Okay… to step back a bit, it’s not QUITE that complex. I’m building a private, small (size and scale) maker space. It’s going to be a bit of an ongoing experiment to lay the groundwork for *actually* building a large scale, public space in a better location down the road and hopefully making it open to public membership. I just secured a current location in Santa Monica in a great area that’s close to the expo line and down town SM. It’s only about 350 sq feet, but its bigger than you’d think and more importantly, the lease is exceptionally well priced for our needs, includes power and internet in the cost, and has its own entrance/exit with 24/7 access that I can apply some smart automation to. We’re going to start off small… I and a few Rioters are going to be moving some 3D Printers, tools, accessories and other related stuff there. Though, no major power tools, no CNCs, nothing that would disrupt the 9-5 lives of our fellow tenants in their tech offices out of both politeness and necessity. My hope is that my fellow Rioters and friends can make good use of a shared creative space for 3D Printing, Cosplay making, prop fabrication, circuitry noodling and other stuff in a relatively small but shareable space via evening creative sessions and fun group weekend collaborations. I’m not going to ask for a specific fixed dollar amount from folks in order to partake in the space to fully offset the cost, but hopefully people will contribute enough in their own way that I’m not fully out of pocket entirely, and it can be self sustaining. It’s a month to month lease so we can experiment as long as its working for us, and either sunset things if they don’t or use it as a springboard to securing a bigger, more appropriate long term space and turning it into a public venture. Additionally the monthly cost is reasonable enough that I am comfortable shouldering it myself as part of the experiment, worst case.
I have no idea what I’m doing and this is probably a stupid idea I’ll regret down the road, but I’m incredibly excited and highly motivated right now! On a related note, if anyone is interested in investing into an *actual* Makerspace on the Westside of LA with me, that changes things greatly and offsets the challenge of bootstrapping an idea from scratch. Let’s talk.. You know how to find me.!
So, several people have asked me for the basics on building a GPU based crypto currency miner (more specifically for Ethereum, or ETH). Previously I outlined how to setup up an Ubuntu one for the very, very basics with existing home PC hardware, but they’ve wanted more details for a full fledged setup. Now that I have mine in a happy place, tweaking-wise, I think I’m ready to share my findings. Please do note I am not a complete expert, and you might find better advise elsewhere or additional tweaks I didn’t suggest. I’m not saying this is perfect guidance, but I will stand by it being good. 😀 Additionally, I’m going to make some assumptions to avoid bringing too much into-scope on this write-up, and expect that you already have an Ethereum wallet setup, and have chosen a mining pool to join mining within.
(Oh and to answer the question that comes up a lot: Yes, mining ETH is still practical. ETH was specially designed that custom mining hardware like ASICs will never be a thing like they were with BTC. It’s designed to always live on GPUs. Eventually the difficulty will curve enough that you need tons to do it practically, but right now my 5 GPU rig brings in 3.9 ETH (or $700 after power costs). Investing and buying it might still be better for the average Joe, but if you are a tinkerer and tech guy, this is just plain fun to do!)
First off, if you want to build a solid, full fledged rig, there’s a few things to keep in mind before you set down this path.
It won’t be cheap. You are going to want to max out at 5-6 GPUs and a quality power supply. My rig was $2400 all-in.
It needs to live somewhere. It will produce continuous fan noise, and you want it to sit where that won’t be an issue. (though quality fans, like Frozrs or the stock MSI Gaming X’s are very quiet. Mine are near silent and sit at 30% at load!)
It needs internet. You’ll need a constant connection, though wireless will do just fine.
It’s going to produce heat. My cards are air-cooled and hover around 70c. This causes my open floor-plan condo to kick over the AC every 2 hours or so for 10 minutes to keep it around 74f. By comparison, last year our AC turned on for 20 minutes once a day, this does have a $$ cost for me here in SoCal!
So first off, let’s talk parts. You’re going to need to get all this stuff gathered from a few places on the internet before you get rolling. The majority can be bought at Frys, Amazon, etc, but some of the parts, specifically the case and risers, may require some extra leg work depending on availability.
Motherboard: any board that supports 5-6 GPUs will do. Asrock makes a miner-specific “BTC” series boards for hella-cheap in the 60$ range, but mining-whales grab up the market every time they ship and flip them for $300, so good luck. I just went with an MSI gaming series. Do note that bus speed bandwidth is important with this many video cards, so try to avoid specifically the z170 line of boards due to their bus low bandwidth.
