Open Source Avalon Gen2 55nm Board

[klondike_bar]

What's the part number of that .9 volt power supply?

PTH12040W => http://www.ti.com/product/pth12040w

If anyone is interested I have some of thus I can sell you...

for only 10 chips, a smaller 30A component would likely suffice, such as the one bitfury hardware uses:  Texas Instruments TPS53355 http://www.ti.com/product/tps53355
it serves a similar (or identical?) purpose, with a much smaller footprint and is almost certainly cheaper to source and install.

[Lucko]

Yes but this is 10 chips because he has 10 chips... And 24W power usage. So yes only bit more then 50% at the moment but there is also efficiently that you need to think about. And overheating. I have a Chili board. And the biggest problem is power supply. It is on the limit and it needs additional cooling or the board slows down by 20%... So because you have save some $ on parts you need now additional cooling(fan and heatsinks) on power supply and that cost more then stronger power supply and uses more power since you need a fan and it runs in less effective part of operational range... Not to talk about lost hashes... So some parts are better if they are stronger...

[ngzhang]

In our under development design, I use a 45A rated, 55A stable, 60A peak DC-DC circuit for 14 chips. I do think this is a good start.

Basic configuration is 7 chips pre bank, 2 banks share one DC-DC. Will open-source when the test is over.

By different operation voltage and clock speed, the chips power consumption is vary greatly. for example, 1GH@0.8V only consume <1.5W, about 1.9Amps. 1.5GH@1V is about 3.75W, about 3.75Amps. There is no doubt this chip can over clock to 2GH (this is limited by PLL at 25MHz XCLK input), but the voltage must increase to maybe 1.1V (NOT tested),  the overall power consumption may increase to 6W and 5.5 Amps.

So, select which DC-DC module is depend on which type of product you want. more power efficient or push the chips to its limit?

By personal views, A3255 chip is cheap, if over clock too much will go with a huge increase of peripheral (DCDC, heat dissipation, etc)cost, then it's unworthy. just add a few chips and run under lower voltage. In other words, this is an art of balancing.

In addition, use Ti power module is a good solution for fast shipping, but they are really costy. I will release a opensourced 2 phase 55A DC-DC design for reference next month(sorry, it's a bit slow but high-current DC-DC circuit is not easy to design and debug), it' maybe only cost 5$.

[HorseRider]

By different operation voltage and clock speed, the chips power consumption is vary greatly. for example, 1GH@0.8V only consume <1.5W, about 1.9Amps. 1.5GH@1V is about 3.75W, about 3.75Amps.

How crappy it is! Ones who dont keep promise usually are stupid. When stupid asses design a chip, they design a chip only worth a shit. 2.5J/GH,   . It is a joke, ng. You have wasted the fabrication capacity of TSMC. If I were the designer, I would have killed myself as I could not bear the shame of this chip.

Everyone who has been burnt by Avalon, please spend 20 seconds of your time and give this stupid and evil ngzhang a negative feedback:

https://bitcointalk.org/index.php?action=trust;u=38132

[aneutronic]

Wherever BkkCoins is now, i wish him all the best, and may he come back soon.

Agreed.

-------

Not sure how much support there is for chips from Avalon given what they have done to everyone. Plus the fact the chips "underperform" and are not as "cost effective" as Bitfury chips which are available right now.

Personally I'd really want to support efforts like this if there are future plans to utilize other chips other than Avalon.

+1

Nice design but pick another chip, why would you want to support scammers with an inefficient chip.

avalon screwed enough people here already,   

please don't help them screw more.

[Wesly]

Thank you for all encouragement and inspiration above, there is no doubt Avalon project will not grow up without your criticisms.

Also, we will not waste a cent on losers.

Decentration is one for the most important spirit of Bitcoin, so we still stick to chip selling, rather than maching making or self-mining, this is our belief but needs the joint efforts of the whole community.

Are you guys seriously working with Avalon who openly called their previously customers losers?

"criticisms" = "encouragement and inspiration"?  What are you?  LeBron James?  You guys are so out of touch with the Bitcoin community it is not even funny.  What's with all these double/passive-aggressive talk?  How can you expect to regain people's trust when you called your previous paid customers that you screwed over losers?  If nothing else, you guys are the one who is responsible for creating these money/bitcoin 'losers'.

You must be hurting for business to have to come crawling back here asking for more money?  And auctioning your chips off with ZERO bid.  Because of your major screw-ups you are now forced to prepaid for overprice, under-performing silicons that you can't sell and trying to keep alive developing the next gen before you run out of money.

Don't give us the BS about "decentralization" (can't even spell it right, let alone know the meaning) and "spirit of Bitcoin", and therefore you are not self-mining.  The reason you are not self-mining is because it is unprofitable by the time you can mass-producing your miner with your inferior chips for a reasonable cost, the constant 40%+ difficulty (soon to be 80% as KnC release their B2 and again in December when HF release B1/B2) increase every 10 days is killing any chance for ROI even with free electricity and you know it...   Besides, even when time is good, it is hard to scale as can attested by ASICMiner.  They learnt real quick it is much more profitable and less headache to pass on the hot potatoes (miners) to the next suckers/'losers'.  If a large and resourceful company with all these profits from previous batches can't built miner (at cost without any markup) profitable, how can you expect an individual to pay for miner with huge markups from you, board builders, assembler using your underperforming chips?  If you don't know that then you are an even bigger fool than you are or think we all are.  But I bet you are smart enough to know that don't you?

