Achieving ASIC Resistance Using Economics

[liberabit]

All anti-ASIC algorithms have failed or will fail to prevent the development of specialized chips.

Instead of creating complex algorithms, I propose creating a new cryptocurrency that adapts the reward based on difficulty or the number of leading zeros in the hash.

The more zeros, the lower the reward. This creates a dynamic where miners can control the supply and favors more efficient machines over more powerful ones.

This approach might slightly affect security in the pursuit of greater decentralization.

Is this idea good? What do you think?

[Mia Chloe]

You point is a Little unclear to me . However bitcoin protocol already has rules that has a majority of the attributes you are trying to refer to.  Chips miners currently use are more efficient than regular computer CPU and GPU. Although they are unable to multitask, they are capable of carrying out a single task more effectively and faster than is what makes miners still being able to find valid hashes with the current difficulty level.

It was during the first for years of bitcoin's launch that GPU and CPU mining was possible because there was lower hashing difficulty. Anyways, op don't forget that although the Hashing difficulty is high, the bitcoin protocol was also created in such a way that if the valid hash is found too soon, the difficulty is increased to slowdown the mining process.
Bitcoin is decentralised enough I would say since everyone using the network has their percentage of control over the network.

[d5000]

"More zeros" means higher difficulty. So in your idea we have the following rule: the higher the difficulty, the lower the reward.

This has already been implemented in some old and small altcoins, however often combined with a PoS element, like in the case of Peercoin.

The idea itself does not look bad at the first glance, putting a kind of "top" on difficulty, but on the second glance it has more problems than advantages;

1)  if the coin loses traction, i.e. its price lowers, it becomes more inflationary because less miners are interested in it and so the difficulty lowers. And this can cause a "death spiral", i.e. the more inflationary it becomes, the more sell pressure from miners, lowering the price and again the difficulty until it approaches zero.

2) the concept would also mean an artificial ceiling for security. This means that if someone sees an opportunity to steal coins and manages to buy a large amount of hashrate, for example using a botnet, it may be vulnerable to attacks.

[HeRetiK]

The more zeros, the lower the reward. This creates a dynamic where miners can control the supply and favors more efficient machines over more powerful ones.

How would this improve ASIC resistance though? ASICs are more efficient as well as powerful, so in the end they'd remain the favored choice; especially given that the reward would be lowered for everyone, regardless of what hardware they mine on.

(assuming ASIC resistance is desirable in the first place)

[ABCbits]

All anti-ASIC algorithms have failed or will fail to prevent the development of specialized chips.

They certainly could try to discourage that by
1. Designing algorithm where such ASIC doesn't offer much advantage compared with CPU or GPU.
2. Periodically change parameter or detail of the algorithm, which makes previously designed ASIC useless.

The more zeros, the lower the reward. This creates a dynamic where miners can control the supply and favors more efficient machines over more powerful ones.

Miner generally already favor efficiency, where they periodically move to newer ASIC type (which typically offer better efficiency) and pay attention to electricity rate.

[MeGold666]

All anti-ASIC algorithms have failed or will fail to prevent the development of specialized chips.

RandomX has proven it is possible, no company in the world will create ASIC for RandomX because you would need almost a full blown CPU and you can't compete with Intel and AMD due to Economies of scale.

Even the so called "ASIC" that was published by Bitmain (X5) turned out to be just a bunch of regular RISC-V CPU's that were less profitable than even older AMD CPU's.

That's why we are still able to mine Monero with our home computers, like we could mine Bitcoin in the past.
With ASIC's on the network we have no chance mining using regular computers and this decreases decentralization because not many people are willing to buy specialized equipment that can do only one thing.

Permisonless system also suffer with ASIC's on the network because the only viable way to obtain coins is to buy from someone else which require permission from someone willing to sell you.
Seeing the increasing centralization of supply, this will lead to very permissible network:

https://bitcointalk.org/index.php?topic=5507809.msg64479357#msg64479357

Seems like Bitcoin is becoming a decentralized cryptocurrency bank that will require KYC in order to acquire and use Bitcoin.

