Klondike - 16 chip ASIC Open Source Board - Preliminary

[xyzzy099]

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 he only mentioned K1's couldn't hash that high. I imagine he will get similar hashrate out of the K16's as burnins.

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

~4.5GH, theoretically.

[1714199719]

1714199719

[1714199719]

1714199719

[1714199719]

1714199719

[1714199719]

1714199719

[BkkCoins]

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 he only mentioned K1's couldn't hash that high. I imagine he will get similar hashrate out of the K16's as burnins.

The K1 is limited by power on a std USB port.  Other than power supply there isn't any reason that the Avalon chips will hash slower on the K16/K1 than other boards. Noise has been an issue but with the new board revision I hope this is behind us. The K1/K16 designs don't clock lower or have some firmware limit that prevents them hashing as fast. The data rates outside the ASIC are quite low (2-3 MHz) and not a speed limiting factor up to short 420 MHz tests I did. But the revision 1 prototype board had no USB or PLL supply noise filtering and that was tested to be very significant in error rates.

The burnin numbers above show that water or air above 409 MHz isn't feasible, and that some clock rates have instabilities within the ASIC. Running at higher voltages may give you some few percent extra hashes but also puts more stress on the ASIC, and could lead to early failure over time. I'm pretty sure that the electron migration issues in GPU over-volting also apply yo ASICs, but I'm no expert in that area.

[BkkCoins]

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

I've been running steady for most of today with very low errors at 330 MHz. No noticeable heat issues or any anything else to report. Chips seems to be running 50-54C. Heat sink is only slightly warm. Fan is 80% but inaudible.

If that works out as typical on a full K16 that would be 5280 MH/z.

[CumpsD]

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!

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

Burnin has posted speeds with air and watercooling, think he gets stable 409 on air

[SebastianJu]

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!

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

He tested water and aircooling:

Fyi, the K16's of BkkCoins are finished. Looking eagerly to see Bitburners get into production

I hope some overclocking tests are done too. And i would like if burnin could make a overclocking test with watercooling so one could check if its worth.

Done some measurements as requested.
But again didn't dare the 1.3V. (maybe tomorrow)
But i am pretty sure that with 1.3V the 450Mhz could be doable.
The stabilizing effect of voltage adjustment is obvious with this dataset.
Core Voltage can be adjusted in fine steps just like your GPU, not just 1.1/1.2/1.3 like planned before.
The Water cooler did not have the impact i had hoped for.
But the undervolting tests were encouraging.
See for yourself:

ambient: 25°C, water temperature 26°C
Chip Frequency - Chip surface temp, cgminer board-temp, Core voltage, power consumption, comments

Air cooling:
282 - 45.8, 40.5, 1.2, 46W, stable
340 - 50, 43, 1.2, 56W, stable
372 - 51.5, 45, 1.2, 61W, stable
390 - 52.5, 46, 1.2, 2% hw errors
390 - 53.3, 47.5, 1.25, 70W, stable
409 - 55, 48.5, 1.25, 2% errors
409 - 56.5, 51, 1.28, 77W, stable
434 - 57, 52, 1.28, 82W, 10% hw errors

Water cooling by Anfi-Tec:
282 - 40, 30, 1.016V, 33W, stable 170Mhash/watt undervolting
282 - 42, 30, 1.085V, 37W, stable
282 - 45, 30, 1.20V, 46W, stable stock 122Mhash/Watt
340 - 48, 30, 1.20V, 56W, stable
372 - 50, 30, 1.20V, 61W, stable
390 - 53, 30, 1.25V, 70W, stable
409 - 54, 30, 1.28V, 77W, stable
434 - 57, 31, 1.28V, 82W, 4% hw errors

Project status:
First batch of boards is in production so I can react immediately when the chips are shipped.
Website - almost ready to launch.

I wonder why the error rate of klondikes still are higher than burnin's design.

[BkkCoins]

I wonder why the error rate of klondikes still are higher than burnin's design.

I've commented on this several times. It's the noise. My first prototype board didn't have the ferrite beads for both USB and PLL (clock) power supply. I scoped some serious noise on the GND plane when connected to my notebook, and when connected to the RasPi the noise was gone, and the error rates dropped to <0.7%, even at 330 MHz. How far the clock can be pushed when the new boards are made is still to be seen, but the error rates were the result of noise injected from the USB connection.

[SebastianJu]

I wonder why the error rate of klondikes still are higher than burnin's design.

I've commented on this several times. It's the noise. My first prototype board didn't have the ferrite beads for both USB and PLL (clock) power supply. I scoped some serious noise on the GND plane when connected to my notebook, and when connected to the RasPi the noise was gone, and the error rates dropped to <0.7%, even at 330 MHz. How far the clock can be pushed when the new boards are made is still to be seen, but the error rates were the result of noise injected from the USB connection.

Ok... then we'll see the results when the noise is suppressed.

[Bicknellski]

Chillax everyone...

Let us see what next week brings with the new revised board.

Wow... seems like some people want to play an us vs them here. What we can learn from both development teams is that the chips definitely have the legs to hit 400 and that is good but nothing really new given what the OC is with standard Avalons. Time to just sit back wait for the revision and see what is possible. Given the limits on a USB hub port those working towards K1s might want to find something that goes to 11 a la spinal tap! I'd prefer building K1s with the revisions you suggested if possible BKK.

