The minimum plot size will be k=32

[hoffmang9]

All,

Though k=30 had been ruled out, it's now time to rule out k=31.

@wjblanke and I were able to get a k=31 phase 1 to complete in about 16 minutes on $4,000.00 equivalent hardware. That means that in two Moore's Law cycles of 18 months, a k=31 phase 1 arguably can be completed in 4 minutes. Since it's theoretically possible to farm at the end of phase 1, that's cutting it too close.

Someday the minimum k will go up from 32. However we'd like that to be more like 10 years from now rather than the 2-3 years that making k=31 minimum would have caused.

[marvic2409]

What was the hardware used in the testing?

[hasbrowow]

Bumping this 👍

[felicityc]

moore's law is dead.

it'll happen even faster.

[dorkmo]

im only doing 33s lol

[dipique]

Yes, it's dead, but we're now slower than Moore's law, not faster.

[vitiaaap29]

Hi @hoffmang9, excuse for troubling
I can select K = 25

Is it ok? Will my farming have a profit?

I use Windows 7, 4Gb RAM, 64-bit, processor: AMD A6-7310 APU with AMD Radeon R4 Graphics

[najchris11]

ok cool! thanks!

[stdiv]

plots with small k won't be accepted by mainnet

Привет! Подскажите плиз, есть ли сейчас смысл делать плоты более чем 32, за них награда будет начисляться? вроде говорили что у 32 и 33 разная вероятность выигрыша - так ли то? Спасибо

[davecarter]

The bigger the plot size the more time will it take to farm. As an advantage for the effort, it will not become obsolete very easily so it will be online longer.

[hoffmang9]

Larger plots do not take more time to farm. The only variable that matters is total space plotted.

[stdiv]

Tell me how the common space is determined, if I have 2 computers in different networks, full nodes are running (the chia application is running), everything works under one wallet, but they do not see each other's rafts, on one 100 on the other 30. Is it possible to do this and depends whether from this the likelihood of closing the block. Thank you
На русском: Подскажите как определяется общее пространство, если у меня 2 компьютера в разных сетях, запущены полные ноды ( приложение chia запущено) все работает под одним кошельком, но плоты друг друга они не видят, на одном 100 на другом 30. Можно ли так делать и зависит ли от этого вероятность закрыть блок. Спасибо

[Icecapital]

Помогите разобраться, для того что бы начать фармить на диске должно быть свободно 239 гигабайт для построения и в конечном итоге занято будет 101.4 ? Или достаточно просто иметь 101.4 гб для K =32, у меня есть свободные диски 240гб

[ilyakooo0]

239 для построения и 101.4 в итоге

[giobit]

239 для построения и 101.4 в итоге

для посева участка размером 101.4 гиб тибе нужно иметь на быстром носителе SSD типа М2 свободное пространство 239 гиб это для создания на нём кэша. например если утебя М2 512гб сможешь посеить только один участок и после посева опять один и так далее если утебя М2 размером 1ТБ сможешь посеить 3 участка паралельно. временной кэш рассчитывай так (K=32- 269 GB (288.8 GB) ) это был пункт 3. пункт 4 где уже из кэша переносица на основной носитель жостки диск (HDD) чем больше тем лудше чтобы много посевов было. после этого твой М2 не нужен он не участвуит в фарминге. главное иметь быстрый ссд для посева ото будешь сеит весь год :Д а там сложность вырастит до небес а тут главное сделать всё быстро чтобы быстрее заработать.

[ChandlerFerry]

@hoffmang9 With k32 plots being completed in sub 10 minutes due to a new multithreaded plotter (https://github.com/madMAx43v3r/chia-plotter). Will Chia-Network be moving to k34 plots to maintain the 10 year buffer?

[BitSec01]

And what stops people from building asics?

memory

edit: At least that's what I've always heard is a hurdle too big for asic creators.

[exander77]

And what stops people from building asics?

memory

edit: At least that's what I've always heard is a hurdle too big for asic creators.

I don't think you need much memory. You have to take into account, that you will not be sorting or storing anything, you will just test if the attempt leads to proof and if not you will throw it away. The memory used for plotting is for sorting, compression and storing actual plot file.

[kanasite]

Well well... a factor of 512, how did I miss that... That changes the game quite a bit, now you got my GPU brain thinking. Some rough calculations tell me it's a close race, I think only an attempted implementation could tell if it's more profitable or not. (If a decent CPU can do phase 1 in 10 min, a decent GPU should do it in ~15 sec, assuming BLAKE3 plays well on GPUs. Such a GPU will probably draw ~250W and it competes against 1024 plots on lets say ten 10TB HDDs, drawing ~100W.)

Does that mean GPU can solve any puzzle in 15 seconds, and winning every reward?

not true, the only thing that a machine is able to compute phase 1 in less than 30secs only break the current plot filter (512 factors). In layman terms, it's just having the benefit of 512x better chance to generate plot on the fly but doesn't guarantee the best proof. However, it's a matter of concern if it's economically profitable to generate the hash on the fly, which defeat the purpose of "proof of space"

[exander77]

Well well... a factor of 512, how did I miss that... That changes the game quite a bit, now you got my GPU brain thinking. Some rough calculations tell me it's a close race, I think only an attempted implementation could tell if it's more profitable or not. (If a decent CPU can do phase 1 in 10 min, a decent GPU should do it in ~15 sec, assuming BLAKE3 plays well on GPUs. Such a GPU will probably draw ~250W and it competes against 1024 plots on lets say ten 10TB HDDs, drawing ~100W.)

Does that mean GPU can solve any puzzle in 15 seconds, and winning every reward?

not true, the only thing that a machine is able to compute phase 1 in less than 30secs only break the current plot filter (512 factors). In layman terms, it's just having the benefit of 512x better chance to generate plot on the fly but doesn't guarantee the best proof. However, it's a matter of concern if it's economically profitable to generate the hash on the fly, which defeat the purpose of "proof of space"

Exactly as you say. If it is economically viable to produce proofs with similar probability on the fly, it basically degrades proof of space into proof of work.

[kanasite]

anyway proof of space literally meant "proof of workED" ;)