[RUN 2 CLOSED][SIDEHACK STICK]GekkoScience Compac Official sales thread

[Mikestang]

(and hopefully teaching the robot how to handle the pot)

Well, you need to smoke every day for a couple years to build up a good tolerance.
 

[1714925757]

1714925757

[1714925757]

1714925757

[1714925757]

1714925757

[1714925757]

1714925757

[cxboyminer]

This is extremely impressiv! May I know the power consumption for the USB when it is mining at 20+GH/s?

[TheRealSteve]

When I changed stuff around the pot to give us clockwise-increase adjustments, there was a single small trace that got left over, hidden underneath the silkscreen, which basically shorts the whole pot to ground and forces the regulator to minimum 550mV output.

These things happen, although I'm surprised it wasn't caught by the ERC

I have to say, that green heat sink did turn out quite nicely as well

[TheRealSteve]

This is extremely impressiv! May I know the power consumption for the USB when it is mining at 20+GH/s?

I don't remember if I took a reading at that level, but at those speeds it's well outside standard USB 3 range (900mA), so you do absolutely need a powered hub that can deliver the juice (or modify a hub to take an external PSU)

[sidehack]

I should have seen it. I made a few changes and sent it off to manufacture before getting test boards because I didn't have an extra two weeks or couple hundred bucks. The changes were entirely cosmetic (like getting the reset cap out from under the heatsink, or aligning the Tx resistors properly, or labeling the backside test pads) and that pot was the only one appears to have not worked right. Friggin' silkscreen layer.

The worst of it is, I bet it wasn't caught on German Guy's fab either so his sticks will probably have the same problem. I've already emailed him about it and hopefully it's easy to take care of.

Yeah, the green heatsinks are pretty darn sexy. I can't wait to get nine of these little suckers lined up in my test hub and see 'em flashing away.

[TheRealSteve]

Well, like you said, nothing a small knife (or a non-penetrating drill hit) can't fix  Literally 5 minutes after I sent out for some boards and started to order parts I realized - from the order process summary - that I had used a part with internal workings exactly opposite of what I needed (think N vs P junction).  Finding a replacement part with similar or better characteristics, same footprint and same pinout was a good hour wasted.  From seeing random jumper wires, PTH resistors crudely soldered onto an otherwise entirely SMD board, parts flipped upside down, etc. in various teardown videos of even high-end electronics, it seems it really can happen to anyone

I'll have to actually find some pink heat sinks now... or see if I can find a place to anodize (or, well, powdercoat more likely) these bare aluminum ones in pink...

[klintay]

(and hopefully teaching the robot how to handle the pot)

Well, you need to smoke every day for a couple years to build up a good tolerance.
 

[sidehack]

Well, it just took 31 minutes to place 5 boards. Admittedly I had to stop it quite often to splice tape. I don't have full reels of everything so I didn't have the foot and a half of blank tape and the end this thing needs to start out with for the auto peelers to work, so I ended up scrounging up spent tape and, well, taping it together. I think I have it ironed out now where I shouldn't have any more breaks, but I ended up burning a lot of time.

Of the five individual test placements (with refined calibration on each one) four of them work and three are currently mining on the Burger address using Novak's custom cgminer driver (still experimental of course). The fifth, well it was giving me fits. I could get either power or data but not both. Not sure what's up. But the most recent two sticks took all of five minutes each to touch up, verify and install heatsinks. It's bedtime so I'm not going to snap apart and test the next five sticks coming out of the oven. Been here long enough already.

I don't know how many we'll be shipping tomorrow. If all 5 work, I'll have nine working which knocks off the first three orders. I'll immediately start working on the remaining 20 on the board, which knocks off the next six. After that is the first bulk purchase, which the next panel will cover (it'll also hit the next 4 individuals. I need to update my queue with recent sales. Anyone that's purchased in the last day, my wallet has to reindex (grumble grumble power outage) so I can't confirm purchases until it's caught up. There are at least two pending.

[edonkey]

Well, it just took 31 minutes to place 5 boards. Admittedly I had to stop it quite often to splice tape. I don't have full reels of everything so I didn't have the foot and a half of blank tape and the end this thing needs to start out with for the auto peelers to work, so I ended up scrounging up spent tape and, well, taping it together. I think I have it ironed out now where I shouldn't have any more breaks, but I ended up burning a lot of time.

