I use Poolchia and am very happy.
1 Like
tell me how you coordinate when working with him PoolChia?
1 Like
PoolChia channel at Discord.
The fact that you spent $700 and only had 8 TB boggles my mind. You would of never made money. I spent a total of $1200 and have 94 TB. That includes all the HDD, M.2, cables, everything. Already have my computer.
Edit: I think I actually spent closer to $1500⌠but still
Maybe you could also partner/sponsor Jeff Geerling, and ask him, whether having one/two HP H240 ($50) would be possible with RPi. Such setup (up to 32 HDs) would potentially make 90% of your clients really happy - maybe all, but those few biggest one (I would switch instantly to such a setup). He already did work on a single HBA, and my understanding is that that work was propagated back to RPi sources.
Such solution together with your code would resolve all those RPi USB issues, and make it a really superb harvester.
Actually, he got 16 SAS/SATA HDs on RPi CM4 already. Maybe you could make your code running on that model? Having that 16 SATA connectors opens up a possibility to use port multipliers on those connectors (they work in X86 Linux, potentially may/should run on RPi, if not maybe he can help, as he is pushing an envelope with RPi and SAS/SATA). That would be theoretically 80 SATA HD drives. Even if the max would be 30-40 HDs, that would be again a superb harvester.
Also, in this video, he mentions Chia project, so he is already familiar with what we do. Maybe he could be really happy to collaborate with you, as that would be again, pushing the envelope.
An issue I had was needing to start and stop the OS due to issues with Chia freezing (GUI and CLI). Would this impact drive performance over time? During the time I was unable to connect due to the CGNAT I had been told not to run the drives as they would be pointlessly worn out until the network issue was resolved.
@lybrary The tutorial is behind a paywall. This sounds too good to be true. Could you post a link to a free version?
@Chris22 Could you post a link to your docs and/or a tutorial specific to RPi?
@cjd9153 $1500 for 94TB drive space compared to my $250 for 8TB is an incredible deal. Where and when did you buy your drives? Could you provide manufacturer and models?
Drives have two mechanical components - the main spindle and the heads. Those components wear out with time/usage, and heat is the main issue. Chia has really small amount of reads per day, that leaves the main spindle to be the most affected. That spindle of course prefers to be turned off, but when on power it doesnât like âconstantâ starts/stops (say every 10-20 minutes). Chia keeps those drives 24/7, so that spindle should not have to many starts/stops. Also, those drives donât know anything about your computer status / network behavior etc. So, I would not worry about that advice. Your 20 or 100 starts/stops are nothing comparing to 2-5 years those drives should last, as the manufacturer assumes that they are connected to your main computer and start/stop several times per day.
Sorry, there is no free version. I spent months testing and ended up buying things I learned were not necessary. In other words, I spent money so that you donât have to. Purchasing my ebook will save you money and time. Asking $15 for it is more than fair because depending on the size of your farm you will save a lot more than that. My ebook can be bought here: How to Farm and Plot Chia Coins with just a Raspberry Pi
Usually, if it sounds to good to be true, it ends up like that as well. For instance, RPi will draw 10+W when plotting, so your 23 hour plot draws about 230W/h of power. On my box, I get 30 min plots on less than 200W box, that makes about 100W/h of power to plot. So, the whole notion that this way you save power is BS. You just spread your power consumption over much longer plotting times, and end up using more power per plot.
Take a look at Flex pool. They ported their farmer to RPi. Also Jeff Gearling on YT makes excellent videos about using RPi with SATA drives (though on CM4 model, that at this point cannot take Flex farmer - hopefully âyetâ). Most likely, if you want to go RPi route, Flex support (discord?) would be a good place to start.
Flexfarmer 2.0 (to be release shortly) supports 32bit devices. It also cuts power and wear down as weâve rewritten the entire PoST library to optimize it as there were several errors and inefficiencies. Reports are that Flexfarmer uses 1-2 Watts on a RPi and under 100mb of ram.
1 Like
HD is OS agnostic. Either it spins or not; either it reads/writes or not. Chia keeps those drives in READY/IDLE mode 24/7, and from time to time does some reading (regardless what OS Chia is running on). So am I missing something that RPi does to âkeep mechanical wear and tear on the HD low?â
Nice calculation.
Now we add the loss by waiting ages for each plot, before it can be used for farming while the HDD already draws full power⌠and then we start loughing about the idea of plotting on an pi. 
