The TypeZero boards will be just about right sized for a 120mm fan, if you screw one down to a square heatsink and mount a fan blowing into it it could be a decent quiet machine. Or you could dump them in an oil tank, yes.
|
|
|
Everyone who ever bought an S1 or S3 and didn't resell it or throw it away. I've got about 15 running around the shop. Or you could fetch one for about $30 and refit it.
|
|
|
Okay, guess I should clarify (by repeating a few oft-repeated points). We will be selling complete and fully functional stick miners, and also miner boards which will require cooling and power. The initial board will be designed to run standalone if you have a heatsink and fan for it (we might look into sourcing some of those if there's enough interest) or strapped four at a time onto an S1/S3 chassis. I don't care what fans you use. A stretch goal would be making boards designed to fit on a Prisma chassis. These could also be dropped in a 3U case with appropriate heatsinking and make a 5TH 2KW miner adjustable down to 2TH 600W or so, if someone was into that sort of thing. I still don't really care what fans you use for it. Initially we're only really worried about making the boards functional.
|
|
|
It's quitting time now, but I was just looking at ~8A buck chips for those crazy people that want to crank their stickminers up to 5W output (or more). We sent off for the breakout board PCBs and I expect to have them in hand hopefully end of next week; I want to have an initial stickminer PCB already done by then. I probably won't get any work done on this tomorrow on account of I'm way behind on a router setup I owe a guy, and maintenance on our own hosting network.
Also, tentative stickminer model names are Compac (single chip) and Amita (two-chip). Likely we'll shoot for a green heatsink, top-mounted if possible.
|
|
|
None of the above. We're not selling breakout boards, nor have we ever stated an intent to do so. We're making them for in-house use for testing and prototyping. If enough folks are interested in playing with BM1384 themselves, and if the boards work as intended, I can release the design files so folks can make their own.
We will not be selling breakout boards. We will not be selling chips. We will be selling complete miners only. I'm just keeping people updated on progress, since this is a project discussion thread.
|
|
|
So because we're cheapskates and buying 10 half-sized one-chip breakout board prototypes is a fair bit cheaper than five two-chip breakout board prototypes, it looks like we're buying 10 half-sized one-chip breakout board prototypes. ![](https://ip.bitcointalk.org/?u=http%3A%2F%2Fgekkoscience.com%2Fmisc%2FBM1384_breakout_v0.1.PNG&t=663&c=x9kiI_Hug1hueQ) Novak's trying to push the order through right now. I'm gonna start on a bit of actual stickminer design tonight since I have other things to do that'll take a lot longer than one hour but quitting time is in one hour. I'm really looking forward to getting this thing moving and playing with some chips.
|
|
|
I've been talking to your people for a couple weeks and that's literally the first time anything NDA has been mentioned. I'll get back to him about that I reckon.
|
|
|
Righto. The price Janet gave me on 1000 was actually about 20% lower than what Yoshi said, which I'm certainly not going to complain about. What would pricing and availability be like for 10,000 chips? We're trying to figure out a good timeline for things. We're sending off for prototype breakout boards today, and later this week I'll be working on a stick miner design which we'll modify as necessary once we have breakout boards in hand to test chip functions. 1000 chips would be okay for a stick miner batch, but as the primary goal is a larger (18-chip) board which hopefully would be designed and prototyped within a few weeks after the breakout boards are here, it'd be nice to know if we could get a large batch of chips to run out a large batch of full boards within the next month or two.
If you need to message me and keep some of the numbers out of the public eye that's fine. I'd just like to get an idea of what's possible.
|
|
|
Janet told me that 1000 chips could ship from stock, but that in general all chips were set aside for miners under construction already. Yoshi did not say anything about wafer orders, just also that 1000 chips could ship from stock and to contact Janet for purchasing.
Every time I've asked for protocol details I was told to just look at the code for the U3 and such. Everything appears to be standard UART comms so it's probably all implemented in cgminer drivers anyway. I think the only real tricks would be fan speed and temp sensor interfacing, and core voltage adjustment. But code is Novak's job, so I can't really say anything for sure.
|
|
|
Might be worth looking into, but when I saw their tapeout announcement I wondered how likely it was they'd have chips in April, let alone a deliverable product. Also, we don't really care about scrypt at all.
|
|
|
What were you told for Avalon pricing? I might be interested in their next chip in a few months, but the pricing versus performance on the 3222 isn't really gonna cut it. By the time we'd have miner to market the chip would be almost a year old and 1.5 generations behind. I'm looking at getting a thousand of BM1384 at the start, and was told [redacted out of courtesy for Bitmain]. Bulk availability I'm told MOQ of 25 wafers, minimum two month lead time, wasn't given a price.
