A decision was made to stick with cuckaroo29 for GPU mining.
That currently requires 5.5 GB for efficient mining.
Duplicating all hash computations should allow for a reduction to 4 GB,
but there are no concrete plans to implement such a miner.
What’s the rationale behind the decision, when clearly there aren’t any drawbacks? IMO it’s a very unfortunate decision to actively exclude hundreds of thousands of GPUs that would be perfectly capable of mining while not even reducing security of Grin PoW (integrity meaning ASIC hardness).
IMO Grin is doing itself a disfavour, actively reducing it’s potential userbase and network growth.
We briefly considered doing cuckaroo29+ or cuckaroo28+ but decided the extra complexity was not worth it.
By being efficiently mineable only by 6+ GB cards, we increase profitability of such cards, and attract a higher fraction of them, leaving fewer of them available for renting through nicehash to join a 51% attack.
I think the majority of small time miners went with 4GB GPUs, the big corporate mines all made the switch to 8GB when GPU prices were high, I watched many of them being sold in massive quantities (I’m talking 50’000 RX 570 4 GB here). I was asking why sell them and many times the answer was along the lines of “we’re upgrading, 4 GB won’t last”.
So all you’ll have by keeping cuckaroo29 is big corporate mines that did the change in time and saw this coming, like Genesis Mining and Core Scientific.
Sure you’ll also have some hobbyist 1080 Ti farms but the big chunk is on lower end GPUs.
Is there a github for cuckaroo28+/cuckaroo29+?
1 Like
Let my clarify my FPGA comment. Existing boards won’t accelerate Roo29 much. That’s all I looked at. Custom FPGA boards and/or Too32 is an open question.
There may be a similar effect with ASIC’s. Too32 may be large enough that only those with access to the 7nm processes at TSMC have a chance. I haven’t done enough research on our ASIC plans yet… still focused on GPU’s for the moment.
In my opinion, the biggest threat to Cuckoo Cycle is algorithmic improvements, and you can’t plan for that.
Certainly 4GB is achievable. With your own Mean Miner, for example, if you just removed the 7 partition bits from U, you need only 51 bits per edge, or about 3.4GB dense. Maybe some extra room for partition unevenness, but it’s less than you think.
@timolson thanks for your thoughts and expertise on this.
What would be the minimum GB requirement of cuckaroo28+ and cuckaroo29+ respectively?
I wasn’t able to find anything neither on cuckaroo28+ nor on cuckaroo29+ (the only post in this forum is the one above that @tromp made)
i.e. what would be needed to make it work on 2 GB cards?
plugin cuckaroo_cpu_avx2_29 uses well under 4 GB. but its data layout is poorly suited to GPUs. https://forum.grin.mw/t/gpu-mean-memory-reductions offers a better strategy to run on a 4 GB GPU.
And obviously there are conflicts of interest in every of these discussions as there most likely is always money behind the argumentation (also behind mine as I stated above).
It was long thought that 6GB GPUs would be cut off and now even 4GB GPUs can participate. I would say just open it up to all modern GPUs that have been made in the past 4-5 years and allow for an even playing field. This also shines an unpartisan light on Grin and invites many more people to the project, I’d say that’s a win-win.
Dan Saltman @dbl_twitter 08:08
so, to be clear- gone is the mantra that this coin would only be mineable by people with high end cards and gamers.John Tromp @tromp 08:08
@engmsf no delay. fairness requires optimal miners public at launchDan Saltman @dbl_twitter 08:09
1060s is one thing, opening it up to 4gb cards is another ball park
Allowing only a certain class of GPUs to participate (in this case very high end) creates unfairness, tensions and likely profits big corporations mostly.
Sooner or later somebody will even figure out a secret way on how to compress it down to 3GB GPUs and those will then also have an unfair advantage.
So best is to go down to 1 GB minimum, as there are not any GPUs that I know of in the last 4-5 years that had less than that in VRAM and let it clash, an equilibrium will form and nobody will have good reason to complain.
The more GPU hashpower is protecting Grin (and the more diverse it is), the more secure its network becomes, so it makes no sense to exclude certain GPU classes, if the exclusion does only serve the purpose of high-end cards being able to run more profitably, that’s actively siding with one side. There has not been an argument made that a potential ASIC would benefit more from a 1 GB implementation vs. a 4GB implementation in terms of efficiency gains in implementation 1 vs. 2.
And to quote you on your previous statement with regards to distribution of GPUs with regards to their “high-end-ness” and VRAM:
There is another potentially large benefit to requiring more than 6GB. Many huge mining farms are dominated by <= 6GB GPUs. Excluding them makes it easier for hobby miners to compete.
In my opinion it’s the exact opposite, as big farms have anticipated this development. Also if I was still a hardcore gamer there’s no way I’d abuse my precious high end card for straining cryptocurrency mining but maybe that’s just me 
Have you taken a look at the link above showing real-time stats of all the GPUs that run on Ethos?
Offtopic: I love the design of this forum and it’s high speed and good interactivity. Is there a template of this somewhere on github or is this a proprietary solution?