CPU: go cheap here. A Celereon will be *too* cheap, but an i3 is plenty power and cheap. If you go with an older board then you can probably get an older i5 or i7 for cheap as well, up to you. In either case, you are *not* going to be CPU mining, it is a complete money sink to try, just don’t do it okay?
Ram: you want 16gb min, but you don’t need high end gaming shit. Quantity not quality here. (Generally speaking you want 3x the number of GPUs in RAM or 12gb, whichever is larger.. but RAM is cheap so just go all the way now if you can.)
GPUs: you are going to want anywhere from 1 to 6 GPUs, the more the better. AMD is light years ahead of Nvidia in this space due to complex technical reasons I won’t get into. If you have some old 7990’s dust those babies off because they are KING,but impossible to find anymore (we’re talking 75% faster than an RX 4/5xx series!). Other wise you want RX 470/480/570/580s. As of right now, 5xx’s run roughly the same hashrate as 4xx’s. but might be improved over time with new BIOS tweaks. However, they use morepower and produce more heat so if you can get some 470/80’s still at this time (it’s hard), go for it. In either case, the only difference between a 70 and 80 is about 1.5 mh/s hashing power and a 70 will use 45% less power, so aim for a 70. Also an 8GB card will get you 1.5 mh/s more than a 4gb, not because of the memory size but due to the types of chips used being better quality on the 8GB resulting in better hashing. In short, get what you can within your budget from the above, aiming at 7990, 470, 480, 570, 580 in that order.
An SSD: doesn’t matter what size, just big enough for the OS and some other stuff, anything over 126gb will do.
Power supply: this is important! There is where those “gold” and “platinum” ratings actually matter. We’re going to put 60%-80% load 24/7 on this supply so the quality does matter here for longevity. You want a gold or higher, and the power output should be relevant to how many cards you have. For a full on 6-GPU rig you can get by with 1100w but 1300w is better.
Case: Obviously you cant fit 6 GPUs in a tower. Additionally, air flow is hella key. That’s why you work with open framed cases for this, meant specifically for mining. I use this one.
HDMI Emulator: You are going to run a headless windows system, and when you remote in with something like Teamviewer, you will be stuck with a desktop resolution matching that of the monitor attached, or a tiny ass 400×600 pixel view otherwise. Plugging in a little 7$ HDMI emulator dongle will trick the system into thinking a full res display is attached instead!
USB WiFi adapter: Pretty straightforward… your miner will need to phone home to it’s pool every 5-20 seconds, and I doubt you will want wires running to wherever it lives.
Here’s my own parts for reference:
Okay now you have the core hardware, let’s talk setup. You’re going to want to hook everything up in the case you’ve obtained. Those risers? the small breadboard goes into you PCI 1x/16x slots and the usb cable connects it to the long board that slots into your GPUs. Each one needs power over Sata connectors, and I would not suggest more than 3 risers per power rail. The GPUs themselves should each be powered via their own cable and rail… this is where having the high quality PSU helps because it should have rails galore… for example mine has 8 dedicated to GPU’s alone. (A power rail is basically just an individual cable all to itself).
Next up, the OS. Now in the word of headless servers, Ubuntu is king.. but guess what? We’re putting Windows 10 on this thing. Yes, I know that sounds insane but bear with me! You are going to eventually want to custom flash the BIOS’s on each card to squeeze out every last Mega-hash per second (mh/s). You’re also going to tweak the settings, overclock the GPU memory, etc. This is all really hard, and in some cases impossible, to do on Linux. Also, the custom Drivers you will need need don’t work so well on there either as while AMD officially supports Ubuntu, their track record is quite spotty.
The nice thing about Windows 10 is that Microsoft really, really, reaaaaaly wants the whole world moved onto it. They are all-in on this OS cycle due to some long term world-domination stuff I won’t get into here, and as a result they were willing to make some concessions. One being, you can install it without a key for free and it’s going to pretty much run just fine for you, so long as you don’t mind an obnoxious watermark down in the bottom right popping up. Now, if you plan to do this long term, I think it’s safe to say you can easily afford less than 1/2 of an ETH worth of money on paying for a key, but for the initial tinkering/learning phase, if you are on a tight budget you don’t have to fork that over right away if it’s going to be a blocker.
