Robert’s Rules for Success, or at least reduced chance of failure

I’ve made a few references lately to avoiding Jurassic failures. In some tech circles, including cryptocurrency projects, it seems very popular to make bad decisions and not claim responsibility. And yes, I’ve been writing and talking and thinking a lot about crypto this year. I’m not alone, but I’m the only one on this blog who you’d be able to make that observation about.

The Jurassic reference of course is to Jeff Goldblum’s character Dr Ian Malcolm in Jurassic Park.

Your scientists were so preoccupied with whether or not they could, they didn’t stop to think if they should.

The Jurassic Park reference is a bit more universal than the one I used to use, which was to Wangdi something. I had a Nepalese classmate in college named Wangdi, and the main thing I remember of him from those years was his disc soccer/disc golf prowess. Well, that, and the time he had someone spread three soccer discs in a throw which he would normally have caught with ease, but in this case he tripped over a sprinkler head about two steps into his run and faceplanted in the quad.

  1. Don’t always do things just because you can

    Corollary: Don’t always do things just because you saw them on the Internet

2. Don’t try to start out at full speed; watch where you’re going and work your way up

Corollary: Set a reasonable plan and try to follow it. Don’t get distracted by squirrels

3. Plan to spend at least one minute for every $100 spent, learning how your item works

Corollary: If you can’t do that, don’t expect others to do it for you for free

Between these two warnings, you can take something away. First, don’t always do things just because you can (or because you saw them on the Internet). Second, don’t try to start out at full speed; watch where you’re going and work your way up.

And third is my One Percent Rule, not to be confused with Arthur Conan Doyle’s Seven Percent Solution. For every $100 you spend on an endeavour, spend at least one (1) minute learning or trying to understand it on your own before demanding help in free volunteer forums or from overworked support staff. So if you’re spending $3,000 on a GPU, but you’re not willing to spend 30 minutes learning how to use it, maybe don’t spend it. See also the first and second rules above.

Continue reading →

Chia hardware starting points

I started writing this section as part of my So you want to farm Chia post, but that post was getting pretty long. So my hardware suggestions will make their own separate appearance here.

If you found this post first, please read So you think you want to farm chia? first. Then come back here. I’ll be waiting.

As I mentioned in the other post, start with what you have, if at all possible. An underused 2-4 cores of CPU, an extra 4-8GB of RAM, 300GB of temp space and a couple hundred gigabytes of permanent storage will get you going.

If you can’t start with what you already have, you can build a very viable system for around $1000. Since it’s not very specialized, you can reuse it for something else (a desktop, another software project, a VMware home server, etc) when you’re done, or when you outgrow it. 

These bills-of-materials will have some pricing and Amazon affiliate links. Prices are current as of the original posting, and will change over time. You can choose to buy them anywhere of course, or change the components up. If you’re in the San Francisco Bay Area, I’d recommend checking out Central Computer to support one of our few remaining local computer stores, and I’ll mention their advertised prices here as well as Amazon’s. Sign up for their Preferred Customer program and you will probably save a few bucks. I get nothing if you buy from them, except warm fuzzies. 

Also be aware that some of these items may be hard to find at times, as Chia is becoming more popular and a lot of popular items are getting bought up.

One last caveat before we get to the configurations:

If you’re using an SSD or NVMe drive for plotting, the plotting process will wear out the SSD eventually. A single plot uses up to 1.8TB of writes. Check your drive specs and figure out what lifespan to expect, and don’t use your boot disk as plotting space. As examples, the 1TB Samsung SSD980 is rated for 600TBW, or 333 plots (including failed ones), for $130. A Seagate Firecuda 520 1TB is rated for 1800TBW, or 1000 plots, for about $180. So spending about 50% more gets you 3x the endurance. 

Continue reading →

So you think you want to farm chia?

Chia is a great fiber supplement for your diet. But have you considered it as a supplement to your cryptocurrency mining?

Please note: While this post was originally published in April 2021, and some screenshots are outdated, it is still largely accurate and not outdated as a whole. New plotters have come out, more data is available on the Chia wiki for hardware configurations and drive options, and in general this is still a good primer. 

A package of Target Simply Balanced chia seeds

Chia is a proof-of-space-and-time cryptocurrency, invented by Bittorrent creator Bram Cohen and inspired/designed after Satoshi Nakamoto’s original Bitcoin paper. You can read plenty of the underlying details and math around it at chia.net, but what I’ll be covering here is an introduction to how Chia works for the person making it, and how to get started with Chia.

You should read this post first, but I have hardware suggestions in another post, and both the chia.net website and chiadecentral.com have plenty of ideas as well.

There is a Frequently Ungoogled Chia Questions post available as of May 11, 2021; it will get most updates, although if anything in this post becomes totally wrong, I will fix it. 

I wrote up my NUC plotter build step by step.


I also need to note that, while I appreciate all of the feedback and the record-setting number of viewers to this post, I cannot provide free consulting and rig reviews to everyone who submits a configuration. I’m happy to try to address specific questions in the comments, but general “help me get started” and “what hardware should I use” questions are answered in these posts and I won’t be able to review everyone. 

 

First, The Caveats

Read this post, and some of the linked resources, but the best thing to do is to start with what you have on hand. A big part of the design of Chia is to use reusable components (unlike ASIC mining and even certain GPU mining, which eventually outgrows the components, which will then be less useful for any other use). 

If you have a modern desktop with 8+ CPU threads, 16GB or more of RAM, at least 300GB of free SSD space and at least a few terabytes of HDD, you’re good to go. 

