GekkoScience is now dabbling with 16nm ASICs for new designs

[BUZZCOINproject2]

So what would be the range of the 15chip?
Gh/watts over
Go/watts under

[1715393321]

1715393321

[1715393321]

1715393321

[1715393321]

1715393321

[sidehack]

Bottom clock, let's say 150MHz 3.2V (minimum stable voltage for chips)
Let's call top clock 325MHz 4.0V (safe current limit of the buck, may require buck heatsink)

That puts us in a range from 120GH/32W up to 270GH/110W at the chips. Adding fan, buck and PSU losses say 20%.

Final estimate 120GH/40W to 270GH/140W

Please note that to achieve that top-end setting will undoubtedly require modification to heatsinking and possibly input power jacks. Likely I'll stick with a "stock" setting of 200MHz 3.4V for 165GH/60W or thereabouts, but again everything will be adjustable.

[gt_addict]

Sounds freak in awesome for a mining desk tinker toy!! Now an important question....is it going to have blinking lights?

I like the fact the gekkos have them so you actually know they are working. Not just the standard on-off-on perfectly monotonous blinking.

[sidehack]

Probably wouldn't be as bright, and likely a different color, but I was planning on using the same flash-on-share system as on the Compac. As fast as they strobe at 20GH, imagine the blinking at 165GH!

Even the new Bitfury stuff with a microcontroller, I've set up provision for a blinking LED and planned on coding it to flash on share. As you said, it's good to know they are working. That flashing LED is really handy during bulk testing to identify the one stick in a block of 25 that has dropped out.

Thinking about it, I should put on an adjustable fan drive circuit good for 2/3-wire fans (that pulse the power line). I can base it off the PWM driver setup on the new Dell 750W boards (much improved fan speed control) and will add about a buck fifty in parts. On the Bitfury pod down the line, fan speed will be integrated into software.

I likely will also put a thermal shutoff on there that disables the core buck if it sees over 80C (or somewhere safe; maybe 90C if I'm taking a more direct-at-chips reading) and resets after cooldown. It'll be a hardware thermostat, no user changing. The Bitfury will have temp sensing and, if it's done in the driver, user-implementable cutoff setpoints.

Also - just sent off for heatsink bids for both sticks and pods. I'll design the BM gear to take the same heatsinks as the corresponding BF gear so I can use 'em for testing and to save tooling costs. How handy is that? The 2-chip Compac heatsink is the same size as the old one, but drilled differently to mount up a longer PCB with two ASICs.

[gt_addict]

Does sound like its going to be a fantastic little miner. I cannot wait to get my hands on one (at least)!! As soon as I have earnt some btc back from an S9 I purchased (spent all my btc) ill drop a donation in the 1burger pot

[Meech]

You know I'm a fan of flashing lights!  Has that retro feel.  2 and 15 sounds good.  I still have almost a complete s2 for you (blades at least) I can send you if you can use those parts.  Possibly sell my compacs and send you the proceeds.  Lmk. 

[KyrosKrane]

So what if I made a 2-chip Compac at $35 and a 15-chip pod at $60? Those are just ballpark prices, especially the pod since I'll have to find a heatsink and fan for it. If it's possible, they'd run off cgminer and have hardware-adjustable voltage just like the Compac. Nothing fancy at all, I know.

My thought is, the pod at $60 makes the stick at $35 look a little expensive. I'm thinking the stick would be a gift-type item or an impulse buy, not really for even a small home farm at that price. With a choice between two sticks or one pod for just $25 more, I suspect the pod would get more attention.

As for the pod, I'd be willing to go as high as $90-100, with performance scaled up to match. I remember when the original Erupter sticks came out, they were ballpark $100 a pop, and they sold like gangbusters. Even in today's depressed economy, I feel that would be a good price. It also puts some distance from the stick price, and gives you enough room to work in both more performance and more features.

[sidehack]

If you want more performance or features, you're gonna need to wait for the Bitfury gear. Or someone find out how to work Avalon6 chips, or convince BW to sell. Honestly BW chips are no harder to work with than BM1384, except the pads are flippin' tiny.

So what do we think? $30 for the stick and $75 for the pod be more better?

Adding more performance (GH, especially J/GH) is going to mean more chips which means more support components and probably more PCB area. Adding more features mostly takes more time and complexity. As this is a fundraiser to help bolster support for a project which will likely prove to be fairly time-consuming because of its complexity, I'd rather keep it fairly functionally simple. Thinking about stuff like the U3 and New R-Box, neither of those had adjustable voltage or fans so we're technically ahead of the curve already.

[murasame2003]

$30 for 2pac / $75 for 15pod works for me.
 

[gt_addict]

I'm happy to pay a little more. As you've said though you need to do minimal alterations to make a product that will hopefully provide enough profit to fund the main project which is more involved. A U3 style miner would be great I think. If it's got the barrel jack it could be run from a breakout board like the grid seeds etc with 6pin pcie input and screw terminal outputs, which I'm sure many people have laying around somewhere or you could develop one (not sure how cheap it would be to produce but it's down your street for psu bits). I'm just thinking and typing at the same time so these ideas are probably over/under thought  

The pod miner would be the favourite I think as it would be easier to run for people and it wouldn't be as limiting as running things of USB, but a 2pac stick is definitely a winner as it doesn't change much from the existing form and would still be feasible to use on an existing setup i.e swap a single chip gekko out for a 2pac version.

