Pages:
Author

Topic: GekkoScience is now dabbling with 16nm ASICs for new designs - page 39. (Read 77011 times)

legendary
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
Yep, you've been pessimistic about it for months. Noted. I, however, choose the other route because it's simultaneously gonna be awesome for the community and I don't go bankrupt if I'm right.

So anyway. The stickminer question. Let's discuss that instead of everything else that's already been rolled over a couple times.
legendary
Activity: 2212
Merit: 1001
Sorry, I'm just one guy and I've already discussed three distinct design projects for miners, on top of everything else I've got going on.

Now if I end up able to make the S1 refit boards, four of those on a C1 chassis would probably hit 1000W. Or pack 'em in a rack case and hub 'em together for like 2KW. You know, whatever. But a standalone 1000W miner is not on my list.

Very doubtful you'll get enough chips to go that far........  Sad

Hope I'm wrong,but I don't see Bitfury doing anything but teasing us  Roll Eyes
sr. member
Activity: 490
Merit: 251
Sorry, I'm just one guy and I've already discussed three distinct design projects for miners, on top of everything else I've got going on.

Now if I end up able to make the S1 refit boards, four of those on a C1 chassis would probably hit 1000W. Or pack 'em in a rack case and hub 'em together for like 2KW. You know, whatever. But a standalone 1000W miner is not on my list.

A S1 refit or any 250W board would be sweet.
legendary
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
Sorry, I'm just one guy and I've already discussed three distinct design projects for miners, on top of everything else I've got going on.

Now if I end up able to make the S1 refit boards, four of those on a C1 chassis would probably hit 1000W. Or pack 'em in a rack case and hub 'em together for like 2KW. You know, whatever. But a standalone 1000W miner is not on my list.
sr. member
Activity: 490
Merit: 251
I don't have solid numbers for the chip yet, but if we assume 0.1J/GH and 5W you're looking at 100-150GH.

I will not build a 1000W miner.

Dam, I tried.  Tongue
legendary
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
I don't have solid numbers for the chip yet, but if we assume 0.1J/GH and 5W you're looking at 100-150GH.

I will not build a 1000W miner.
sr. member
Activity: 490
Merit: 251
No stickminer. Build a 1000w miner.
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
Would anyone be terribly upset if I did not build a stickminer around this chip? Interfacing to the ASIC directly kinda sucks, and the minimum buck voltage is at the top end of what's likely the practical range for the chip.

It'd be better to use a proper interface chip, but that's kinda wasteful without multiple ASICs. Of course, multiple ASICs also adds to the cost. Doesn't double the cost of course. I don't know how much it'd add.

So, what's a general consensus - a two- or three-chip stickminer (for, say, nominally $35 or $40 versus the original Compac's $25) or no stickminer at all?

2-3 chip miner built as usb form factor sounds good.   


hash range would be 40 to 80gh? 
legendary
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
Would anyone be terribly upset if I did not build a stickminer around this chip? Interfacing to the ASIC directly kinda sucks, and the minimum buck voltage is at the top end of what's likely the practical range for the chip.

It'd be better to use a proper interface chip, but that's kinda wasteful without multiple ASICs. Of course, multiple ASICs also adds to the cost. Doesn't double the cost of course. I don't know how much it'd add.

So, what's a general consensus - a two- or three-chip stickminer (for, say, nominally $35 or $40 versus the original Compac's $25) or no stickminer at all?
legendary
Activity: 1722
Merit: 1000
This sounds perfect.
hero member
Activity: 2534
Merit: 623
Hopefully everything else cooperates and I actually get past prototyping. Right now I only have enough sample chips to build one; it'd be nice to get enough to make several and send them out for testing and review like I did with the BM1384 Compacs. It'd be really great if I could have 'em out in time for Christmas, but even if I had a working design now it'd be two weeks before I could build one and at least five weeks before production started.