So get your parts installed, and with only one video card connected, boot it on up up and go ahead and install Windows like you normally would, and we’re almost good to go. First up, we need to install some very specific AMD drivers. Now, you can install the latest and it’s going to work fine, but if you want to squeeze every last mh/s out of your cards then you want to tweak your GPU’s BIOS. Unfortunately, last year with driver v. 16.12, AMD added a BIOS check to the drivers. Moving forward from there, if you have a custom BIOS on your card, then windows will not detect the cards properly because “reasons”. To avoid this, we’re going to install OLD drivers from AMD, 16.11 ones to be precise since we need the 16.x drivers to detected 4xx/5xx cards, but not past 16.12. You can grab those bad boys right here.
Lastly, there is one other little Windows tweak you need to make: You will want to go into your system profile settings, find the “Virtual Memory” option, and you’re gonna crank that baby all the way up to 16GB. There is a long technical explanation behind this around DAG files, GPU memory etc… suffice to say though, this is a very important step.
Once installed, we want to go ahead and do a test run before we start to screw with the BIOS on them. Both to just rule out variables if something is wrong, and two, to help set a baseline so you know what your improvements did. Go ahead and shut the system down and hook up the other cards now, and then we’ll move on to the software side of things.
Next up you want to get the actual miner software installed. There are a ton of options but one of the best in class is Claymore’s Ethereum miner. We’re gonna grab the latest one from here. As of this writing,. 9.4 is the newest, but the main post is updated each time he changes it, this forum thread is the definitive resource for all things Claymore Eth Miner. Download it, and extract the folder to wherever makes sense for you.
We’ll need to make some quick tweaks to the default start.bat file to make it’ll mine for you and not Claymore’s default wallet first though. By now I assume you have an Ethereum wallet address and have chosen a mining pool to be a part of. Getting one, setting up a wallet, choosing/joining a pool, etc, is beyond the scope of this, so I’m going to assume you have one. Open up the start.bat file and edit the existing command line to read like the following:
(Worth noting here, Claymore has a *plethora* of command line options you can explore and experiment with. Custom tweaks per-gpu.. dual mining for SIA or Decred with unused GPU cycles, fan, power, and heat limiters and optimizations, auto restarts, logging, remote management, you name it! Make sure you dig deep into that Claymore forum thread and learn how to really have fun, once you’re comfortable with the basics!)
You should get some pretty output in a console window that looks similar to this example:
(This is just a generic example, you’re version, card identifiers, and other data will be different!)
If it doesn’t look like this, or you have a lot of red, then we’ve got some errors that you’ll need to triage, most of which should be self explanatory thanks to the error reporting, and easy to find answers to in the Claymore thread you got the file from.
Let it run for a bit, and what you’re going to want to look for, is a reporting of stats showing that all of your cards are working as intended. It will be the teal text (for Claymore that is) reading total speed/total shares followed by a second line with each GPU listed in numerical order with it’s individual stats. Pressing the “S” key will force this to appear rather than wait for it’s timer:
Congratulations! You are now mining Ethereum!
If all of your cards are showing similar numbers then things are looking great! Now, you can call it a day here and let it do it’s thing, and life will be good. If you’re a real min-maxer though, you’re gonna want to do the “fun” stuff and play with the BIOS settings of your GPUs by tweaking the memory timings.
The problem is, this is where it gets hard for me to walk you through things directly, since it all depends on the specific cards you have, and even the types of memory chips among lines of cards. Some cards have custom BIOS files the community has released, ready to go for you. Some have people who have trialed and error many different memory timings and settled on accepted good ones. Others need you to explore on your own, it varies. To be honest, your best bet here is to do a LOT of googling and research before taking any actions… a wrong step here can brick your GPU.
Just to give you a rough idea of what to expect though, I’m going to walk you through exactly what I did for my own setup, which has MSI RX 480 Gaming X 4Gb cards, but I cannot stress this enough… DO NOT DO THE EXACT STEPS BELOW OR USE THE EXACT BIOS FILE BELOW IF YOU DO NOT HAVE THIS EXACT CARD! This is meant to be a learning example to point you in the right direction only. If you flash the wrong BIOS on the wrong card you now own a $300 brick, okay?