Of the five individual test placements (with refined calibration on each one) four of them work and three are currently mining on the Burger address using Novak's custom cgminer driver (still experimental of course). The fifth, well it was giving me fits. I could get either power or data but not both. Not sure what's up. But the most recent two sticks took all of five minutes each to touch up, verify and install heatsinks. It's bedtime so I'm not going to snap apart and test the next five sticks coming out of the oven. Been here long enough already.

I don't know how many we'll be shipping tomorrow. If all 5 work, I'll have nine working which knocks off the first three orders. I'll immediately start working on the remaining 20 on the board, which knocks off the next six. After that is the first bulk purchase, which the next panel will cover (it'll also hit the next 4 individuals. I need to update my queue with recent sales. Anyone that's purchased in the last day, my wallet has to reindex (grumble grumble power outage) so I can't confirm purchases until it's caught up. There are at least two pending.

This is pretty exciting stuff! Even with the mystery failure.

I've never had a (nearly) front row seat for manufacturing before. Thanks for letting us be part of this. It's very cool!

I find it interesting that you're actually mining with the working units. Who do you think you are, Bitmain? (Totally joking of course)

[sidehack]

Three of them are mining. One of those three is the first off the line and I'm gonna keep it forever. The other two, well Novak needed something to test his driver with multiple devices. Yeah it's a bit cheating, but we aren't stressing them I promise. The one we pushed to 450MHz this afternoon was one of my old V0.3 test sticks.

[Luke-Jr]

Can you test that the working production units are good to go with BFGMiner's "compac" branch on GitHub?

[sidehack]

I mentioned that to Novak. He implemented the same ID thing in his cgminer driver and it's working fine. We haven't looked at the new BFG yet but I'll make sure it gets done tomorrow and let you know.

[novak@gekkoscience]

Can you test that the working production units are good to go with BFGMiner's "compac" branch on GitHub?

bfgminer tested working on the compac branch.  I haven't had time to test a lot of options or anything since I just have my craptop at home but I can confirm that bfgminer is working on my setup as well.

I've been using Luke-jrs hex file script to reflash the cp2102s as well and that is also working.  The way it should work is, on stock or old cgminer you will not be able to set freq > 250 MHz, and it will enumerate as an antminer u3, although it should run just as well.  On bfgminer or our branch of cgminer (still in the works, I'll try to post it fairly soon) you can set any frequency and it will enumerate as a compac.

--
novak

[Flep182]

This is extremely impressiv! May I know the power consumption for the USB when it is mining at 20+GH/s?

This is actually a very nice result from just one chip! Makes one think what you could achieve with a big series of them as in the S7 with decent cooling!

[alienesb]

Will this work with Minera? That would make it nice and simple for some of us.

[Flep182]

Will this work with Minera? That would make it nice and simple for some of us.

You'll just have to add the adapted version to the raspberry, but then you're golden. I'm going for that option too!

[prodigy8]

Can somebody tell me how much time does it require to get your money back (ROI) ?

[TheRealSteve]

Can somebody tell me how much time does it require to get your money back (ROI) ?

You won't - these are not intended to break even, let alone give you a profit.

Calculators are whack, but you can start with this sample simulation:
https://coinplorer.com/Hardware/Simulate/55e83fe0226bf410a453eeb0?key=4802c38f-0b0d-42da-a01c-5c288480c125

It assumes $25 to buy it (no shipping added), 16Gh/s at 0.35J/Gh (you can over/underclock), a constant difficulty (won't be so), BTC25/block (won't be so come half-way next year), no pool fee and no electricity costs.  You can see that the time for it to break even in those unrealistic circumstances is still 2 years from now.  Almost all of the aforementioned assumptions changed to more realistic values is only going to make that worse.

[alienesb]

Can somebody tell me how much time does it require to get your money back (ROI) ?

You will get your ROI right away when you turn it on and start smiling.

[chiguireitor]

Can somebody tell me how much time does it require to get your money back (ROI) ?

You will get your ROI right away when you turn it on and start smiling.

This is the way to think about it!