But it'll be worth the wait I'm sure. I'm trying to keep a lid on the excitement for the USB stick I won in the auction  Cheesy
legendary
Activity: 3822
Merit: 2703
Evil beware: We have waffles!
Not entirely OT here but looks like EUV systems I've talked about in the past to create the 7nm and lower nodes are 'almost there'. Great article on semi tech in general and of course focused (pun intended) on the EUV system at Global Foundries.
http://spectrum.ieee.org/semiconductors/devices/leading-chipmakers-eye-euv-lithography-to-save-moores-law
legendary
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
Hopefully everything else cooperates and I actually get past prototyping. Right now I only have enough sample chips to build one; it'd be nice to get enough to make several and send them out for testing and review like I did with the BM1384 Compacs. It'd be really great if I could have 'em out in time for Christmas, but even if I had a working design now it'd be two weeks before I could build one and at least five weeks before production started.
hero member
Activity: 2534
Merit: 623
Glad to hear that the auctions did well and that you're planning to cram so much information into the new pod miner. Its been long overdue to have a miner that gives all the details/features you want.

legendary
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
So I finished up the power and basic layout of a new Bitfury-based Compac and handed it off to Novak to add controls. Looks like he's become a lot busier than expected and I haven't heard anything in a while. The suck part of the stickminer is interfacing to the chip directly, as it uses a quasi-SPI comm with multiplexing and weird voltage levels so there'd be a lot of bit-banging.

I plan to focus on the pod miner, which shouldn't take much hardware dev since the power system has already been messed with a lot in previous versions. Bitfury has their 250nm comm multiplexer chip that takes in pretty much stock-standard SPI and handles all the chip-level comms by itself, so the microcontroller programming end shouldn't be too difficult. That's handy because I'm out of practice on embedded, but I have been getting back into it with various projects the last few months.

So my focus for the next week, when I'm not required for manufacturing, will be to work on the layout for the pod miner and get some sample PCBs sent out for testing.

Here's what I'm hoping for out of the pod miner. It's got a practical power-draw upperbound of 70 to 80 watts, probably more like 50W in normal use. Board dimensions 8x12cm, the chip field will mount up an 80mm fan and heatsink. Power in from a barrel jack or PCIe, signal from USB B or mini. The hardware will include temperature sensing, dynamic voltage control, fan speed control and I'd like to see an input power monitor. An automatic autotune would have to be implemented in the driver (which is not my job) but if we had a reading of input power you'd be able to specify how much heat you could afford (like if you had a 12V 4A brick you could tell it not to exceed 40W draw or something) and it'd dial up the frequency (and fans if necessary) until the overall power draw was maintained at your desired level, kinda like Spondoolies did with their PSU limits.

Once this is done, it should be fairly trivial to extend everything to a larger board. I know someone's already putting out an S5-compatible board, or at least claims to be, but I'm gonna keep working on mine because I think it'll be better. In any case theirs is lacking some features I really want to see, and judging by the 3x power jacks it's going to run hot and loud (possibly incompatible with S1 open-chassis single fan) and won't be compatible with S3 housings at all, so probably not a universal upgrade or necessarily good for at home.

Also, regarding the auction, we brought in some decent money so thanks everyone because it'll really help get this project going. We also got 0.66BTC for vh for all the driver work he did on the failed BW stick project, so that's pretty cool.
hero member
Activity: 723
Merit: 519
would be nice to see something happen good for a change. I am tired of getting repairs for my s9s.

Sidehack could you check the bids out  there is a question I asked on that thread.

My sentiments as well.

I want to put my financial support and purchase from the community rather than a bloated giant that squeezes the market from all angles to make the consumer basically as non profitable and competitive as possible.

I am so in for this development.  If there's some way I can lend a hand, I would/will.

You can help by bidding on some of this gear:
https://bitcointalksearch.org/topic/m.16598805
legendary
Activity: 1848
Merit: 1166
My AR-15 ID's itself as a toaster. Want breakfast?
would be nice to see something happen good for a change. I am tired of getting repairs for my s9s.

Sidehack could you check the bids out  there is a question I asked on that thread.

My sentiments as well.

I want to put my financial support and purchase from the community rather than a bloated giant that squeezes the market from all angles to make the consumer basically as non profitable and competitive as possible.

I am so in for this development.  If there's some way I can lend a hand, I would/will.
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
would be nice to see something happen good for a change. I am tired of getting repairs for my s9s.

Sidehack could you check the bids out  there is a question I asked on that thread.
donator
Activity: 4760
Merit: 4323
Leading Crypto Sports Betting & Casino Platform
Good luck sidehack!  Do your thing.  Thanks for the info.
legendary
Activity: 1484
Merit: 1004
Awesome! Im ready for Canadian Group Buy!  Grin
Pages:
Jump to: