Скачать или смотреть is k=32 broken!? 15 minute plots with madmax and Intel Optane SSD

Is k=32 broken now that people are completing sub 15 minute plots!??? (hint...no) but learn about the new madmax plotter in the video.

Chia is about proof of space, in the end, all the matters ishow much storage capacity you have farming. The plotting can be very fast, hopefully, fast enough where writing the plot sequentially to the disk for farming is the actual bottleneck. There is some weirdness if people can complete phase 1 in under 30 seconds, this is called a replotting attack.

The death of k=32? Not so simple. In the consensus paper, it discusses a replotting attack where an attacker gets a challenge, tries to create a plot in real-time that has a proof to that challenge, and throws out any results immediately that do not pass the plot filter. Currently, this has to be done in the same time the farmer / harvester have to get a valid proof out, which is just under 30 seconds. I'll be talking about this at length next week with the Chia team in a public zoom, so I actually am not going to cover it in this video.

The new plotter makes use of available cores for multithreading more of the phases. The actual amount of data written is still around 1.3TiB for the total 101.3GiB k=32 plot, but this now can be done with a considerably lower amount of temporary storage. The new exciting thing is the temp 1 & temp 2 also operate differently, and getting 100GB of DDR4, SLC, or Optane is much cheaper than buying high TBW and high capacity enterprise SSDs.

On systems that are already optimized and have lots of SSDs, this may actually be a step backwards in performance, but it opens the door to many more interesting system configurations for plotting that can achieve high output per day per system.

https://github.com/madMAx43v3r/chia-p...
https://docs.google.com/document/d/1t...
https://github.com/Chia-Network/chia-...
https://docs.google.com/spreadsheets/...