If I have $1M, I get a quarter million chips and build my own miners, then blow the rest on power supplies and breaker panels. But then that's just me.

Novak and I agree about pretty much everything regarding what does and doesn't suck. He's actually a bit younger than I am too - and meaner, if you believe that, but he keeps a tight leash so it's okay I guess.

Anyways, hey guess what arrived this morning!

Some PCBs. I bet you guessed that right. PCBs for the Compac V0.4 and barebones 18-boards. I'll test Compacs first and see how those guys behave, and if they work I'll hand off a bunch to folks for testing. Lighting up the 18-board is going to be a bit trickier because I'll have to verify everything node-by-node while juggling power so it doesn't burninate, so I might not have news on that for a few days.

Bump for we've got a fresh batch of 750W gear in stock (PSUs, boards, whatever).

I'm pretty much out of stock on DPS2K hardware. If Scrappy Do (mentioned above) still has some gear, buy it from him. He got rolled over pretty hard by Technobit and, after buying a bunch of PSUs from us (that did get delivered) and a butt-ton of Minion boards from Technobit (that probably never actually existed, and certainly were never delivered) he's not doing so hot. So if you want 2KW stuff talk to him first.

We're also working on something for the cheapskates, a basic Dell 750W PSU with hard-soldered 18" cables, zero fan speed control and a jump header to turn it on and off. $32 apiece, minimum order ten, allow at least a full day for assembly.

Also I don't know if anyone here is following our miner dev thread on the Hardware forum (https://bitcointalk.org/index.php?topic=995675.0) but the BM1384 (AntMiner S5 chip) stickminer is nearing completion. I should have a few sent out within a few days to eight testers to verify they work properly and do in fact exist, then I'll announce the full product specs and start taking orders. It will, unfortunately, be a preorder scenario (at least initially) because economies of scale and limited-funding dev project and such. But if a USB stick miner that'll do about 8GH off stock USB power and is clockable up to about 12GH off a 0.9A USB port (or higher if you can keep everything cool), with built-in voltage adjustment for over and underclocking, is something you're interested in (if nothing else, for a great learning tool), keep an eye out in the next couple weeks.

I know where about a dozen are, so I'd be surprised if there were only 200 here.

Yes, if a patent is to happen it'll have to happen with the help of someone experienced in that neighborhood already. We can talk about that later, sure.


Yep, the KISS (or we like to say SDR) is part of why I hate high-power-density stuff like KNC and HashFast put out, because when the only way to cool something is with water and the best "home scale" way to do that adds a hundred bucks in strap-on equipment that leaks and dies entirely too often, I prefer to dodge it. Yep, if someone can solve convective flow liquid cooling in a small, clean, reliable and self-contained manner that'd be pretty nice but until then I'm going to lean on monolithic heatsinks and easily replaceable reliable fans.


I'll make boards that can bolt onto the S[odd] chassis, but I won't exceed the power density which their native heatsinks can handle on air cooling. Not until everyone has something else, anyway. The point is to make things cheap, efficient and accessible, so building things designed to fit onto very simple reliable and plentiful chassis is dumb if they don't actually work on that chassis.



Yep, the Prisma is nicely dense. One of the issues with the Prisma, as you've mentioned, is that it was perhaps too dense and the tail-end chips were running pretty hot. I speculate that top-cooling a designed-to-be-top-cooled chip will probably have a better Tcs than the Prisma chips cooled through the board onto a heatsink with no TIM. I also speculate that a design made to be underclocked can probably last a lot longer on those heatsinks than one running full power constantly. I also speculate that there are thousands of bare Prisma heatsinks running around the US waiting for someone to either recycle or repurpose them. I also speculate that the form factor of the board would work nicely with rectangular heatsinks blade-style in a 2KW 5TH 3U machine. It should also continue to work well in whatever the heck liquid cooling scenario you want to come up with. So I think I'll continue that plan for now.


&c, &c...

I'm going to avoid making anything for which liquid cooling is a requirement, at least on my own project time. If you want to commission a design that dense, we can talk numbers, but for the reasons given above I won't ask the community in general to buy watercool-required hardware. If it's something like the C1 where you strap air-coolable boards to a heatsink in higher density, sure that's just fine. I won't build a single S[odd] sized board that pushes out more power than a single S[odd] heatsink and fan can handle though, because that does not serve the community at large which has a million heatsinks and fans available but may not want to pay even more for some newfangled specific setup just yet.

