The question is in the topic: are Gigahorse compressed plots more energy efficient to farm solo? If pool difficulty affects how much energy is used, then isn’t it best for everyone to go solo if the variance is bearable? I have 2 weeks block time and could take the variance if energy efficiency with compressed plots is better then.
Sorry this is in the wrong section, should’ve been in farming!
Set difficulty to over 9000 if your worried.
i don’t know if it’s just me but gigahorse plots seem harder to win a block it’s been a month and my 4000ish plots haven’t won any
With a 0.5PB farm, changing the difficulty to 10 results in hardly any partials (and it is those that use extra power)
That should be true, yes.
Super good question because even though it’s a partial, the plot table still has to be calculated like normal right, so would be using a lot more power?
Partials do require your GPU if your GPU farming. That’s why we recommend setting your difficulty high. At high enough difficulty, the use is minimal.
It seems like the optimal way would be to join a pool and set difficulty to a level that confirms your ability to produce partials (which take the same amount of energy as actual blocks, right?) without drawing too much energy.
Is it possible to set minimum difficulty at Space Pool?
The only way to find out is to run a controlled test. I have already started testing different Harvester power usage (kwh from wall) based on difficulty. Recommended difficulty is ~1100 for my 4.80PB farm. I am testing 500, 1000, 1500, and 2000 over a 24 hour period. Once I am done I will also test solo.
But do understand this is not bank breaking numbers since the GPU is only using $270/year at twice the recommended difficulty. Obviously this would matter for inefficient systems that are using multiple GPUs to farm, but I only use one and my system is built from the ground up to be highly efficient since I come from 6 years of GPU farming where enrgy consumption makes or breaks you in bear markets.
a nice test, i can say from my experience that it doesn’t really make much difference in consumption as long as you don’t go below 100 (i have effectively 2,1Pib).
I would argue that no 217 C7 plots with 8 C8 plots fit on an 18 TB disk, it’s more the other way around.
as far as efficiency is concerned, I’m at about 0.3 Watt/Tib (all 18 HDDs with 235 C8 plots), so there’s still plenty of room for optimization with 0.8 Watt/TiB.
have a nice day, i’m looking forward to your complete results
Here you go sorted by size, smallest files first. You can see the 8 C8 plots and then the rest are C7. 225 total files.
Because of my farm size, future expansion, and plot filter halving next year I went with C7 after extensive chiapos testing. This is why the majority are C7 not the other way around.
yes was my mistake I have calculated with 235 total and not with 225 why ever
but what I’m still a little surprised is your power consumption with your farm size you should have about 300 18TB HDDs and even if we now say each consumes 10 watts which is way too much would be only 3kW/h and not 5.9 kW/h
what consumes so much extra?
I think you are confusing kW with kWh. I am using an HS110 wall meter to track total watts used over a 24hour period. So the 5.9kWh is 246Wx24 Hours. This is important because the GPU power fluctuates rhythmically during idle and proof lookup times. This is why it is better to track kWh and then if you know the period you can divide it by that to get the average power the harvester uses. In this case it is 246W.
Also I am only measuring the harvester power with GPU and not hard drives since harddrive power should not fluctuate since they never sleep and reading uses negligible power.
Finally my total farm power is 2.1kW for 267 x 18TB enterprise drives. My systems are set up to be extremely efficient and I never use server racks that are heavy, waste space, scream like jet engines, and are inefficient.
ok now i get it
but if you only measure the harvester consumption, then you do not consider the minmal increased consumption of the HDDs in the proof accesses which could then possibly have an influence at very low difficulty.
Whether this is so I can not say because I have only ever measured my complete system.
ok your system is then exactly on the same level as mine in terms of Watt/TiB, I also use only consumer hardware
It looks like checking for plot quality still uses GPU even if no partials are produced. Is this because of the plot filter and only going to get worse when it decreases? On top of that, producing partials still amounts to a considerable additional energy demand?
Are you sure you are getting 235 C8 plots on 18 TB hdds?
I formatted in ext4 my drives and im getting only 234 plots.
See this :
yes and there are then still about 9-10 GB free. but I had already written elsewhere that only works with the cuda plotter the cpu plotter makes me the C8 plots on average 50MB larger and then only 234 plots fit on a 18 TB disk and there are about 68GB free.