Topic: GekkoScience BM1384 Project Development Discussion - page 47. (Read 146720 times)

I would say that anything you can do to make it as close to a plug-n-play swap with an S1, the more you will move when resale time comes.

I'll probably end up designing the ArchetypeZero to about the same width as a TZ S1 board (so half width of an actual S1 board) but it'll be longer/taller and screw holes in different places. So it actually would fit on an S1 but you'd have to redrill the heatsink or figure out some other way to attach it. Unless... I put S1 screw holes on it. Which might be possible.

So these wouldn't fit on my S1 heatsinks then, it doesn't look like.

yeah I am thinking I would build a custom one in a case.

1 rasp pi
1 usb hub
1 atx psu

? boards 4-6

heat sinks
fans
case

The New R-Box ones?  Chunky rail ones: https://farm6.staticflickr.com/5599/15274308559_2107d80ed6_o.jpg

You know I'd be interested, but numbers depend (mostly) on Amita plans.

Does this still apply?  Wouldn't mind a GekkoScience original or two, myself, but if you're not catering to the European market, would be good for people to know before filling your inbox

4 R-Box cases are very similar in size too a m-ITX case.

Anyone interested in large batches of Compac sticks, please PM me. I want to get a ballpark for how many I need to make. This is not a reservation or pre-sale.

What do the heatsinks look like.  I have so many spare parts in house.  I am wondering if I could build a 4 to 6 board  unit in a pc case.

6.375x6.75 inches

What are the dimensions on the r-box boards?

Alright, so I talked to Novak a bit and we have an idea. We could build a 2x7 bucked string that would get 230GH at 300MHz off 100W machine-level. Clocked down to 150MHz you'd see about 115GH off around 40W, and if you could keep it cool 400MHz would run you over 300GH but pushing 160W so you'd definitely need to beef the fans. If we wanted to make voltage software-adjustable we could use it as an intermediate build between stickminers and the TypeZero, with mostly full software and such, and still use BM1384. It wouldn't take much to retool the design a bit for TypeZero once new chips are available. I posted a poll at https://bitcointalksearch.org/topic/new-r-box-upgrade-kit-1114329 in case anyone wants to submit an opinion.

Naw, you knew exactly what you were doing. Presenting awesome ideas at midnight is always gonna do that.

sorry... didn't mean to throw that idea up. just count them 0s salted on the heads of them hash-string sheep.. think about them boards tomorrow, plenty of time

Yeah, anything I design will have adjustable voltage integral so you can downclock without the lameness that is "pencil modding" and double the effective service life with no extra cost.

I'd set the stock power dissipation to within a few watts of the actual board, which should ensure the heatsinks would work. Cranking it up higher will be your own concern.

A single run, like a couple hundred boards, would increase the per-unit cost from materials and dev recoup. Probably still come in under $100 a board, which seems kinda high to me for the hashrate but whatever. Also it's past my bedtime but now y'all got me thinking so I'm gonna be up for another hour scheming instead of being passed out.

i was thinking one of as in a single run to give us dead NRB owners something to use.

it would be a little harder to retrofit a 120mm fan in, you'd have to cut some holes (make it look like a Prospero X1 clone), or even 2 more 72mm fans on the other end without shrinking the board or strapping them to the outside, none the less, the 2 fans pull a lot of air no issue there, the sinks though, i many have to measure up and calculate the potential Watts/Centigrade dissipation (used to do this a lot with old amd atlonsXP heatsinks, with my maaad overclocking skillz, so now im a bit rusty on the calculations)

but the board itself looks to be used in R3-Boxes, and the RK-box, same layout, almost all are the same sinks, different controllers

yeah, i would be buying them if they were available. id be up for 3 of them off the bat. then i'll buy up as many NRBs as possible

also Screw ROI, you cant predict the future, i already made 6BTC off a 200kh/s(scrypting) videocard many moons ago, they all said i couldn't do it.
who says when btc halves its reward things don't get easier? price rise, hashing drops, much fud in the fan. The GiggleWATdolla police can shove it, a better designed miners will outlast and outplay the cheap cost cutting fire-per-watt hardware, yeah, im looking at you AsicMiner Tech >_>

i was looking into dev'ing up my own board for the NRBs, maybe throwing a spondoolie chip or 2 (stealing it off the "lotto dice" thing or cramming up the BE200s in the same footprint (instead of 12, maybe 24). but i don't have the resources atm.

By the time I could roll them out, Bitmain (and possibly Avalon) would be on new chip generations and I'd probably have to retool anyway.

But yeah, under $100 per board shouldn't be a problem.

If its reasonable to get 250Gh/s per board and stock R-Box fans are enough to keep it cool, 4 boards per 800watt PSU, under/at a $100/per board delivered to Texas,... I got 12 boxes available! By the time you could roll them out, I'd have the spare cash/btc too.

Screw those guys, it's my thread. The talk so far has been on stickminer dev and there actually hasn't been much talk about ROI anyway.

Which, by the way, if anyone wants to talk about ROI make sure you don't get it confused with positive ROI and/or breakeven. That's super annoying.

I'd be interested to hear from AJRGale, CrazyGuy and maybe a few other folks about this particular side-venture. In the meantime I'll focus on the Amita and TypeZero boards. An R-Box board could be built off an 18-board base design if we wanted it to be a bit more complex but ultimately probably better.

If I put a number on it, then this thread would get run-over by the GiggleWAtt-per-dollar police!

I like selling bare-metal boards.

So, what's it worth for them to exist?


Also, let me re-evaluate this for a minute. If I do string, let's say we run 12 chips instead of 15 and runn a dual-phase buck like the 18-board. That drops the VRM parts cost to about eight dollars (instead of about 20) and drops a further ten off chip cost. Let's say stock setpoint is 350MHz at 780mV. This gives us a stock hashrate of 230G off about 120W. Not quite as good but saves a lot on board cost - about twenty dollars. Shelf price maybe $70 a board? Takes more dev time, however.

Any of this is contingent on the existence of chips.