I'm definitely aiming to serve people in the sub-500TH market, but I'd prefer to serve people in the sub-50TH market. Sub-5TH market even. One thousand miners in one thousand homes is a heck of a lot less central than one thousand miners in ten warehouses. It spreads out a lot of things, encourages people to learn and do for themselves, and doesn't assist in rich people getting richer (at least not near as directly). Yet another reason we will not be offering bulk discounts on our Compacs, Amitas and TypeZeros - the price is the price no matter how much money you do or do not already have.

Yeah, I'd really like to know who's in charge of the BE300. That chip is the reason we wanted to build miners in the first place - as soon as the sample data was released, Novak and I spent a full day and a half talking over how we'd do it if we were going to build a miner using those chips.

What I meant about costs regarding patents is, if the budget for the project so far has been about $600 I don't think we're gonna be able to scrape together however many thousands of dollars it takes to file a patent claim. I'm also sorta allergic to BS, and every patent I've ever [tried to] read makes almost no sense because any technical information is completely obscured in stupid legal language. I mean if someone wants to go through what I put together, decide if it's patentable, translate the technical information into Vogon and push it through the system pro-bono, sure we can patent some of the TypeZero stuff. But right now I'm a little more concerned with being one of two engineers in a two-man business than being a paperpusher. You may be right about everything, but that doesn't change the fact that I'm broke and undertrained in legal documentation.

Also, and no offense to PlanetCrypto because I know he really likes liquid cooling, but I really don't care much about it. There's just something reliably low-tech about a heatsink and fan that makes things so much easier for regular people to work with, and since making things regular people want to work with is our priority, I'll stick to considering those needs. That said, if our Spec1 boards are compatible with S[odd] heatsinks, it also means they're compatible with the waterblocks folks are using for them, so there you go. The Spec2 board would be built dimensionally as a Prisma board, which I know PanetCrypto has worked with in immersion cooling attempts. Something like that might be better for an oiltank. If y'all want to figure out the best ways to high-density or watercool the works-with-aircooling boards I want to make, I won't get in the way. But I'm not going to retool designs for MinerEdge-style boxes unless someone wants to specifically commission that project and pay for the dev. The miners I want to make are for people, not corporations.

Also, no progress. Family reunions all weekend. Hopefully I have PCBs in the mailbox on Monday though - barebones 18-boards and version 0.5 Compacs.

I didn't really reverse-engineer much. Most of it was just common sense, though I did look at their board a bit to figure out the inter-node level shifters (and I think I improved on some parts of that just a little bit).

I've standardized a design/interface? Which part?

If by "record time" you mean "over a month behind the initial timeline", then yes.

I think trying to patent anything we come up with will probably cost more than, basically, the whole rest of the project. I'm not sure tuning code is actually a patentable idea, since it's been done before on other hardware. I dunno, I think you're complimenting our efforts too highly - but maybe that's because I'm pretty disappointed so far.

Yeah but now it's going to be really embarassing if we actually can't make per-board tuning work.

Right, but at 850W that 5% adds 13 cents a day at ten cent power, plus the additional 55W of heat needs to be done something with. 60 bucks could get you a Gold-rated 1KW server PSU and not have to screw with a rebate or worry whether or not it had the right cables, since it seems information is pretty contradictory.

60 bucks is decent for an 850W PSU, but you gotta remember you get what you pay for.

Also isn't 80PLUS Bronze rating kinda terrible?

I was just calling attention to that I'd previously stated we'd try to make a per-board autotuning, so assuming it would not exist because nobody else had one and therefore we should make our firmware a lot more complex instead doesn't make a lot of sense.

Also, Phil, there will be no pot adjustment on the 18-chip board. Software adjustment only, but explicit manual control of that software setpoint will be included.

No, because


You can't say cgminer will optimize all boards to the same point until you know what driver code we end up writing. If we can do per-board optimization, we will - and we probably can, because since we're using USB connection, each board should actually enumerate as a separate device. I seem to recall Habanero code could set each die's frequency and voltage differently. This should be no more difficult.

What I'd like to see is a cgminer command line that can set the clock and voltage per board (I really like the -S flag they used to have before hotplug support, where you explicitly stated what device was where) or a -autotune flag with parameters for voltage or clock (fix one, adjust the other) and an error threshold to shoot for.

The stick has a small pot for manual voltage adjustment. Phil's got lots of pictures in his review thread. The TypeZero will be a software-controlled voltage, but we'll make sure to have command-line control well before we have an auto-tuning feature so you can put it wherever you want it.