Another problem is with importing ASIC's from China which are usually the most efficient, the transport costs alone are like ~50% (Import Tariffs,Shipping Costs,Customs Brokerage Fees,Sales Tax) and on top of that this companies mine for them self before releasing the hardware to the public only after developing newer ones and again mining then selling used hardware as has been proven many times and the very limited, short warranty on this hardware is ridiculous (not that you would send it back to China and pay all the transport fees again).

Monero community knows how important is to stay ASIC resistant due to this reasons and Bitcoin whitepaper reflects the idea:

The proof-of-work also solves the problem of determining representation in majority decision
making. If the majority were based on one-IP-address-one-vote, it could be subverted by anyone
able to allocate many IPs. Proof-of-work is essentially one-CPU-one-vote.

Source: https://bitcoin.org/bitcoin.pdf

[vjudeu]

All anti-ASIC algorithms have failed

Test networks, like testnet3 and testnet4, didn't fail at being "CPU friendly" (which is probably what people actually want: instead of preventing ASICs from being used, CPUs and ASICs co-exist in these test networks).

The more zeros, the lower the reward.

It should be reversed: you put more power, to protect the network, so you receive more coins for doing that. In this case, CPU miners can actually receive something, consolidate their dust for free, or even experiment with non-standard transactions.

RandomX has proven it is possible

It works, until it doesn't. How many times Monero hard-forked to a different hash function? There is a huge risk, that if the value of Monero will increase significantly (for example because the whole crypto will increase), then it may become profitable to create just another ASIC for RandomX, and then Monero will try hard-forking again.

you can't compete with Intel and AMD

I can see more and more people, trying to run x86 emulators on different architectures, for example M1 or M2 Apple chips. The world is no longer "x86-only for home PCs", people are trying to switch to other solutions, also because of some security bugs, and backdoors, like Intel Management Engine.

Another thing is that you don't have to execute x86 code, instruction-by-instruction. You can translate it from x86 into some other architecture, by using some kind of assembly transcompilers, and then run your code natively.

we have no chance mining using regular computers

That's just another proof, that people want "CPU friendly" consensus rules, and not "ASIC resistant". Currently, SHA-256 is not broken. And as long, as it is the case, CPUs and ASICs should co-exist.

[ABCbits]

All anti-ASIC algorithms have failed

Test networks, like testnet3 and testnet4, didn't fail at being "CPU friendly" (which is probably what people actually want: instead of preventing ASICs from being used, CPUs and ASICs co-exist in these test networks).

That only works until someone bring financial incentive to test network. But i certainty hope such people keep using testnet3 or their own "test" network.

[vjudeu]

That only works until someone bring financial incentive to test network.

Why? The last price for 1 tBTC from testnet3, is 0.00000235 BTC (at the time of writing). There is some financial incentive. And people are still using ASICs, instead of mining CPU-only blocks. Also note that people didn't get all coins from all halvings, even if there is a bigger incentive for that, than for mining honestly. If people would really want to get all testnet3 coins, in the most efficient possible way, then we would already have around 7 million blocks.

So, why it is not the case? Is there a bigger incentive in mining a single block per 10 or 20 minutes, and taking bigger transaction fees instead?

[ABCbits]

That only works until someone bring financial incentive to test network.

Why? The last price for 1 tBTC from testnet3, is 0.00000235 BTC (at the time of writing). There is some financial incentive. And people are still using ASICs, instead of mining CPU-only blocks. Also note that people didn't get all coins from all halvings, even if there is a bigger incentive for that, than for mining honestly. If people would really want to get all testnet3 coins, in the most efficient possible way, then we would already have around 7 million blocks.