If you don’t have anything like that, I’ll share my own configurations below, as well as some hardware lists in an accompanying post to help you find a manageable starter farm for around $1000 and another more scalable one for under $2000 (less if you start from an existing system or components) . You can also find hundreds of tested configs on the Chia subreddit and various blogs. 

If you decide later to invest thousands of dollars in hard drives, SSD, big CPUs, and lots of RAM, you’ll better understand how the process works and where to put your money. Or if you’re like me and have a pile of old servers in the garage, you’ll be able to upgrade what you already have and make the most of it. 

If you’re using an SSD or NVMe drive for plotting, the plotting process will wear out the SSD eventually, although it will take a lot of effort to do so.  A single plot uses up to 1.8TB of writes. Check your drive specs and figure out what lifespan to expect, and don’t use your boot disk as plotting space. As examples, the 1TB Samsung SSD980 SSD is rated for 600TBW, or 333 plots (including failed ones), for $130. A Seagate Firecuda 520 1TB is rated for 1800TBW, or 1000 plots, for about $180. So spending about 50% more gets you 3x the endurance.

But be aware that the TBW rating is more of a warranty consideration, rather than a “this drive will die after X TBW” consideration. You can dig into details of how they’re rated, and you’ll find that drives will often last way beyond the TBW rating especially when used for large sequential writes or heavy read ratios (both of which apply to Chia plotting). 

And don’t expect to get rich on day one. For starters, Chia transactions don’t open up until May, so you can’t do anything with your Chia until then. And it may take you days, weeks, or months to find Chia, and none of us know yet where the value of the currency will go. Calculator sites are estimating $20. Some speculative sellers are looking at a lot more than that. But today, it’s just pre-farming and getting ready for the full launch next month.

Continue reading →

Colin Quinn and the Art of Mining Rig Maintenance

I’ve been doing a lot with cryptocurrency mining lately, so there’s been a lot of activity here on rsts11 about it as well. For this post, a lot of the concepts can apply to a home lab in general, although it was inspired by several conversations on mining channels on Telegram and Discord over the past month.

In the late 1980s, actor Colin Quinn was an announcer on a game show called Remote Control. It’s not terribly remarkable in context, but as I’ve had a few people of differing technical levels ask about remote control for their mining rigs, it seemed relevant. 

Whether you’re managing a single server, a home lab, a mining rig, or an entire farm, it’s likely that your gear will not be in  your home office or possibly not even in your home at all, but you won’t want to travel to where the gear is everytime something needs to be power cycled. In the case of one of the people I consulted for recently, he was going to deploy mining rigs at a friend’s house in another country, and would not be able to be hands-on with the gear more than 2-3 times a year.  

So the considerations I’m making here are tuned for a relatively small number of systems that may not have IPMI console access, and may not have convenient access or remote hands services available. In a datacenter or similar environment, you may have contracted services to send someone to hit the reset button or look for issues. For many of my readers, this is not a luxury you’ll enjoy. 

My own remote access adventure happened when my oldest mining rig was a few miles from home. Thanks to my Meraki networks, I did have direct VPN access, but the rig would occasionally crash and require a physical power cycle. It was only 5-10 minutes each way, and I did have 24/7 access to the facility, but it got very inconvenient at times. 

The considerations I’ll layout here fall into a couple of categories. 

  1. Network access (usually via VPN of some sort)
  2. Console access (remote KVM)
  3. Power control (remote power controller over Ethernet, via #1 above)

Disclosure: Neither Robert M Persig, Colin Quinn, nor Ken Ober have any association or endorsement or even awareness of this blog post. 

Continue reading →

Wrestling with an ONDA B250-D8P-D4 mining motherboard

Most casual crypto miners use a conventional motherboard, especially if they have a PC/case/power supply with sufficient PCIe slots for their GPUs. But when you get beyond 2-4 GPUs, you either need a rat’s nest of riser extenders, or maybe (just maybe) a dedicated mining motherboard.

I recently got a new-to-me mining motherboard, and found it painful to find some information and resources I needed. I’m aggregating this information in this post, and it will get updated as I get more relevant experience with the ONDA motherboard in question. If you have any info to share, feel free to comment below and I’ll update. (Last update 2021-03-14)

I’ve mined with an Octominer 8-slot motherboard for 3 years now. In addition to an onboard Celeron 3855U and a single DDR3 SODIMM RAM slot (max of 8GB), it has eight PCIe x16 slots, so you don’t need to use the common x1-to-x16 risers. It’s complicated in that you have to power the motherboard with a number of additional power connectors (in this case, 6-pin PCIe power leads from the power supply). But it sits flat on a custom frame I ordered in 2018, and it doesn’t have much that I don’t need (like lots of drive controllers, extra memory slots, audio, etc). And if you get a custom mining power supply (or breakout board) with only 6-pin connectors, you’re in good shape. 

Octominer has discontinued their 8-slot boards, and the boards may not support the latest GPUs on the market (much like the Ethos mining distro I used on it until this past week). I couldn’t get the board to boot with an AMD 5500XT GPU (Amazon, eBay) in the first slot, for example. So it’s chugging along with eight Sapphire Nitro+ RX580 8GB cards (Amazon, eBay), seven of which have been chugging for almost three years now.

While they still make custom boards, the only ways to get their products are either to find the rare used item on a marketplace, or to buy the one integrated rig they currently sell in quantities less than 10 (their x12 rig with everything but the GPUs, which runs almost $1,000 shipped to the US). 

Another company making custom boards is ONDA. You can usually find them on eBay or other marketplaces for a couple hundred dollars, with a range of slot support. I found a good deal on the B250 D8P-D4 recently, and since I wanted to aggregate a mess of old GPUs, it was an easy way to go. Continue reading →