GekkoScience BM1384 Project Development Discussion

[sidehack]

We don't plan on building any miner without voltage control. Even our stick miner had it integrated, there's no reason to leave it off bigger machines where being able to undervolt is significantly more valuable in the long run.

[1714209291]

1714209291

[1714209291]

1714209291

[RichBC]

Very pleased with your stance on voltage control. But... are you going to go with a cost effective, efficient string design and expect people to find an efficient supply that can have it's voltage adjusted? Or go for a more conventional, but less efficient and more expensive VRM? Or something else? 


Rich

[sidehack]

When I say voltage control, I mean software-controlled integrated voltage control on a per-board basis using cgminer flags. I make no assumptions about what PSU anyone uses, and certainly don't mandate you have special equipment to unlock any special features. With the current design I'm working on for the 30-chip board, you should be able to run it off anything between about 9 and 24 volts without problems.

[RichBC]

Fantastic... I guess the background problem / opportunity here is that Bitmain has very little interest in providing voltage control as all it does is complicate, add cost and add life to a product, and why would they want to do any of those things when they can just release a new model?

Their history of voltage control is very poor and shows very little commitment. S3 there was a board version but later withdrawn. S4 some excelent VRM modules but again firmware support withdrawn. S5 had great promise with the 9V , 0.2J/GH announcement, again withdrawn. S7 no word yet except I see "Rated Voltage: 12V+5%, should not be less than 12.0V"

We are in your capable hands. 

Rich

[sidehack]

"Capable" has yet to be proven. All I've done so far is put out a decent stick miner. Like I've said though, I do have parts inbound to test out all the control features on a 4-chip. I'll build it around our test BM1384 chips to prove the things, then move onto a 30-chip with BM1384. Voltage levels and some parts will be a bit different but that'd put us about 90% of the way to a final design once BM1385 become available. Once that's done, maybe then we can start passing around congratulations.

[VirosaGITS]

Sound like a plan, sidehack. But if you add leds, which could be useful, i hope you put one that are dim. The lightshow on the S1's is kind of ridiculous xD

[heslo]

Whatever happens, sidehack... thanks.

Seems at this stage you're the only guys fighting for the home miners. I've got no S1,3 or 5's left over but I hope somehow I can use your new boards (if they do get made) by getting creative with some heatsinks (perhaps like Bitmain do it now on the S5+ and S7)

I'm still excited for your sticks to arrive, I've got a couple on the way and while they're not going to make me a lot of BTC it's still going to be nice playing with some new hardware for a change

[sidehack]

Our robot should be here Monday or Tuesday and I'll immediately start working on getting those sticks out. Heatsinks arrived Friday and they look pretty good. Less blue to the color than I was expecting after the somewhat "evergreen" sample stock we received. I like it. Maybe tomorrow or Monday I'll have pictures of a couple up and running. Novak's test stick already has the new heatsink outfitted but he took it home to work on driver code over the weekend.

Between now and when the robot arrives, I'll be using my time hopefully to start getting board control systems roughly laid out (checking on chip pinouts and figuring out what wires to what) and try to finish up initial design of the TypeZero power.

I probably won't bother to put together a test implementation of the improved level-shifting I had come up with, and just use the standard method I've tested with so far (which is a slight alteration of what's used on the S5) for 1384-based prototype boards. It's fairly moot since it appears the 1385 has integrated level shifters.

I don't think we're the only guys still fighting. PlanetCrypto is a somewhat industrial customer, but he's been working very much for the community as a whole. Guys like Jabberwork are also coming up with a lot of great stuff. There's philipma1957 and TheRealSteve and a lot of people like them who aren't always working directly but help out the guys that do with advice and resources.

[heslo]

I think that's the best thing about the bitcoin community; the ingenuity and sharing spirit around it. Of course there's always those out there looking to abuse it and twist it but overall, it's a great place to be!

[Mudbankkeith]

The Pi's are out there, all we need now are some chips to build a Fan Chassis(tic) Miner  

[PlanetCrypto]

"... PlanetCrypto is a somewhat industrial customer, ..."

With the nonavailability of affordable efficient hardware, something with "legs", we are becoming less "industrial" with every difficulty increase. If the trend that the community has experienced over the last six months continues we will be out of the game, so to speak. It is in our best interests, as a company, to promote and assist in anyway possible GS et. al. Simply put, it's a matter of survival for the BTC mining department/division of our organization.