So, why it is not the case? Is there a bigger incentive in mining a single block per 10 or 20 minutes, and taking bigger transaction fees instead?

It's more than price of testnet3 coin. There are other factors such as,

• Token/altcoin airdrop, which either request testnet3 coin[1] or making tons of people try their service on testnet[2]. I recall this makes more people beg for testnet3 coin or how to mine it.
• Certain people or group who give financial incentive to testnet3 attack Jameson Lopp after he demonstrate attack on testnet3 and threaten to do same on testnet4[3]. Who knows how far they'll go to maximize profit.

And while I don't know why those people don't abuse time warp and other similar attack, the moving average of hashrate on testnet3 has grown drastically in past few years[4].

[1] https://bitcointalk.org/index.php?topic=5286384.msg55522090#msg55522090
[2] https://bitcointalk.org/index.php?topic=5484836.msg63650336#msg63650336
[3] https://blog.lopp.net/griefing-bitcoin-testnet/
[4] https://mempool.space/testnet/graphs/mining/hashrate-difficulty#all

[NotATether]

RandomX has proven it is possible, no company in the world will create ASIC for RandomX because you would need almost a full blown CPU and you can't compete with Intel and AMD due to Economies of scale.

Even the so called "ASIC" that was published by Bitmain (X5) turned out to be just a bunch of regular RISC-V CPU's that were less profitable than even older AMD CPU's.

That's not even an ASIC, it's a bunch of general-purpose CPUs.

It's not only that if you made a CPU like Intel or AMD it would be almost impossible unless you have billions of dollars, but it's also the fact that even if you do make a general-purpose CPU, and you optimize it for mining a single algorithm, is that really a CPU anymore? The more you optimize something into an ASIC, the easier it can be thwarted by creating a new algorithm.

[MeGold666]

It works, until it doesn't. How many times Monero hard-forked to a different hash function? There is a huge risk, that if the value of Monero will increase significantly (for example because the whole crypto will increase), then it may become profitable to create just another ASIC for RandomX, and then Monero will try hard-forking again.

Monero has history of hard-forking to fight off centralization that ASIC's bring along but RandomX is the final frontier and it's been working since 2019.

If you read more about RandomX you will understand there will never be a profitable way to build specialized hardware for it because economies of scale does not allow it.

Unless you think some company can compete with Intel and AMD and create more efficient and cheaper CPU  
And even if it did, it's not a problem because it would practically be general purpose CPU due to the RandomX requirements.
Competition is good, what is not good is a specialized hardware that can do only one thing more efficiently because it means nobody but miners will buy it and destroy profitability of everyone else which makes it more centralized.

We can see how it negatively impacted Bitcoin, it's not only hashrate that counts, but also decentralization and there's nothing better than being able to fairly mine on regular PC without the need for some expensive hardware that can't do anything else.
It's also much more green if you care about that.

That's just another proof, that people want "CPU friendly" consensus rules, and not "ASIC resistant". Currently, SHA-256 is not broken. And as long, as it is the case, CPUs and ASICs should co-exist.

Please, be my guest, go ahead and mine Bitcoin with CPU  
It is "possible" but it's not fair and practically impossible. They can co-exist the same way small fish can with a shark in a small tank.

I can see more and more people, trying to run x86 emulators on different architectures, for example M1 or M2 Apple chips. The world is no longer "x86-only for home PCs", people are trying to switch to other solutions, also because of some security bugs, and backdoors, like Intel Management Engine.

Emulation will never be efficient enough to compete, otherwise RandomX would be mined with FPGA.
If some other big chip manufacturer comes along that will be more efficient, it will surely be available for everyone, anywhere in the world and it will be general purpose and not some 1-thing-only brick.
For sure it won't be Bitmain  

This conversation we have here is because you never did appropriate research about RandomX, it's a shame I have to repeat my self over and over on something that should be obvious.
Leaving this thread as there is nothing else to say.