1 Like
Yes, you are missing quite a bit. Maybe what you wrote should theoretically be true, but in practice it is not true. My discs, when they were badly setup, had to spin up for each proof/partial which meant the harvester was in some cases too slow to deliver a valid proof. Once they were properly configured I had continuously spinning discs and very fast delivery of proof/partial. It all depends on what you know about HDDs, Linux, and RPi.
I am afraid you donât see the full picture. You were correct in the very early days of Chia with strongly growing networkspace when literally every day counted. In those days plotting speed was very valuable. Now the networkspace grows only slowly. If the price of Chia continues to drop the networkspace might even shrink. In a situation like that speed of plotting is not anymore as valuable. The cost of plotting therefore factors in much more prominently. The RPi is the most efficient one for farming. Being able to also slowly plot to add to your plots without any additional plotting computer cost is an advantage. Of course, it depends on how many new HDDs you want to fill in a given time period. In my economical assumptions slow and steady wins the race. Your assumptions may be different.
Letâs assume that I donât know much about those HDs / Linux / RPi. Are you saying that the magic is to disable HDs spin down (hdparm -S 0 /dev/sdX)? That was your mistake / secret sauce?
No, not at all. It is only one of many troubles I encountered and things I figured out. Just to be clear, I donât claim any âsecret sauceâ. Anybody with the time and willingness to try, test, measure, and figure things out can figure this out on their own. But I have found that many either donât have the time or donât have the willingness to experiment and work through difficulties, errors, and things not working properly. My ebook offers a shortcut through all the troubles I encountered because I describe them how they manifested themselves and then what I did to fix the issue. You will also learn how my sytsem is setup for maximum convenience and efficiency. I donât have all day to babysit my farm.
Assuming that you want to fill up 10TB drive, and you are getting one plot per day on RPi, that will take you three months (~90 plots). During that time you pay for 6W/h of drive usage, where that drive has not that many plots (less chance of winning). That would be 6W * 24h * 90d = 13kW just to keep that drive up until it is fully plotted. On my desktop, I can make those plots in 2+ days, as such 6W * 24h * 2.5 = 360W. So, yes you spread that over a longer distance, but will earn much less (while belonging to a pool).
During those 90 days, my drive will earn (through a pool) about 0.2XCH, and your 0.1XCH. With XCH at 160 today, that is $16 difference - enough to cover the electricity cost, plus another $16 is the gain that you donât have. Also, my drive has 2x chance of winning anything (yes, a slim chance for both of us).
Again, the RPi draws ~10W or more when plotting, that is 240W/plot power draw. On a desktop (like mine), that job takes 30 minutes, where the box maxes below 200W, so it is only 100W/plot. Therefore, with 90 plots for that 10TB HD it is a 100W/h * 90plots = 9kW difference - your extra cost.
RPi is fantastic as a harvesters, as that job requires virtually no processing power, so it may as well take just 5W to do that job, where a desktop will draw about 50W. However, that CPU is by definition not energy efficient when it comes to using full power.
So, for someone that doesnât have a desktop that can be used for plotting, and wants to get just a couple of HDs worth of plots, RPi may be a good solution to both plot and farm. However, when you want to push it a bit (people on this forum are talking about ~20 drives sitting on RPi), that setup will need 20HDs * 90days = 1,800 days - uhm, about 5 years. I guess, they could get 20 RPis and shrink that time to just 90 days, and sell those 19 extra RPis when they are done. That may be also a model, no doubt.
Maybe that is how it should be advertised, then? It is clear and concise description what you have there. Agreed that for some that can save a bit of time.
You are assuming you have a free PC sitting around doing nothing. That is not an assumption I made. I assumed one has no free hardware. Everything that is used in plotting and farming needs to be purchased. I am further assuming that HDDs will be bought bit by bit over time not in one big swoop. If you understand time value of money, price swings and opportunity cost, then you will understand that this is a big advantage. Your assumptions about RPi power consumption are wrong. And yes, one can scale this to more than one RPi.
Yes, I think that I donât understand that part. Although, I am not really enlightened by your writing, so am potentially still lost.
You have provided none, so âŚ
Listen, all (I mean most, as you are apparently not) of us on this forum are trying to help each other. Sometimes we help, and sometimes we are being helped. This process benefits all of us, and I would assume that you have learned most of what you cover in that eBook here. So, my take is that you are an outlier that peddles that ebook making false claims.