Someday if I ever get money I need to find a 4-barrel for my truck. That, and recam the engine and find it a 5-speed and regear the rear end and replace the diff with a manual locker. If some of the stuff we have going on works out, I could look into it this winter but heck I probably won't have time. You know how it goes - when you have the time, you don't have the money and when you have the money you don't have the time.

But I prefer ready maintainability over complexity. Complexity, especially digital complexity, adds a lot of failure points not manageable with ordinary tools - as well as cost, both in design and materials. I also don't like automatic transmissions, because it's less fun to spend extra money on extra weight and mechanical complexity to *remove* control from the driver.

http://gekkoscience.com/philosophy.html

Yes, and I need food and sleep so I'm not addressing sales concerns until tomorrow.

I just kinda feel sorry for anyone that bought S4+. I mean 0.6W/GH is pretty great for a year-old chip, but the same machine with BM1384 could have run six strings instead of four off the same PSU and gotten 1.5x the hashrate. Depending on the specifics, it probably wouldn't have cost much more. Just seems kinda strange to me to have the best chip available for half a year with no competition and not make any new products with it, while rolling out stuff that's not as good with old technology. I guess they're taking advantage of having no competition to burn off old stock maybe? But that still doesn't make it cost-effective for anyone at their selling price.

I do hope they roll out some S2 kits with the 18x2 strings. 10 of those clocked moderately could get over 3TH off the 1KW PSU. Or maybe they'll retool the design for a new chip soon and get even better?

About the best Bitmain could feasibly do with an S[odd] size miner with the BM1384 is probably a 17x3 string, getting 1.4TH at 500W and 70% higher chip cost. That's pretty dense. A new chip with 30-40% better top clock would do the job better and probably at lower cost. I'll be surprised if they roll out another S[odd] miner with the BM1384 instead of a new chip.

They could do another S[even] miner with it, and some 18s unregulated strings could get around 3.4TH off a 1KW PSU in a rack box which would be pretty darn sexy.

Even if we build with BM1384, 18-chip boards won't be available for months. If I had a design right now, I'd have to raise the money for a chip batch and then wait for them from the factory. If I had a design right now, S7 would probably be selling before I had the money all together. And since it'll be probably a month before I have the design ready...

But all that is discussion for the project discussion thread. Let's try and keep this one to stick facts instead of non-stick speculations.

Wait, the S4+ is good for something? But I thought it was an overpriced inefficient machine made with tech about two generations old. If they wanted to maximize profits they should have built something actually worth buying.

The sticks will never have auto tuning (for voltage control). The required overhead is stupid for a one-stick miner.

The 18-board will not be strictly auto-tuning. There's no reason to cripple a thing in software when the option for "more control" is already there. The primary reason I got into bitcoin mining hardware to begin with was because I liked messing with the hardware, so why would I give you guys something you can't play with to the max?

I also prefer carburetors to EFI.

So I had a really long and thoroughly quantitative post written up addressing points of discussion from the previous 10 or so posts but the forum ate it and I don't really feel like going back over all those numbers again. Call it a lack of motivation if you want. Novak and I got back to the shop this morning after a 37-hour marathon drive (1630 miles) for supplies to find the power had gone out, all the UPS were depleted so the servers were powered down, the internet wasn't working and the hosting router had eaten its own hard drive, and also I haven't really had anything to eat except Twizzlers and sunflower seeds in about 23 hours. All in all it was a pretty fun couple of days.

Anyway. The point was, the S5 was designed at a time when the hardware scene was pretty different than it is now. It's fairly optimal when you consider the size constraints of the S[odd] miners, prioritizing hashrate first and efficiency second (since top clock on the BM1384 is still more efficient than bottom clock on just about every other chip brought to market). With higher chip density they could do better, but at increased cost (17x3 like in the S4+ could get you 1.4TH at 500W with 70% more chips, whooptee doo).

The 18-chip board will be more efficient than 18 sticks, because you have 1x support hardware instead of 18x support hardware and the regulator should run closer to 95% than the 87-90% on the stick.

I will probably also not release an 18-chip board with the BM1384. I'm assuming, with the dev timeframe we have, that Bitmain's next gen chip will be out and if I can design around that for the larger stuff I certainly will. Maybe if the BM1384 chip prices go way down it'd still be viable, but it'd be starting off at a disadvantage when the next chip will probably see 0.3W/GH around top clock instead of bottom.