Topic: Klondike - 16 chip ASIC Open Source Board - Preliminary - page 63. (Read 435369 times)

~4.5GH, theoretically.

sorry for newb question since i've only now started paying attention but how many GH per K16?

That hash rate would require their optional water cooling, I believe. That adds significantly to the cost and complexity.

So potentially Burnin's boards will be hashing around 35% higher per chip? That is a massive difference and with the difficulty rise even a small difference counts.

Btw, great effort so far!

I think 350 may be possible but is pushing it. Above that you would definitely need to change parts on the K1 board. Replace AP6502A with AP6503A for 3A output (probably a good idea now but beyond what a std USB port can supply anyway), and replace the inductor with one that can handle more current. I haven't looked that up yet but it would need to fit on the same pads.

Clock can be set from cgminer. Voltage is not settable via software with current design. That would need to be a future revision (which I doubt I'll do). You can change one resistor to change voltage.

+1000

Yes. The idea is to be able to flash the unit with a standard USB cable to OC. Initial testing will dictate what our boards are shipped at, but will almost certainly be 300+mh/s.

The over-clocking potential is pretty exciting.

Sorry if this has already been covered. Will we be able to set the clock frequency and chip voltage with options supplied to cgminer, or how does that work?

Hey, clearly it was a good idea.  I doubt I'll be able to win the BitMit auction, but it's all in good fun anyway.

Sorry for the thread clutter 

great news, great work lads! 

In terms of a K1 is 350+ possible with the current design with a hub that provides the right power?

4 chips.

Nice. Great work. Is that printscreen with 5 avalon chips? Sorry for being newbie

Maybe I can meet up with you once I get my hands on the avalon chips.

Some data from Burnin's testing of his boards, may be useful for BKK. It seems that HW errors are not linear with frequency:

I just pushed up the QFN variant. I renamed the k16 directory to k16s and added k16q.
Also added a new K16Q-3D.png image file in docs.

***
I could do a limited edition signed and numbered run of 3x K1 and 3x K16 and put them on BitMit for auction.
They would be shipped out when the first batch gets to me with free registered air mail, or EMS at added cost (usual minimum EMS is around $20). Or something like that. Maybe run for 7 days?

The first batch of K1 will be special WHITE boards. How many I run like that will be determined by initial pre-orders placed on the web store. There will be single and 4up options. The 4 up you will need to cut.

***
I ran my board for a while at 350 MHz but it was getting too high an error rate so I stopped. It was about 20% errors. I think the new board may help with that as I have wires hanging around and no filtering on data/power.

Now I've been running at 320MHz with very few errors for a few hours. Seems to be very stable at 320.



Thanks for handling that. It's all correct info.

How many can be chained depends on testing not yet done. In theory the data rate allows for a lot, the addressing allows for ~100, and the signal length/degradation and also how much impact it has on the master mining. If that works out really bad then we can always use a K1 without ASIC as a master. It has the same I2C function (allowing K1s to be chained without USB, or with USB as power only). A K1 as master would be very cheap since it only needs a few parts installed. (PIC, 2 resistors and only 1 reg. with a few caps).

Might want to keep requests like this to a min and maybe we get back to the thread. LoL .... let the hero win the battle first.

1) Yes. A new chip from Mouser or Digikey (or others) won't have Bkk's code on it, so how will it possibly interface with the ASIC chips? If you don't want to flash it yourself then you should wait for Bkk to announce sales of flashed chips. If you are buying a completed board from a vendor then they should handle the flashing for you and it will "just work".

2) Yes. The boards have the capability of linking together and using only one USB connection. I cannot tell you the max amount but the general idea is that the "slave" boards will communicate their results to the host through the "master" board that the USB cable is plugged into. There won't be any wires involved, there will be connectors on the edges of the boards that will plug into each other. Although if you wanted wires instead of a connector I suppose that would work too.

Yeah... No. I already called it. 

I'm afraid that would be beyond my means.

I agree it should be a limited run, though. Maybe 10 or 20.

Do I really need the flashing as a user? I mean, will it work, if I just assemble the board and try to connect it via USB?

So for users your i2c will allow us to connect few boards (placed one-by-one and connected by wires) into one USB miner?
Then how many K16 boards can we combine? What wires do we need to connect the boards?