The first 2 years of planned hardforks are all the time we have for tweaking an asic resistant pow. So we better use that time to let some healthy ASIC competition evolve.
Monero differs in that they’re happy to keep on hardforking for many years, while Grin feels that is too much of a centralizing force.
asic manufacturers splitting the bitcoin community and bitmain now reportedly holding over 1 million bch are not centralizing forces?
the problem facing all of cryptocurrency is how to create a fair distribution of grin coin rather than concentrating it among those with access to silicon fabrication or wealth.
Wealthy people will always have the ability to get a larger share of any cryptocurrency. That is not a fairness problem.
And you’re right most grin owners will not have mined their grin; they will instead have bought it. The fairness comes from no-one being able to get grin vastly cheaper than what it sells for on the open market.
The decisions should be made early, very early; changing direction when herding cats is very very hard.
When money gets involved, your herding catniped cats; something like segwit starts a cultist offshoot and a few months of backtracking on adoption.
@tromp, You’re suggesting that ASIC development is less of a centralizing force than hard forking? I’d say hard forks are the beauty of decentralized crypto, as the community is free to reject any fork it deems unnecessary. The community is also free to fork the project away from leadership with questionable or misguided intentions.
@monkyyy, you just have to make sure you’re looking out for the “cats” in your analogy (and not the herders), otherwise you might be in for a big surprise.
Also:
Why do we only have 2 years? What happens after that?
Further changes to Grin will be attempted as soft-forks (that need majority miner support).
Hardforks will be reserved for critical bugfixes (hopefully exceedingly rare) or changes that cannot reasonably be implemented as soft-forks, the latter requiring consensus across the Grin community.
I agree that more than 2 years of scheduled AR hard forks would be better, and not even close to the level of centralization as secret asics, so long as it’s decided ASAP and echoed regularly as the plan …
The known and scheduled POW transition is just about as smart as it gets is trying to level the playing field. ASICs are a net security/decentralization good if they’re cheap and accessible (aka lots of competition).
So exciting to design in this economy and its adolescence… or toddlerhood? hard to know.
Are there numbers for how extreme asics improvement will be for cuckoo?
Without an estimate of some kind couldn’t 6 months plan be a disaster or completely pointless?
Like lets say that grin does well for 6 months, and asic’s are exterme(where bitcoin had 150,000x; grin has 100,000,000x improvement) if its at such an extreme scale it could be still more then profitable to design an asic during the 2nd half that will work for the 3rd and 4th. And well the 6 month plan recreated bch fork attack
Alternatively, if asic’s are much much weaker; the 6 month plan makes asic manufactures yawn at the 6 month return and they just don’t brother, the open source asic doesn’t have competition meaning its not actually a source of stability for investors as it may just be shit and the asic races are just being delayed.
I think you misunderstand. GPU frindly PoW has not been selected yet, it won’t be cuckoo/cuckatoo as that is the ASIC PoW. Of course it can be just modified a bit, but those changes can range from not affecting asic at all to slowing them by factor of 100x.
Also Bitcoin has _000x efficiency gains over GPUs, not 150000x. I would estimate raw Cuckatoo at _00x.
_000x efficiency gains over GPUs
If you exclude xeons phi (super computer hardware that is 50x better then other gpu seems out of place to me, where they actually used?) its 10,000x
That number I gave was low end cpu to modern asic from my earlier post.
I would estimate raw Cuckatoo at _00x
Why? And which gpu are we talking, bargain bin hobbist, or the super computer weirdness, which is its own factor of 1000x in bitcoin and I assume somewhat applies here
Antiminer S9 is 14 000x faster compared to a GPU, but only 1000x more efficient. You can’t compare a multichip (189 chips) 1400W box with single chip gpus, otherwise you can just buy and hot-glue 10 S9s together and call it 140 000x “better” compared to a gpu.
Another example is Antminer E3 being 6x faster than a gpu, but just as efficient when it comes to hashes per unit of energy.
The newly announced Linzhi miner for ethash (see linzhi.io)
is about 6x more efficient than GPUs though.
Talking past each other on what happened in bitcoins asic race doesn’t seem worth either of our time.
Why do you give an estimate of 100x and can you define what you mean by that number a bit more concretely? low end hobbist vs professional miner favorite gpus, first gen asic or theoretical best? What methodology?
This is an interesting point, and makes me think the solution being taken (supporting two PoW algorithms, one favoring GPUs, one favoring ASICs but not so much as to allow for single chip designs and 100,000x efficiency improvements, but not so little as to discourage ASIC development altogether, while slowly phasing out the GPU algo over time) is much more complicated than simply resisting ASICs, or at the very least, minimizing their efficiency gain.
What am I missing?
If SHA-256 would be used, would that put Grin subject to attacks ?
Since so much hashing power is already deployed for Bitcoin, just a small fraction of it would be huge for Grin…
True, but what I meant to say is I didn’t see really much difference in whether to implement Cuckatoo and expect ASICs for it to emerge or just to use the existing SHA256. Grin is going to increase ASIC-friendly algo mining profit gradually, to make sure noone breaks into the network with huuuuge mining power suddenly and takes control over the whole network.
Bu then I thought of it deeper. Perhaps it makes sense, to be honest. Cuckatoo ASIC production will take time and it will make ASICs entry much smoother than it would be with existing SHA256. Even if Grin dev team made SHA256 PoW and limited it to be 99% less profitable than second ASIC-resistant algo gradually increasing profitability, it would still be dangerous. It could happen that noone would mine Grin with SHA256 for years and then someone pups the coin price and the huge hashpower switches to mining Grin - it would be no good at all.
However, the other thing I wanted to say is my deep and steady concern that there won’t be just one or two, or five powerful ASIC manufacturers like for Bticoin SHA256 - Titmain, Titfury, who will split the whole Grin’s hashrate between them and take over the network.
The purpose - is to have a lot of competing manufacturers that would result in more decentralization, but I’m still not seeing why the odds of it happening are greater than the chances of Titmains centralizing ASIC production and the whole Grin network as the result.
I think we can all agree ASIC safe algo does not exist. So I think Grin should not even try but use SHA256.
Cryptocurrency mining ASIC is useless piece of silicon born to do useless job.
No, the job is security.
To mine with CPU or GPU is a waste of electricity and less secure even if the distribution can be better, but really there is no proof that GPU owners are better distributed than ASIC owners.
The real discussion we should have before release is inflation and supply, are these decided? I think a fixed supply is the only interesting option.
So there you have it, SHA256 and fixed supply and you will get my mining (in the winter)! No brainer.
Fixed supply is very boring option, IMO.
After some time spent on thinking I must agree that it’s wise enough to create PoW as simple as possible. But only because there is no provable way to invent a PoW that would keep itself in line with widespread commodity hardware forever and because tweaking PoW every N years - is not an elegant solution, altgough, quite feasible as for me.
Regarding PoW simplicity - yesterday I read that Intel had embedded FPGA into one of their Xeons some time ago, although it’s not available for retail yet. So yes… Chips are/willbe cheap. In some time everyone will be able to be an owner of decent FPGA with quite fast RAM access. It means even dynamic PoWs are doomed. Not sure about such as RandomJS, though, but time will tell.
But as for now:
Dude, are you serious?
As for SHA256: There are huge number of SHA256 miners in the world now. Doesn’t it mean for you that accommodation of SHA256 as a PoW makes chances of taking over the network and 51% attacks a lot higher than if it would be new PoW, ASICs for which aren’t printed yet?
I think a fixed supply is the only interesting option.
Why?