When one pool get a majority that creates an environment for a 51% attack does it not? Not suggesting this pool can get to those heights, but never assume never. It’s conceivable that if a pool has a 30% added profit, it gets the farmers. Eventually. So seems evident to me that Chia needs to come up with something. In theory this pool is a liability for the blockchain. Not today, but nothing is about today. As the pool is showing itself as legit, and growing, it will be a likely force to be reckoned with, especially with any new Chia farming participants. As they say, the plot thickens. Either that or I have no concept of a 51% attack.
I am confident Chia comes out with reasonable plot compression or similar before this pool gets to 51% of the netspace. Even if plotting is 2-3x faster, plotting EiBs of plots is a lot. Their solution may not reach 30% but carving off double digits from the 30%, I would not be surprised at all.
1 Like
I’ve been plotting for few hours already and I have to say that performance of the plotter is superior compared to Chia branded plotters (assuming it pumps out valid plots).
Right now I’m plotting on R820 4x4650 V1 with 448GB. I’ve ended up with 3 instances of nossd plotter running under Docker. Each instances allocates around 128GB RAM and uses 225GB tmp file on NVME (Samsung 980 Pro 2TB). It is unclear to what degree the TMP file is used. Plots are written directly to HDDs in JBOD by the plotter (I’ve attached 2 HDDs to each instance so far). Plotting speed (for sparse plot with C5) is around 11 minutes per plot (averaged over 3 instances). On bladebit I had around 18.5 minutes per plot. I don’t know how much finalization step takes, but I would assume around 15-20 minutes per plot. With 3 instances running CPU is not maxed out, however I don’t have enought RAM to start 4th instance.
I guess that after CAT1 drama, Chia team will need to focus on plots compression, because current situation (existance of closed source superior plotter connected to a single pool that owns keys) is just plain bad for the Chia blockchain.
@Dawson I have a question. I first plotted without -c 5. The plot size was about 100 GB. After I added -c 5, the plot size (SPT) was about 92 GB. If the disk will be full then and I keep -c 5, will the 100 GB plots also be reduced to about 78 GB or do I have to replot them to get 78 GB plot size?
You have to replot those first sparse plots to get -c 5 compression, but you can wait for the disk to be full and keep mining on them for now
3 Likes
I will do that and thanks for the quick reply!
Even if you don’t plot with a high-end system, the performance gain is very high. With my relatively weak 4-core CPU and 64 GB RAM on an SSD and with Madmax plotter, I previously managed about 10 plots per day. Now it’s about 25-28 per day, and that’s just with hard drives. They definitely delivered what they promised.
@Dawson
Can you explain what the 225GB TMP file is used for and what its usage pattern ? As I’ve said in the previous post, I have it (3 of them) on NVME. Does it make any difference (compared to having it on HDD) ? I’ve tried having it on tmpfs and it didn’t make any difference.
With 128GiB of RAM temp files are used during save/load only. If you don’t need to save/load quickly, you can put them on HDD.
1 Like
After your drives fill up (or get close) the plotter will start to convert them to FPT files (and be 78GB).
Dawson wrote that I have to replot for that.
Can the pool explain why they are winning blocks twice as fast as expected?
1 Like
ppl still track this scam topic, and some say they delivered with zero proof or you all sheeps here???
you do not even know what space they have in their pool, they can fake all
1 Like
getting lucky, quite simple
That was a joke question, I guess.
Some still believe, “Expected time to win” is a guarantee of consistent reward interval…
2 Likes
Agreed. Actually, checking some pool block wins where they specify what was the luck rate for every block is an eye opener. I was looking at a couple of pools, and getting 8x behind was not that uncommon. Also, getting a couple of blocks within a minute was happening a lot. When looking at Poisson distribution, those should virtually not happen.
You got me. I was expressing the hate nossd was getting when there were no blocks by making a specious claim about too many blocks
1 Like
That was a great job, by the way. 