I have no problem teaming up on discounted chips, especially if the end result is home miners get two more decent machines on the market and the net cost to them is reduced.
|
|
|
The cost per unit will be as low as we can make it. Sure we like having money, but we also like taking care of customers and home miners have spent too long getting shafted.
USB sticks are definitely a go. The question is whether we want to try to move on a TypeZero with BM1384, or wait the extra months for a BM1386(?) board instead. We'll probably be redesigning for the newer chip anyway, as soon as we're able. But will it be worth the time and investment to continue a 1384 design, at least for a single batch. I guess that's something we'll need to figure out over the next month or so that we're prototyping.
|
|
|
Hey, don't overlook the guys that basically started the server PSU breakout board industry for consumer miners. We still have some stuff worth looking at. https://bitcointalk.org/index.php?topic=940317.0Any large PSU worth anything will have active PFC, which typically gives you a power factor of .99 or better. That's one of the reasons why efficiency is better with 240V, because active PFC is a boost converter that always drives current proportional to line voltage into a >300V DC bus. If the line voltage is already getting into that neighborhood it doesn't have to work as hard to push the power uphill. Also, as mentioned, currents are reduced for the same amount of power consumed, which resistive loads (from wiring and such) are proportional to current squared - meaning if you halve your current you quarter your resistive losses. Another benefit not always directly intended is load balancing. Your house's mains feed transformer will have two 120V legs, and drawing different amounts of current from each leg is hard on the utility system so hooking up high-demand machinery to 240 will mean its current is being pulled from each phase equally.
|
|
|
No they haven't. All I've been told is [something something new chip in the future sometime maybe, redacted out of courtesy to Bitmain]. I would bet the comms are pretty similar to the BM1384, which was identical to the BM1382 as far as I can tell, but the pinouts and packages haven't yet been the same from one chip to the next so I don't expect them to do it again. Hopefully it wouldn't take long to adapt a functional BM1384 board to a BM1386 (or whatever) board if they don't change their IO protocols and voltages, basically just redrawing footprints and rerouting some lines.
|
|
|
So, looks like we'll be able to at least get a short run of about 1000 chips from Bitmain. The price is higher than I was hoping but still in the neighborhood of "reasonable" for a stick miner - which most of them will go on. We'll keep some for building probably ten TypeZero boards so we can send them out to some reputable tester/reviewers, and then have about 800 chips to put between one- and two-chip USB sticks. I've looked back over and verified all the breakout board PCB, so hopefully we'll be ordering a prototype batch of them today and have them in hand probably end of next week. Between now and then I'll run out a tentative stickminer PCB and get dimensions to heatsink people. Once the breakouts are populated and tested, we can finalize our stickminer design (having had a chance to find any problems with how we understand things to work) and get some test PCBs of them, and then finalize the TypeZero board and get test PCBs of it as well. We're probably looking at having stickminer prototypes in 3 weeks would be nice, but 4 is more likely.
So, question. If it's going to be about 5 weeks before we have a prototype TypeZero board, even if we ordered a bulk run of chips at that time (with 2 month leadtime before delivery) it'd be about 3.5 months before we had chips and boards in hand to start building. That puts us right around the time Bitmain will be announcing the next better chip. Would anyone buy the TypeZero Spec 1 board with BM1384 chips, or would we end up sitting on them and going broke? As soon as information is available for the new chip we can start designing for it instead, but we wouldn't have access to samples and prototype PCBs until later and it'd shift product availability back to maybe September instead of July.
|
|
|
Dang. If they were still intact I'd have bought a couple off ya for the museum shelf.
|
|
|
I never said BlackArrow. I said Bitmain and Canaan-Creative (Avalon). You couldn't pay us enough to work with BlackArrow hardware at this point.
|
|
|
Please RTFT before asking questions which have already been answered. Minions are also BGA and require ~100A VRMs. We very likely will never design a miner around that type chip (which appears to rule out Spondoolies' gen3 chip) because it makes literally every step of the process unnecessarily complex. Complexity means added cost and also increases the odds of machine failure.
We have never expressed any interest in working with BlackArrow chips, anywhere, at all.
|
|
|
Yeah, using a CPU cooler makes zero sense for the chip style that everyone except Spondoolies uses. We likely won't be building any boards conducive to CPU coolers anytime soon.
|
|
|
Yeah, I know Marto is the Technobit guy. I've not seen a Technobit product that had Bitmain chips on it. I have not actually seen any Technobit products personally, though we were supposed to have about 50 boards for hosting last August. I'd just as soon not do business with Technobit if I don't have to. I don't trust them at all.
|
|
|
|