In any scenario however, mine or yours, we will want to grab ATI-Flash to copy the BIOS files off the card. I then made a backup of that BIOS in a folder somewhere safe and put an additional working-copy somewhere easy to get to. You can grab the utility here: ATI-Flash Download
I then used the Polaris Bios Editor to open the working-copy of that bios file, and made the following changes to the memory timings before saving the BIOS to a new file with an easy to reference name:
Copy the 1500 strap string to the 1625, 1750, and 2000MHz fields.
Then copy the 1425 strap string to the 1500 MHz, 1375 and 1425MHz fields.
I then re-opened ATI-Flash and flashed this newly edited BIOS back to the cards, one at a time. Once done, I rebooted the computer and assuming all was well, fired back up Claymore’s miner by launching the start.bat file.
Voila! ~25.5 mh/s per card!
This was by no means fully comprehensive, and I’m not perfect… but this should be enough to get the average person going, and I plan to come back and make edits from time to time as needed!
Oh by the way.. if this helped you at all and you want to thank me.. you’re welcome to donate any small amount of ETH you would like to the following address! 😉 0x53705931eC12cAd966B7E9Fdc6FEEc3756B71beb
So i have this little hobby, 3D Printing. You may have seen one or two (or fifteen) mentionings of it before. I love it, and I even make a reasonable amount of money on the side that covers the costs of my hobby and then some.
The one thing I love is numbers, tracking, and data all nicely packaged up in nice report with charts and graphs. Since plastic filament is the one main consumable in this venture, and it’s important to know how much filament you have left on a spool before starting the next print, I made a mildly complicated spreadsheet to help with this.
It’s a fairly straightforward spreadsheet. One tab holds the data for each spool of filament I own, another is a list of every print I have made over the last year (time spent, amount of filament used, profit/cost of prints, and so on. Lastly, a pretty little tab that shows cost/expenses, wear and tear on particular nozzle heads and other consumables. All in all it does the job, but the one thing that bugs me is that it’s just a spreadsheet. I want a web app that I can use anywhere any time, even mobile, to help make it easier (so I’m not batch logging prints days later because I’m lazy or was in a hurry) and more importantly one that supports multiple users so that others can use this same thing. I tried poking around on the web, but just couldn’t seem to find anything of the sort.. I find this kind of weird, but also may just be that due to some of the common search terms that have some crossover meaning here I might just have a blind spot in my google-fu.
Enter my project, AnvilTracker. It’s basically an attempt to convert the functionality of that spreadsheet to a Ruby on Rails app. The fun part?
I have no idea what I am doing.
Basically I’ve gotten it to a mostly functional state. It’s not pretty (virtually no styling or UX work at all has been done) and the functionality is basically 75% of the spreadsheet minus a few cosmetic enhancements. I more or less started off with following some tutorials, cherry picking stuff that did what I wanted (just had a different use case) and renamed some stuff, gave variables more contextual names, and commented things up properly. So far so good..
I think the biggest thing from holding me back is the freaking UX. I don’t have the patience to learn CSS on top of everything else. I’ve also created a monumental backlog of features I want to add, learning them incrementally as I go. So far it’s going ok, but the backlog is getting progressively harder and sooner or later I’m gonna hit my skillcap… oh yeah, don’t even ask me about Unit Tests ok? 🙁
So a few months back I saw that someone finally put together what might hopefully be the first awesome 3D Printing monthly box service. A fan of Loot Crate, I thought it should check it out and see for myself. There was still time to sign up for the inaugural kit before the deadline, so after mulling over the rather pricey (but understandable) $69.99, I committed. After taxes though this brought it up to $76 and some change. :/
(It’s worth noting, we’re talking about a 1kg spool of filament, which typically retails for $30-50 plus shipping on said 1kg and other stuff, so the price didn’t seem TOO unreasonable at the time.)
This is what Maker Box touts:
A new specialty filament from only the top renowned and tested manufacturers including: Made Solid, NinjaFlex, Taulman3D, 3D Fuel, Form Futura, 3Dom and more!
Upcoming materials may include Glow In The Dark, Metal, Wood, Bronze, Carbon Fiber and Flexible Filaments.
Maker tools one would need for tinkering such as screwdrivers, scrapers, pliers, wrenches, duct tape, etc.
Fun 3D downloads, reference sheets and guides.
Exclusive deals on free apps, printers, website coupons, courses and more.