[jdape]

I've designed some Klondike 16 computer case/motherboard adapters.  The designs are open source.  Use these adapters to mount Klondike 16's into any of: 3.5" Drive Bay, 5.25" Drive Bay, PCI Express slot, or PCI slot.

Thread here:
https://bitcointalk.org/index.php?topic=246638

For sale here:
https://cp.virtualvm.com/cart.php

I would not cross post, except that the designs are free and I'm doing this as a service to the community -- not for profit.

Personally, I'm using these adapters to install a dozen K16's into a $90 4U case from newegg.  Details and pics in the Adapters thread linked above.

[Bicknellski]

Nice Jdape. LOL love the glow in the dark one!

[jdape]

Nice Jdape. LOL love the glow in the dark one!

Thanks! That one belongs to massnerder, who placed the first order yesterday

[SebastianJu]

@Bicknellski... if you meant me... my interest lies in chosing the best product because i have chips i want to be built into a miner. And im not sure yet if i should chose burnin or marto74's(Klondike) offer. Thats why i want to know the performance of both designs. I see that Klondikes has to fight the noise first to get real results. Im not sure yet but i think overclocking seems to be a bit easier with burnin's design. No soldering for overclocking or overvolting. But im not sure yet, because my question wasnt answered yet.
On the other hand i think the development in another design group is interesting to another designer too. So he can compare and maybe see problems.
Nothing bad intended...

[bigbeninlondon]

@Bicknellski... if you meant me... my interest lies in chosing the best product because i have chips i want to be built into a miner. And im not sure yet if i should chose burnin or marto74's(Klondike) offer. Thats why i want to know the performance of both designs. I see that Klondikes has to fight the noise first to get real results. Im not sure yet but i think overclocking seems to be a bit easier with burnin's design. No soldering for overclocking or overvolting. But im not sure yet, because my question wasnt answered yet.
On the other hand i think the development in another design group is interesting to another designer too. So he can compare and maybe see problems.
Nothing bad intended...

The one thing I don't like about the bitburner is that if one chip goes, the entire bank of 10 stops hashing.  That's a killer right there.  With BKKCoins, dead chips don't always mean the rest stop hashing.

[SebastianJu]

@Bicknellski... if you meant me... my interest lies in chosing the best product because i have chips i want to be built into a miner. And im not sure yet if i should chose burnin or marto74's(Klondike) offer. Thats why i want to know the performance of both designs. I see that Klondikes has to fight the noise first to get real results. Im not sure yet but i think overclocking seems to be a bit easier with burnin's design. No soldering for overclocking or overvolting. But im not sure yet, because my question wasnt answered yet.
On the other hand i think the development in another design group is interesting to another designer too. So he can compare and maybe see problems.
Nothing bad intended...

The one thing I don't like about the bitburner is that if one chip goes, the entire bank of 10 stops hashing.  That's a killer right there.  With BKKCoins, dead chips don't always mean the rest stop hashing.

Thats a point. Ill ask burnin if he could change this or not. On the other hand... are there reports of breaking avalon asics anywhere at all?

[trigeek]

Thats a point. Ill ask burnin if he could change this or not. On the other hand... are there reports of breaking avalon asics anywhere at all?

Yes.  I can't find any exact threads, but I have read reports of chips not performing correctly in avalon-built units.  Of course it depends on what the actual failure mode is, but I wouldn't count it out.

[turtle83]

Thats a point. Ill ask burnin if he could change this or not. On the other hand... are there reports of breaking avalon asics anywhere at all?

Yes.  I can't find any exact threads, but I have read reports of chips not performing correctly in avalon-built units.  Of course it depends on what the actual failure mode is, but I wouldn't count it out.

paraphrasing someone : Everything that can break will eventually break...

Possibility of chip breakage should be low, but in the event that it happens, you wouldnt want the board to become totally useless, waiting for repairs...

[SebastianJu]

Thats a point. Ill ask burnin if he could change this or not. On the other hand... are there reports of breaking avalon asics anywhere at all?

Yes.  I can't find any exact threads, but I have read reports of chips not performing correctly in avalon-built units.  Of course it depends on what the actual failure mode is, but I wouldn't count it out.

Did the chips break while runtime or where they broken from the start?

[Jumpsu]

I have pcb ready to flash, how i flash pic with pickit3?
what files are needed?Order?

Can someone make step-by-step guide?

[nekonos]

Hi BKKCoins

I'm uploading the gerbers for make 4 prototypes.

Gives some issues with anular rings and some drill holes.
16 drill are not aligned correctly are the center via hole for every asic chip.





I use the k16 qfn version. I analyzed this in eurocircuits.com price calculator and uploading all gerbers.

Maybe its a problem from manufacturer importing the gerber files...

Whats the class of PCB  6-7?

Many thanks

Regards

[Taugeran]

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

I've been running steady for most of today with very low errors at 330 MHz. No noticeable heat issues or any anything else to report. Chips seems to be running 50-54C. Heat sink is only slightly warm. Fan is 80% but inaudible.

If that works out as typical on a full K16 that would be 5280 MH/z.

crank it up a little more and you be kicking  BFL's as*, i mean toosh