I ran the ROI calculations on the new S7, and for us, it ROI's in early May of 2016. By the time the block reward drops to 1/2 an S7 exhibits $200-$300 net retained earnings on a roughly $2000 investment. That's crap. Effectively amounting to ~10% APR. There are other divisions of this entity that could use that same capital investment and return much higher retained earnings over the same period (Solar farming, VPN, Public Electrical Vehicle charging network, etc . . .).

To distill it down to it's simplest, if one or more chip manufacturers does not sell state-of-the-art chips for inclusion in a GS hash board (whatever that may be) we will be forced to shutdown our BTC mining operations and pursue other profits centers.

I work under moral, ethical, and contractual obligations to run this entity in the most profitable manner.
Sorry guys/gals, but we do it to show a profit for our shareholders and not for altruistic reasons.
I can absolutely guarantee that if/when profitable BTC mining becomes impossible, I'll shut it down in a New York nanosecond.
Till then I'll investigate any avenue that permits this division to continue operations.
Right now, as of today, this instant, I believe our best shot is to obtain some chips from either Innosilicon or BM1385's and have GS manufacture hash boards.
The next best "best shot" is to pursue fabricating our own chip(s) or mother a community effort to the same ends.

The S7 is NOT a viable option for us and our operations.
And at ~$2000 each we will likely not buy one, even for testing, favoring to spend that capital elsewhere.

What I'd like to see/buy is a retro fit S(odd) physically sized board, USB interface, that is voltage and clock adjustable, and has current monitoring at either the chip, string, or board level. That hashes somewhere in the .05W/GH/s to .3W/GH/s range.

[VirosaGITS]

The Pi's are out there, all we need now are some chips to build a Fan Chassis(tic) Miner  

As far as i know there would be some firmware work to be done to make sidehack's PCB work on it, but yes that and connected to a PC through a USB hub sound great!

[sidehack]

Firmware work what what now? My boards will work on a Pi so long as you compile cgminer with the right driver. That's the only requirement.

PlanetCrypto, I'm glad you're looking for "a retro fit S(odd) physically sized board, USB interface, that is voltage and clock adjustable, and has current monitoring at either the chip, string, or board level. That hashes somewhere in the .05W/GH/s to .3W/GH/s range." because that's pretty much exactly what I want to build. Our motivations are a balance between profit and altruism - we want to see the community prosper instead of, you know, being raped then ignored. We also got bills to pay, and the crazy stuff Novak and I want to do in the future requires that the business grow a bit to afford better research and equipment. The amounts of money required to pull off manufacturing batches of things feasibly or affordably unfortunately pretty much requires us to have some sort of industrial partner to help foot the bill, but I figure if doing good for one non-crappy entity can allow me to do a greater good for a thousand other folks it's probably worth it. As far as non-crappy industrial partners go, you seem to fit the bill pretty well.

I'm not sure how feasible chip-level power measurements will be without some form of actual chip integration, but board-level actual DC power output is pretty straightforward. My plan is to try and iron out the rest of TypeZero power systems design today after lunch so I'll see what kind of hits I'd expect from including that. I was already thinking about it anyway, and the onboard control IC certainly has enough spare interface pins and ADC channels to handle it.

If I had $2000 to spend on something to make back 10% of my money in one year, I'd probably get DPS2K interface board PCBs. Or maybe some sheetmetal equipment for case manufacture once we iron out the rack design. Maybe Novak would rather have some milling equipment, which'd definitely be fun. Certainly I wouldn't buy an S7. I'll buy one next year when they're getting tired and cheap, for the museum shelf.

[VirosaGITS]

Firmware work what what now? My boards will work on a Pi so long as you compile cgminer with the right driver. That's the only requirement.

You said the other USB project might be parallel but not exactly what you would need. Ergo there will be some software side work to be done.

[sidehack]

I don't know what that means.

According to the first post of this thread, "The primary goal is to build a simple board which would be USB-connected to a controller, and capable of adjusting both core voltages and clock speeds using cgminer flags ... with a 4-port USB hub and a Pi or something as the controller, and you have an 'S1 Upgrade Kit' ".