1 Like
Mining algos that can attract large GPU farms is a huge risk.
Look at what happened to Bitcoin Cash and the recent hashwar… majority hashrate was influenced by multiple wealthy individuals in a very short period of time.
The hash thrown at Bitcoin Cash was almost certainly temporarily diverted (rented) away from the larger sha256 coin Bitcoin. For coins where a ‘larger’ coin also exists using the same algo - this changes the economics of Nakamoto consensus… hash can be ‘rented’ from a larger coin to achieve a short term means.
Bitcoin itself does not have this issue. Bitcoin has 90% of the sha256 miners. There is no hash to rent. Essentially, one algo… one coin.
Requiring 6GB+ gpus is an excellent idea. There are too many large farms with GPUs < 6GB. It is not about giving the hobby gamer a ‘chance’, but rather stopping the hash rent market from coming in.
If anything, the 2 year time frame should be algorithmically adjusted based on growth ratio between gpu/asic algo difficulty adjustments.
Most coins are in a state of flux. Most GPU coins are susceptible to hash rent attacks depending on market value. Bitcoin does not have this problem.
To clarify, simple mass produced commodity ASICs of a specific algo where Grin has majority share would be the best possible situation.
3 Likes
it seems it’s discourse pure vanilla.
1 Like
@stsbx I disagree with on your argumentation.
Comparing SHA256 hashrate of one coin with SHA256 of another coin that defined itself as the one that wants to replace the former, I expect stuff to get a little shady.
A better analogy would be Ethereum and Monero. Both are GPU-mineable but Ethereum has about 10-15x more GPU hashpower than Monero. However, we have never seen a malicious attack from Ethereum GPUs on Monero. Besides there’s also no incentive to do so because by destroying the value of one coin they cause migration of more GPUs to Ethereum, therefore lowering their own revenue long term.
There are a ton of 8GB+ GPUs on Ethereum and in the future more will be added as the DAG increases, so if Grin doesn’t attract an instant $15B market cap it will be a much smaller network and would potentially be vulnerable to large mega-scale Ethereum farms.
Besides, as already mentioned in this thread, it’s already attainable to mine Grin with 4GB GPUs with a 20% loss in SPS. However, I’m very confident somebody will find a (secret) way to mine without a SPS loss on 4GB GPUs, suddendly somebody will find a way to (secretly) mine it on 3GB GPUs and so on…
So the best and most neutral course of action is to allow an even playing field for all modern GPUs and I’d say lowering the requirement to 1GB would do just that.
It would not make it any easier for ASICs either.
The only thing a high requirement for VRAM leads to is concentration of profits into megafarms that have already seen the move to 8GB coming in 2016/2017, aka Gensis Mining, to name one.
Allowing everyone to participate is the only fair and unbiased solution in my opinion.
1 Like
However, we have never seen a malicious attack from Ethereum GPUs on Monero.
Isn’t that a touch disingenuous?
@monkyyy 1. That article has nothing to do with what I said. It’s not talking about an attack a majority network (Ethereum) attacking the minority network (Monero).
CPUs and botnets have their firm place in Monero as it currently stands.
- Grin will not have that “problem” (if you want to consider it one), as it’s no CPU mineable.
>botnets have their firm place in Monero as it currently stands.
The line between agorism and illegalism is somewhat fuzzy but can I suggest its an important one.
Botnet mining is harmful to a coin as its making the game more zero sum and social trust is weakened, not to mention some of the theory of pow is damaged and given its immaturity as a theory of money who the fuck knows if its doing meaningful damage inbetween the noise of this market
And I’m not advocating for botnet mining. I said create an even playing field across all GPUs because the current way it’s governed creates an environment of secrecy where low VRAM mining will happen anyway on large scale farms but will exclude hobbyist people.
I’m focusing on the implied claim that xmr doesn’t have mining attacks, when in fact has had an ongoing one for years; its one of the reasons I consider the project a failure and been avoiding it.
I may be stretching the term a bit, but its still an undesirable behavior that allowed by the flaws of the protocol
That is assuming the attackers goal is revenue. Hash rent markets allow bad actors to pursue other nefarious purposes over a short period… e.g., influence forking/upgrade mechanisms. Nakamoto Consensus was supposed to keep this type of behaviour out. Attacking with rented hash over the short term is orders of magnitudes cheaper and easier than acquiring equipment, securing real estate, power etc.
This can’t happen to Bitcoin as it has 90% of all sha256 ASICs in the world. There is no hash to rent.
ETH is competing for gpus from other coins also. Also susceptible to hash rent market. Less so while market cap is high.
Gpu mining long term is a joke. Highly gameable and subject to secret efficiencies as mentioned by yourself.
ASICs require effort. The status quo is not enough. Even money is not enough. Building ASICs requires ingenuity, entrepreneurship, capital and having a level head. Even the largest miner builder in recent years it seems is experiencing issues as they may have strayed too far from their original goal - building ASICs.
ASICs FTW. Do whatever is necessary to keep hash rent market out. The process is fuzzy though and not perfect.
1 Like