[sidehack]

I'd probably build the pod with barrel jack and 6-pin. I mean I definitely will do barrel jack, and because it can be pushed past the 60W typical for bricks (12V5A is incredibly common) the 6-pin would be handy. Also since, as someone mentioned, a customer could run one or more pods off 12V ATX and use 5V for a Pi controller.

[segaklon1]

Some date when you will have the 2-chip stick for sale, I am interested in buying some of these and collaborating with the project.

[Morguk]

Nice! We want the pod!

[QuintLeo]

I'd probably build the pod with barrel jack and 6-pin. I mean I definitely will do barrel jack, and because it can be pushed past the 60W typical for bricks (12V5A is incredibly common) the 6-pin would be handy. Also since, as someone mentioned, a customer could run one or more pods off 12V ATX and use 5V for a Pi controller.

 Quite a few bricks out there in the 12v 7A and 12V 10A configuration that were used on Gridseed 80 "Blades".

 The efficiency levels on that potential 15-chip pod though are.... not impressive, seems like it should be eating a lot less power.

 4 pods 1080 GH at 560 watts is a quite close match to the S5 (it IS the S5 at a slight underclock give or take manufacturing variation).

  What would it look like if you put all 15 chips in series on one string?

[sidehack]

It would probably be less stable and harder to design for, and also use a heck of a lot more node-level support and data components and PCB space. Basically, it's worse enough that I didn't even consider doing it.

Comparing the top-clock efficiency of this pod to an S5 is comparing an S5 to an S5, since it's basically the same setpoint (800mV). Please also note that I specifically said it would only achieve that hashrate with modification to cooling and risks blowing up the regulator. I will probably never run one that fast.

At stock, you'd be seeing S5 hashrate at 400W. At bottom clock, you'd get it from 320W. And that's assuming your brick efficiency kinda sucks. Bottom clock is almost on par with 135-chip S7.

And yes, the efficiency is not impressive because the chips are two generations old. If I could do better, I'd be doing better. In fact, this whole thread is specifically about all the effort being put into trying to do better. If you want something better, it's way more constructive to try and help than to whine about what's common knowledge.

(sorry, that was rude. I'm kinda grumpy right now.)

[ZACHM]

I would be very willing to get several 2pacs or whatever the names ends up being.

I also have 10 S3's sitting here at my house being used as heaters right now.
If I could upgrade them to something more useful that would be great.

[sidehack]

Sent out a quote for 2pac and pod heatsinks today.

Just finished 2pac layout. Of course it'll need to be gone back over to verify everything before sending off for prototypes. I've got an idea how I want to lay out the 15-pod but that'll be another day. I'll probably start on it sometime and finish it up during forced downtime over Thanksgiving.

Now all I need is a go-ahead from vh about the driver. Setting up a testbench will be trivial. Prototyping probably won't; I've got all the parts in enough quantity to build a few, but even cheap proto PCBs is like a hundred bucks.

Next couple days will be mostly burning through a manufacture backlog so probably not a lot of dev news. Unless Bitfury calls me up or something, which would be pretty cool.

Posted a sales thread with a bunch of miners and miner-related stuff and miner-unrelated stuff from around the shop if anyone's interested. https://bitcointalk.org/index.php?topic=1684735.0

[QuintLeo]

It would probably be less stable and harder to design for, and also use a heck of a lot more node-level support and data components and PCB space. Basically, it's worse enough that I didn't even consider doing it.

Comparing the top-clock efficiency of this pod to an S5 is comparing an S5 to an S5, since it's basically the same setpoint (800mV). Please also note that I specifically said it would only achieve that hashrate with modification to cooling and risks blowing up the regulator. I will probably never run one that fast.

At stock, you'd be seeing S5 hashrate at 400W. At bottom clock, you'd get it from 320W. And that's assuming your brick efficiency kinda sucks. Bottom clock is almost on par with 135-chip S7.

And yes, the efficiency is not impressive because the chips are two generations old. If I could do better, I'd be doing better. In fact, this whole thread is specifically about all the effort being put into trying to do better. If you want something better, it's way more constructive to try and help than to whine about what's common knowledge.

(sorry, that was rude. I'm kinda grumpy right now.)

 OK I'm confused - I thought you were talking about designs for the new BitFury chips?

[sidehack]

Oh. Well I'm in a holding pattern with Bitfury (and by that I mean I've got concepts and layouts and some proto PCBs on the way, and asked them about bulk sales and additional samples three or four times but haven't heard a word back since the end of September) so to help raise money to keep the lights on and get some dev funding for Bitfury projects if they'd ever get back to me I'm working on a couple quick-and-dirty BM1384 ideas - a 2-chip Compac and a 15-chip pod.

The 2-chip Compac should see 22GH around 7W, and the 15-chip pod would see about 165GH off 60W. A 2-chip Bitfury should be more like 70GH at 7W and an 11-chip pod would see about 500GH off 60W. But until I can get some dev funding for prototypes, and more sample chips so I can build more than one of everything, those are gonna have to wait.

In short - the conversation shifted from Bitfury projects to holdover projects while I'm waiting for Bitfury to wake up and make good on everything they've said about chip sales.

[valkir]

With what is happening with the Avaon 7, you are the only hope to bring new gear for home mining.
Hope Bitfury will come back to you soon.
Kepp your good work Sidehack!