[vjudeu]

Unless you think some company can compete with Intel and AMD and create more efficient and cheaper CPU

It is not that unreasonable assumption. For example, it happened in 2020: https://www.imore.com/tests-show-apples-m1-emulates-x86-faster-intel-can-run-it-natively

Also note, that x86 in general is quite bloated. If you download some official PDF from Intel, for example there, it has thousands of pages: https://www.intel.com/content/www/us/en/developer/articles/technical/intel-sdm.html

And you probably don't need the full CPU functionality for mining, but only a small subset of that. Which also means, that an ASIC could be created, given enough incentive. Also, you probably don't need things like "A20 line" or "segment registers" for mining. Because maybe you don't know, but x86 don't have 16-bit, 32-bit and 64-bit addressing, but instead it has 20-bit, 36-bit, and 68-bit respectively, strictly because of segmentation (address=segment*16+offset), which cannot be turned off, because of backward compatibility. And there are more such things, slowing CPUs down, which is why some companies can compete with x86, if they simply disable some legacy code, which they don't need to handle.

Edit:

Please, be my guest, go ahead and mine Bitcoin with CPU

Of course I am mining testnet coins with my CPU. The 20 minute rule works nicely: blocks are always first-seen-safe, so they are not affected by things like full-RBF. To reorg a single CPU-mined block, you need two blocks. And there is also a two hours limit, which means, that only six future blocks can be created at a given time. And then, everything is left to network propagation.

Also, if your mining is more complicated, then it affects verification: if you have a complex hash function, then verifying the chain is hard, and sometimes verifying N blocks of the chain is as hard, as mining a new block. And at that point, people stop caring about verification, because it is more profitable to mine things without verifying them, when it requires comparable resources.

[MeGold666]

Unless you think some company can compete with Intel and AMD and create more efficient and cheaper CPU

It is not that unreasonable assumption. For example, it happened in 2020: https://www.imore.com/tests-show-apples-m1-emulates-x86-faster-intel-can-run-it-natively

Also note, that x86 in general is quite bloated. If you download some official PDF from Intel, for example there, it has thousands of pages: https://www.intel.com/content/www/us/en/developer/articles/technical/intel-sdm.html

And you probably don't need the full CPU functionality for mining, but only a small subset of that. Which also means, that an ASIC could be created, given enough incentive. Also, you probably don't need things like "A20 line" or "segment registers" for mining. Because maybe you don't know, but x86 don't have 16-bit, 32-bit and 64-bit addressing, but instead it has 20-bit, 36-bit, and 68-bit respectively, strictly because of segmentation (address=segment*16+offset), which cannot be turned off, because of backward compatibility. And there are more such things, slowing CPUs down, which is why some companies can compete with x86, if they simply disable some legacy code, which they don't need to handle.

I have been programming in Assembly for past 25 years... if you still think creating an actual ASIC is possible for RandomX, then you know nothing about RandomX nor about ASIC's.

RandomX is developed in a way it requires most of the regular CPU parts to be present, that's why it's not feasible for companies like Bitmain to create their own true RandomX specialized CPU.
You could cut down on some things but it still wouldn't be feasible to compete.

Again... I'm talking about obvious stuff and you are arguing with me because you didn't do any true research.

Over & Out.

Edit: ohh, and you are wrong about the addressing schemes. (a bit) 

[DediRock]

All anti-ASIC algorithms have failed or will fail to prevent the development of specialized chips.

Instead of creating complex algorithms, I propose creating a new cryptocurrency that adapts the reward based on difficulty or the number of leading zeros in the hash.

The more zeros, the lower the reward. This creates a dynamic where miners can control the supply and favors more efficient machines over more powerful ones.

This approach might slightly affect security in the pursuit of greater decentralization.

Is this idea good? What do you think?

Interesting approach! Balancing rewards with difficulty could incentivize efficiency and decentralization. However, it might be worth considering how this could impact long-term security and miner incentives.