Unfortunately, there were quite a few delays leading up to the shipping, (though they were proactive in explaining the delays to weather, and promised a free gift to be sent out after delivery of the box.) It finally arrived however, and I have to say I’m pretty disappointed. Make no mistake, I’m not unrealistic… I expected like most monthly subscription crates for there to be a bit of wiggle room and certainly NOT to hit some kind of “wow this is worth way more than I paid” jackpot. That said, the results have been lackluster at best.
The box itself was packaged in what was clearly meant to be an attempt to be clever, and I applaud that. The painted on Maker Box name was witty and neat, though the actual execution left it looking messy and cheap. I am not a designer, and I am not well versed in how to critique the application of a design idea to function in a well verbose way, but I do know when something has “missed the mark” which this sadly did. I can tell what they aimed for, but sadly it fell short. (There’s also excess blue paint accidentally smeared over other portions of the box as well.) The mailing label was duct taped to the box, on only two sides, leaving the rest of the label loose and exposed. Anyone who has ever shipped something USPS knows this is a recipe for disaster, with the way they handle packages.
Moving into the box, there were two flyers, one advertising something and the other being a code for 3 months service for Trinpy.com. Not terrible, but considering these are 2 of the 5 things in the box we’re running out of value real quick here. There was also a scraper, a common and much needed tool to remove stubborn prints from the bed, everyone needs one.. but really once you have one or two, you’re pretty much set. No harm in sending one, but it’s definitely of the exceptionally cheap variety and likely less than $1 or $2 in value at most, especially considering a very stubborn print will easily bend this type of metal instantly.
Next up we have an interesting tool. It’s a hand held, I guess, hole boring tool? Hard to say really, especially since there is no literature at all tagged with it? After a little digging around, it turns out it’s a pin drill set. Quite neat, but not really useful for me and my applications. At first I thought it was a set of nozzle cleaning heads, but when I opened the case and looked closer I saw all the bits were drill bits… it’s neat, looks nice, likely of some decent value, but.. well, a miss in my book.
Lastly we have the intended crown jewel of the box, a spool of filament. This one appears to be from 3D Fuel, and is a biodegradable Algae filament. Interesting for sure, but nothing to write home about, and I don’t really consider biodegradable PLA to be on par with say, Glow In The Dark, Wood, Metal, or other “alternative” filaments. That said, I *do* recognize it’s technically a “unique” filament, and currently I understand it is rather pricey compared to PLA as a result. Unfortunately however, even though I am signed up for the 3mm diameter plan (3mm and 1.75mm being the two standard formats) this filament is 1.75mm… making it completely useless to me unless I swap out a significant portion of the hardware on my Ultimaker 2 in order to support it. :/
I tried to look up the filament to get a sense about what it is more, and a better understanding. There is a URL on the box, but unfortunately the site shows this when you visit it:
Definitely not at all confidence inspiring… Even less inspiring is all my research I can conduct indicates this filament ONLY comes in 1.75mm, which means they more or less just said “fuck it” to everyone who signed up with a 3mm plan. Thanks guys.
In the end, I pretty much canceled my subscription about 5 minutes after going through the box. It was a neat idea and if they can refine the process down the road I may get back on board, but right now I really feel like I got far less value out of this than what I paid (especially when their site claims “The MSRP value of every box will match or exceed the price.”, and when combined with a clear disconnect between them and the customer experience, I just felt like this isn’t for me.
I had a lot of fun building my SC2 Battlecruiser. When I was 5-12 ish in age, as much time as I spend now playing video games, I spent then building with Legos. Now with the advent of the various Lego creation software, it’s even quicker and easier to prototype things as well!
I’ve come up with a project that I am really digging the idea for, I just need to flesh it out more. Here is the deal. I really want to build a 3 to 4 foot model of the USG Ishimura from Deadspace. The problem is, a model of that size will cost roughly $800 to $1200 in Lego parts alone. So I have decided to take it a step further.
What if I was to setup a KickStarter fund? Crowd source the money, and then turn around and build it. The entire process I would document and record, making an amazing time lapse of the project as it’s built over (presumably) a week or two. When completed, I hope to have an awesome video that I can then post it on youtube and set it up for ad revenue. I would then throw the model up on auction, taking the proceeds from it, the video, and any left over KickStarter money and donate it all to the Child’s Play charity for the 2012 season. What do you guys think?
Before *anything* happens though, I need to roll up my sleeves and flesh out a solid design plan in one of the common lego design tools. My goal is to have something up by time I return from the UK at the end of Jan so I can start raising the funding then!