According to a post some 26 hours old as of right now, "My boards would be USB-connected so you could hook them to anything with USB and cgminer (including the Pi)."

The intention has always been, and will continue to be, that the boards will connect via standard USB to any controller which also has standard USB, and when cgminer is compiled with the driver-gekko we will write, a set of command-line flags will control all aspects of the board without requiring any other changes. The only software work to be done is our job, and no firmware changes will be required for the board to go between a Pi or a Cubie or a computer running Windows - provided you have the right cgminer. If the "firmware work" you're talking about is in fact writing that cgminer driver, then it looks like we're arguing the same point with different nomenclature.

Right now I have two "USB projects" - a quality hub (which is stalled, and not relevant) and the control system for a mining board, the implementations of which for the 4-chip and 30-chip boards will be almost identical. Differences between the two will be number of temperature sensors and the specific electrical hardware to control fan speed (for 3-wire versus 4-wire fans). I plan to dev the controls on the 4-chip, and it would take almost no time at all to port them to the 30-chip.

[VirosaGITS]

compiled with the driver-gekko we will write

That ^. That would be the hardware interface software that's missing. Yes? The missing part with PCB connected to USB to RasPi/PC.

I was just telling the other user that there was a software/firmware part missing for this to work. Nothing more.

[sidehack]

Okay. From my end of things, "firmware" refers specifically to the embedded code on the board itself rather than the higher-level generic cgminer code ("software") which has nothing to do with the PCB itself. Hence the confusion. Machines with stock controllers built in might call their software firmware, but that's because they're treating the controller as an embedded module rather than as a generic computer attached to the system. So, nomenclature issue.

It happens probably at least once a week that Novak and I will end up arguing over something but we're both arguing the same thing without at first realizing it because we're assuming the words we use to summarize it mean the same thing to the other guy. Taking a moment to step back and define a few terms solves the problem right quick.

[VirosaGITS]

Okay. From my end of things, "firmware" refers specifically to the embedded code on the board itself rather than the higher-level generic cgminer code ("software") which has nothing to do with the PCB itself. Hence the confusion. Machines with stock controllers built in might call their software firmware, but that's because they're treating the controller as an embedded module rather than as a generic computer attached to the system. So, nomenclature issue.

It happens probably at least once a week that Novak and I will end up arguing over something but we're both arguing the same thing without at first realizing it because we're assuming the words we use to summarize it mean the same thing to the other guy. Taking a moment to step back and define a few terms solves the problem right quick.

Cheers. I understand. To me drivers are pretty low level, so maybe firmware was not the right term. Maybe i should englobe everything as "code" haha.

[sidehack]

"Sorry to inform you that we don't have spare chip to sell. Please kindly consider the miners directly."
- Bitmain

"I do not want the miners Bitmain is currently offering for sale. I'm working on a design for a miner which meets needs your machines do not. One of the businesses I am working with is currently using S3 and S5 boards on C1 waterblocks for both residential and industrial heating applications and would like to continue using these installations with upgraded electronics, but your new S7 does not meet the mechanical specs (in addition to other system requirements) to be used. The board I am designing will do the job. I would like to use your BM1385 chips, as I had a fair amount of success working with the BM1384 chips I purchased earlier this year. Your chips are good and I was very glad to see the design improvements from the previous generation, both in efficiency and ease of interfacing. If I can't purchase sample chips from you, which will lead to a final design and likely mean factory batch orders of chips purchased from you in the next several months, well I'll have to buy someone else's chips instead.

Would Bitmain rather have me and the groups I'm working with as customers, or would they rather see us purchase from Bitmain's competitors?

I am right now designing around the BM1384 and intend to retool the final design when new chips are available (from Bitmain or someone else). Having chips in the near future would be nice, but I could wait about four weeks (when your S7 is supposed to start shipping) without causing any real delays. I would be willing to pay up to US$200 for fifty BM1385 shipped to my shop in the US. This is less than one S7 board worth of chips."
- me

[chiguireitor]

"Sorry to inform you that we don't have spare chip to sell. Please kindly consider the miners directly."
- Bitmain

What a bag of dickheads

That would be like suggesting buying a full V8 engine when all you need is a timing chain.

Nonsense