory for continuation. I've discussed this issues above-thread in more details.

So I see anti-ASIC as ASIC-resistant, not ASIC proof. Fully cognisant that there can be no ASIC proof.

This is what Bitmain wants us to believe and unfortunately is adopted by many community members but it doesn't change anything about its fallacy.

Firstly, by ASIC proof we mean an algorithm being ASIC resistant enough to de-incentivize special hardware designers and keep them out of the mining business. This is it! We are not discussing in a pure scientific context, it is game theory and costs against incentives.

Secondly:
Every cpu/gpu on the market could easily be called an ASIC! GPUs are ASICs, definitively! They use a lot of specialized circuits for many jobs , so what the heck?

What makes an ASIC miner bad is its status as a specialized device made by a company that is targeting mining industry and nothing else. To understand why it is bad you need to read following paragraphs very carefully.

For a company to be a legitimate player in a market, besides anything else, it needs to offer a 'utility' to the value chain which it is present in, eventually. Innovations should help people (other entities present in the chain) to do their job more efficiently in the first place. Innovation is not about competition between consumers it is about competition between producers.

When Intel produces a chip with higher MIPs and more reasonable prices while competing with AMD, it is offering a utility to its customers, this is how  and why competition is justified as a good thing for consumers and the society as a whole. It is how capitalism is justified and is tolerable in spite of lot of problems and paradoxes.

Bitcoin mining but is an exceptional industry. There is no need for efficiency, actually bitcoin mining is designed to be inefficient. When companies offer more efficient miners they are not doing anything good to the industry or human being or the planet. Miners don't find themselves happier when new ASICs are announced, on the contrary a new ASIC is always bad news for professional miners because they need to pay a lot of money to the manufacturer without any improvement in their incomes, even worse with a long period of pain and stress because of the consequent rapid difficulty growth before they could manage to liquidate their assets, take loans, ... to pay manufacturer for new generation of miners.

ASIC manufacturing for bitcoin is a vanity fair, a black hole that sucks coins from miners and does nothing good for them, a serious centralization threat because of its privileged access to large hash powers.
Bitcoin should get rid of this situation, it leads us to nowhere, sticking with crackers and people who do nothing useful and still suck last coins from miners.  

[Wind_FURY]

My non-technical logic says: we apparently have historical evidence that, given enough motivation and resources (shall we say, Bitmains appetite and technician workforce as a prime iteration of this), there will always be ability to develop a piece of specialised hardware that will deliver what an ASIC does.

And my technical logic says: it is not the case, imho.  

First of all 'historical evidence' shows otherwise: Ethash was fully implemented in 2015 and used to mine billions of dollars worth of crypto in 2016-2018 and ASICs never managed to take over Eth mining scene in spite of high incentives and Bitmain's almost unlimited resources. All they could manage to gain was less than 2x advantage against rigs built of mid-range commodity gpus.

What assurances are there that it won't improve? Or maybe Bitmain will not commit itself on Ethash because of Ethereum's planned switch to Proof of Stake.

I believe a better example would be Litecoin's use of Scrypt. Wasn't it used because it was supposed to be designed to be resistant to GPU, FPGA, and ASIC?

So I see anti-ASIC as ASIC-resistant, not ASIC proof. Fully cognisant that there can be no ASIC proof.

This is what Bitmain wants us to believe and unfortunately is adopted by many community members but it doesn't change anything about its fallacy.

Secondly:
Every cpu/gpu on the market could easily be called an ASIC! GPUs are ASICs, definitively! They use a lot of specialized circuits for many jobs , so what the heck?

Not in the context being discussed.

[aliashraf]

What assurances are there that it won't improve? Or maybe Bitmain will not commit itself on Ethash because of Ethereum's planned switch to Proof of Stake.

Technical assurances. You can't just do anything disruptive about memory bound algorithms with large footprints by ASICs. You need to improve the bottleneck, memory buss, and it is already pushed to its highest boundaries and out of the ASIC business scope.

I believe a better example would be Litecoin's use of Scrypt. Wasn't it used because it was supposed to be designed to be resistant to GPU, FPGA, and ASIC?

Scrypt is not a good example, they needed to keep the memory footprint small because otherwise the validation process would took too long and it is the basic weakness of this class of PoW algorithms. Scrypt didn't fail because of the magical power of Bitmain, it failed due to its inherent conceptual and design flaws.

[Wind_FURY]

What assurances are there that it won't improve? Or maybe Bitmain will not commit itself on Ethash because of Ethereum's planned switch to Proof of Stake.

Technical assurances. You can't just do anything disruptive about memory bound algorithms with large footprints by ASICs. You need to improve the bottleneck, memory buss, and it is already pushed to its highest boundaries and out of the ASIC business scope.

I believe a better example would be Litecoin's use of Scrypt. Wasn't it used because it was supposed to be designed to be resistant to GPU, FPGA, and ASIC?

Scrypt is not a good example, they needed to keep the memory footprint small because otherwise the validation process would took too long and it is the basic weakness of this class of PoW algorithms. Scrypt didn't fail because of the magical power of Bitmain, it failed due to its inherent conceptual and design flaws.

