4 minutes for phase 1…ok it’s still not 30 seconds for a plot, but we’re only a few months in.
Still not using GPU, Asic or super high end Epyc/Xeon stuffs. Seems like it can be pushed a little more
What do you guys think?
Soon most of us will will have to start replotting for pools, and I really don’t want to have to do that again a few years from now because by that time I will probably not even have a plotter anymore.
Edit, interesting discussion on this topic with Bram and Madmax also commenting:
On the question of risk to k=32 it isn’t enough to simply get the proofs in fast for k32 to be in danger it must also be that it’s more cost effective to have that hardware generating plots on the fly against the existing netspace. A few things complicate this: You only need the first phase to be done and you get a factor of 512 because of the filters, but you’re competing against well below cost disks at equilibrium. Even with those factors it still appear k32 is quite safe, and if it become unsafe for plotting reasons it isn’t clear that cranking up the minimum k value will be the right countermeasure.
Here is the good news,… this is all playing out before we need to replot for official pools.
I would think Chia will come out with a statement, that ensures K32’s are still good (and explain why) OR that K33 or k34 or K35… etc. is going to be the new minimal size starting x months from then… just before pooling protocol comes out.
If they don’t by themselves, the community will demand it
Chia cannot come out with a statement that K32s are still good because they have a guaranteed shelf life.
The estimates I have read run from 3-15 years, but the time WILL come when K32s become useless.
The company will have a Zoom call soon to address this, but here are my thoughts (I don’t work there).
You would need to be able to plot Phase 1 in ~10 seconds on a machine that costs ~$1000 (the cost of storing 512 plots) for this grinding attack to be cost-effective. If that happens, it does not mean that Chia is no longer secure; it just turns Chia into a PoW chain.
Countermeasures would likely be either to raise the k limit or reduce the filter size, or both. For example, lowering the filter from 512 to 256 would mean your plotter would have to cost $500 or less. Increasing k by one would make the plots take twice as long.