I’have different machines with different characteristics and after some tests is probable that the best for me is to mix plots types, some non-compressed plots, some compressed and farmed with CPU and some compressed and farmed with GPU.
And my doubts are :
How the fees will be calculated?
Are the fees calculated per partial/plot, worker or miner?
Fees are only on Gigahorse plots.
For FlexFarmer run two instances, one with GPU for the 3% fee and one with CPU for 1.5%. Use each instance for the gpu/cpu plots.
With Flexfarmers client for compressed plots it’s a percentage of partials.
With the gigahorse client it’s as follows:
* 3.125 % when using GPU(s) to farm compressed plots
* 1.562 % when using CPU(s) to farm compressed plots
* 0 % for regular uncompressed plots
When you find a block there's a chance the 0.25 XCH farmer reward is used as fee, this is a random process. In case of CPU farming it's 1 out of 8 blocks on average, and for GPU farming it's 1 out of 4 blocks on average.
When the fee is paid from a block, you will see a log entry like this:
full_node: WARNING Used farmer reward of block 2187769 as dev fee (3.125 % on average)
It will show the block height as well as the average fee that applies, depending on if the proof was computed via CPU or GPU.
So dev fee is taken from the Farmer Reward and on a per proof basis. It will show in the log.
My understanding is you can mix and match compressed and uncompressed plots over harvesters using cpu and gpu to farm the plots. When a block proof is found and rewarded that specific proof will decide on the method for dev fee application. In case of GPU farming a compressed plot a 4-sided dice is rolled with 3 sides ‘you win’ and 1 side ‘max wins’. In case of CPU farming a compressed plot an 8-sided dice is rolled, with 7 sides ‘you win’ and 1 side ‘max wins’. In some clever C++ code of course;-)
As it’s from the Farmer Reward it applies the same for farming solo as well as pooling!
Then if I’m not mistaken, in a hypothetical scenario where I have: 100T NFT OG plots, 100T compressed (c3) farmed with CPU and 100T compressed (c8) farmed with GPU, everything farmed with flexfarm client, mixing plots in between workers, fees would be calculated like this:
* 0 % fee over partials of 100T NFT OG plots.
* 1.562 % fee over partials of the 100T c3 CPU farmed plots.
* 3.125 % fee over partials of the 100T c8 GPU farmed plots.
And lastly but not leas because is the flexfarm client I keep my blocks, as usual.
Once more thank you for your help and clarifications @Chris22 and @xkredr59.
Don’t know… I read @Chris22’s earlier reply different, he talks about running separate instances for gpu and cpu farmed plots, having different percentages 3 and 1.5% of partials respectively.
So as I understand you have to run separate flexfarmer clients for uncompressed, c3 plots farmed with cpu and c8 plots farmed with gpu.
But I don’t know Flexfarmer, maybe one client can support separate instances…
My explanation was explicitly for the gigahorse client, so not Flexfarmer.
FlexFarmer is lightweight so you can run two instances fine although the Gigahorse version may have a bug we haven’t discovered yet, if so please report it.
Just disable GPU for the FF instance you want to CPU farm.