From what I have learned, those mining algorithms that were designed to use more RAM were designed to prevent GPU mining, but it doesn't prevent the development of specialized hardware for them.

Plus my point truly is not to debate about the technical. It's about the naiveté of some people believing that "this algorithm is ASIC-resistant", but history has shown us that nothing is truly resistant from specialized hardware development. Developers have tried and failed.

[aliashraf]

@WIND_FURY, you are rehashing the same argument ever and ever.

Neither any established technical argument nor historical events support your assumption/religious-belief-in-ASICs that "nothing is truly resistant to specialized hardware development", it is a myth, nothing more, no matter how many journalists/uneducated-bitcoiners have faith  in such a claim, it is just false.

[buwaytress]

First of all 'historical evidence' shows otherwise: Ethash was fully implemented in 2015 and used to mine billions of dollars worth of crypto in 2016-2018 and ASICs never managed to take over Eth mining scene in spite of high incentives and Bitmain's almost unlimited resources. All they could manage to gain was less than 2x advantage against rigs built of mid-range commodity gpus.

Most importantly, ASIC design is no magic, you need to break down an algorithm to giant instructions and implement them on die besides conventional pipelining and Single Instruction Multiple Data techniques to accelerate parallelization. When an algorithm happens to be truly memory hard, no matter how smart you are, you never manage to break it down to large enough macro instructions because they need to halt and fetch memory for continuation. I've discussed this issues above-thread in more details.

I suppose I have to admit that I know very little deep understanding of Bitcoin so cannot offer but the most superficial comments for anything outside of Bitcoin, but I wonder if that's part of it? Ethereum miners very seldom are as well serious Bitcoin miners (in my understanding and vice versa) so it was simply to me a case of deciding to focus resources on Bitcoin vs an alt people like me still need convincing about.

And with that lessened priority, to gain still a 2x advantage seems a lot to me!

Say if we took as much interest in mining Ether as we did Bitcoin, mayhaps these mid-range miners would very quickly find themselves edged out? Maybe we'll never find out with their impending algo switch.

Bitcoin mining but is an exceptional industry. There is no need for efficiency, actually bitcoin mining is designed to be inefficient. When companies offer more efficient miners they are not doing anything good to the industry or human being or the planet. Miners don't find themselves happier when new ASICs are announced, on the contrary a new ASIC is always bad news for professional miners because they need to pay a lot of money to the manufacturer without any improvement in their incomes, even worse with a long period of pain and stress because of the consequent rapid difficulty growth before they could manage to liquidate their assets, take loans, ... to pay manufacturer for new generation of miners.

I don't think anyone disagrees! But I've always seen ASIS resistance as directly attacking Bitmain and their ilk and maybe that's simply not necessary because Bitmain's model, or any chip manufacturer model for that matter, cannot possibly sustain because of eventual inefficient mining. We seem to be witnessing this unfolding even now.

[kano]

@WIND_FURY

We have been thru this for a while, your assumption about ASICs-eventually-win is false and an advertised hype. There is no way to make FETCH_MEM operations to be ASIC friendly, no scientific way, not speaking of magical and mysterious powers of Bitmain  

Above-thread I've extensively discussed why and how ProgPow is provably safe against ASIC optimizations higher than 1.2x and I've provided supplementary resources as well, why don't you bother researching a bit more instead of rehashing Bitmain ads ever and ever?

You are missing a rather obvious point ...

Some recent miners cost as much as $1000 or more.
So, what stops someone making an 'ASIC' that has the necessary hardware to match some other PoW similar to CPU+RAM+GPU
Of course it would have aspects similar to a full computer, but would end up costing less, and well, that's not a sudden jump in cost or price vs current Bitcoin ASIC miners.
Edit: and anyway, all ASIC miners have a small computer in them already anyway.

--

FYI this is the Bitcoin part of the forum.

[kano]

What assurances are there that it won't improve? Or maybe Bitmain will not commit itself on Ethash because of Ethereum's planned switch to Proof of Stake.

Technical assurances. You can't just do anything disruptive about memory bound algorithms with large footprints by ASICs. You need to improve the bottleneck, memory buss, and it is already pushed to its highest boundaries and out of the ASIC business scope.

I believe a better example would be Litecoin's use of Scrypt. Wasn't it used because it was supposed to be designed to be resistant to GPU, FPGA, and ASIC?

Scrypt is not a good example, they needed to keep the memory footprint small because otherwise the validation process would took too long and it is the basic weakness of this class of PoW algorithms. Scrypt didn't fail because of the magical power of Bitmain, it failed due to its inherent conceptual and design flaws.

Sorry, that's false information.

Scrypt was kept small purposefully by Charlie Lee so that he could GPU mine it while claiming it could only be CPU mined.
Go read the Litecoin thread (coblee is Charlie Lee)
I even made comment early on about the scrypt settings being too low and should fit in a GPU, thus the CPU-only claim being incorrect.

He did this for a very long time.
Then finally the GPU code, he'd been using all along, was released.

Litecoin using scrypt was a scam by design by Charlie Lee.

[Wind_FURY]

@WIND_FURY, you are rehashing the same argument ever and ever.

[b_2Neither any established technical argument nor historical events support your assumption/religious-belief-in-ASICs that "nothing is truly resistant to specialized hardware development", it is a myth, nothing more[/b], no matter how many journalists/uneducated-bitcoiners have faith  in such a claim, it is just false.

What? Hahaha. History has proven of just the opposite of what you said. ASIC-resistance IS a myth. Once a cryptocurrency becomes successful enough, there will always be dedicated hardware made for it because it's simply profitable for some companies to do it.

[aliashraf]

What assurances are there that it won't improve? Or maybe Bitmain will not commit itself on Ethash because of Ethereum's planned switch to Proof of Stake.

Technical assurances. You can't just do anything disruptive about memory bound algorithms with large footprints by ASICs. You need to improve the bottleneck, memory buss, and it is already pushed to its highest boundaries and out of the ASIC business scope.

I believe a better example would be Litecoin's use of Scrypt. Wasn't it used because it was supposed to be designed to be resistant to GPU, FPGA, and ASIC?

Scrypt is not a good example, they needed to keep the memory footprint small because otherwise the validation process would took too long and it is the basic weakness of this class of PoW algorithms. Scrypt didn't fail because of the magical power of Bitmain, it failed due to its inherent conceptual and design flaws.

Sorry, that's false information.

Scrypt was kept small purposefully by Charlie Lee so that he could GPU mine it while claiming it could only be CPU mined.
Go read the Litecoin thread (coblee is Charlie Lee)
I even made comment early on about the scrypt settings being too low and should fit in a GPU, thus the CPU-only claim being incorrect.

He did this for a very long time.
Then finally the GPU code, he'd been using all along, was released.

Litecoin using scrypt was a scam by design by Charlie Lee.

Thanks for the comment. I Haven't followed Litecoin and Charlie Lee story in details and Lee being such a douchebag doesn't surprise me.

Still Scrypt is not Litecoin and it suffers from a more general problem:

source: https://medium.com/shokone/hash-no-not-that-kind-the-crypto-kind-2e8bf616aa24
All of the different variables of the Scrypt algorithm are designed to prevent the kind of centralization and ASIC development we have seen with bitcoin. Unfortunately, making the function extremely memory intensive is not efficient when it comes to the verification of computed proofs in blockchain systems. Therefore, cryptocurrencies like Litecoin, have simplified their Scrypt implementations to maintain a lower rate of memory usage. Due to this simplification, Scrypt mining ASICs reentered the picture, and now we are back to where we started with centralization and hash power monopolization.

[Wind_FURY]

I believe Charlie Lee used Scrypt because of Tenebrix's claims of its resistance to GPU, FPGA, and ASICs? How was it Charlie Lee's fault that Scrypt was not resistant to the development, and creation of specialized hardware for it?

It implements a proof-of-work scheme based on scrypt, a cryptographic construct specifically designed to resist creation of efficient GPU, FPGA and even ASIC implementations. You can find more about scrypt and how it can improve your stamina, masculine appeal and performance here

https://bitcointalk.org/index.php?topic=45667.0

We really liked Tenebrix's Scrypt proof of work.  Using Scrypt allows one to mine Litecoin while also mining Bitcoin.  We humbly offer a big thanks to ArtForz for the implementation.

https://bitcointalk.org/index.php?topic=47417.0

[tromp]

I even made comment early on about the scrypt settings being too low and should fit in a GPU, thus the CPU-only claim being incorrect.

How much memory should scrypt have required in order not to fit on GPU?
And how long would it take to verify then?

[stortz]

the real reason behind ProgPoW is to eliminate bad actors from Bitmain
It's much more accessible to make an 'arms race' using GPU than having to rely on buying chips from China

IMO this will be an inevitable step on bitcoin's road map
Ethereum and other coins might be the first to experiment with, but it will eventually make it's way into bitcoin's code

[aliashraf]

@WIND_FURY

We have been thru this for a while, your assumption about ASICs-eventually-win is false and an advertised hype. There is no way to make FETCH_MEM operations to be ASIC friendly, no scientific way, not speaking of magical and mysterious powers of Bitmain  

Above-thread I've extensively discussed why and how ProgPow is provably safe against ASIC optimizations higher than 1.2x and I've provided supplementary resources as well, why don't you bother researching a bit more instead of rehashing Bitmain ads ever and ever?

You are missing a rather obvious point ...

Some recent miners cost as much as $1000 or more.
So, what stops someone making an 'ASIC' that has the necessary hardware to match some other PoW similar to CPU+RAM+GPU
Of course it would have aspects similar to a full computer, but would end up costing less, and well, that's not a sudden jump in cost or price vs current Bitcoin ASIC miners.

Sorry,but I overlooked this post. Your argument is wrong, Bitmain just can't make a more specialized thing for ProgPow without ending in a complete gpu technology development project and competing with AMD, and Nvidia it is the same for RAM and memory bus architecture.

FYI this is the Bitcoin part of the forum.

Good to know   but who told you bitcoin is SHA256?