[Bogart]

Nice design but pick another chip, why would you want to support scammers with an inefficient chip.

avalon screwed enough people here already,    

please don't help them screw more.

I have to agree.  Avalon/BitSyncom deserves zero support from the community they have screwed over repeatedly, all while continuing to show no remorse.  They even call us losers.

I personally will not buy such a board because of this, no matter how attractive the pricing.

[jtothevds]

design boards that fit in standard avalon casing! so we can scrap the old cards and get 4x80x1.5GH per unit.

This probably isn't too much of a challenge if you lost the TI part. I have converted the Eagle files to Altium, they need some cleanup but would happily commit them into the repository if you like...

[form]

Some progress to report...
At 1.03 V my chips seem to run pretty stable around 14 GH/s.
I can't do power-measurements right now, because the voltage-drop on my current-meter is so big that i cannot calculate the power without a big error. (Can't measure actual voltage at the same time now, maybe on monday.)

Firmware and cgminer-driver seems to run pretty good at the moment, besides some Idle-Warnings from time to time and a decent HW-error rate of ~1.7 %.

http://my.root4u.de/avalon2-secondhash.png

I have converted the Eagle files to Altium, they need some cleanup but would happily commit them into the repository if you like...

Very good. If you tell me how i can gain you access to the github repo, i would be very interested.
But be aware that this is only a test-board. If i make a final one (if here is some interest), a few things will change. For example the number of chips, dimensions, or the placement of the decoupling caps and things around the chips - It would be better if the components are a bit lower then the chips or are not in the way for a topside-heatsink. I wasnt aware that the chips need more cooling from the top side as from the bottom side. At the moment i have some copper distance pieces and little chipset-heatsinks on the chips, but that just doesnt scale.


https://raw.github.com/formtapez/avalon/master/Photos/CIMG4612.JPG
https://raw.github.com/formtapez/avalon/master/Photos/CIMG4613.JPG

design boards that fit in standard avalon casing! so we can scrap the old cards and get 4x80x1.5GH per unit.

That sounds interesting, but unfortunately i don't own such a avalon case. It would be helpful when someone who has one, tells me the exactly dimentions of the board, heatink positions, etc. and can provide some detailed photos.

[klondike_bar]

Some progress to report...
At 1.03 V my chips seem to run pretty stable around 14 GH/s.
I can't do power-measurements right now, because the voltage-drop on my current-meter is so big that i cannot calculate the power without a big error. (Can't measure actual voltage at the same time now, maybe on monday.)

Firmware and cgminer-driver seems to run pretty good at the moment, besides some Idle-Warnings from time to time and a decent HW-error rate of ~1.7 %.

I have converted the Eagle files to Altium, they need some cleanup but would happily commit them into the repository if you like...

Very good. If you tell me how i can gain you access to the github repo, i would be very interested.
But be aware that this is only a test-board. If i make a final one (if here is some interest), a few things will change. For example the number of chips, dimensions, or the placement of the decoupling caps and things around the chips - It would be better if the components are a bit lower then the chips or are not in the way for a topside-heatsink. I wasnt aware that the chips need more cooling from the top side as from the bottom side. At the moment i have some copper distance pieces and little chipset-heatsinks on the chips, but that just doesnt scale.

design boards that fit in standard avalon casing! so we can scrap the old cards and get 4x80x1.5GH per unit.

That sounds interesting, but unfortunately i don't own such a avalon case. It would be helpful when someone who has one, tells me the exactly dimentions of the board, heatink positions, etc. and can provide some detailed photos.

looks good - whats the blue component, a wifi antenna? It may be better to scrap that component to save a few dollars on parts/assembly since most users prefer to run an ethernet line (my 2c - either way i really like the board and applaud its working!!

[Lucko]

That is for regulating voltage. Trimpot...

[klondike_bar]

That is for regulating voltage. Trimpot...

whoops, now that i look i actually see the markings on it - makes sense! At first glance i thought it was a little antennae connector.

[driksson]

design boards that fit in standard avalon casing! so we can scrap the old cards and get 4x80x1.5GH per unit.

That sounds interesting, but unfortunately i don't own such a avalon case. It would be helpful when someone who has one, tells me the exactly dimentions of the board, heatink positions, etc. and can provide some detailed photos.

It should be readily available for avalon gen 1. there all designs are public in their github.
https://github.com/BitSyncom/avalon-ref

I believe its the controller that avalon uses , since the avalon 1-2-3 units have up to 320 chips have to be quite advanced?
Im not sure, this is things i have heard.

[DPoS]

Avalon are a bunch of worthless apes!   

[Spotswood]

Please enlarge the board in order to make room for mounting holes in all four corners (like the Klondike board).  That way the boards could be mounted properly in an open-air case:

[allinvain]

Open air cases are nice, but personally I'd prefer if these modules were a drop-in replacement for existing Avalon modules. The Avalon cases are a pretty solid design (with the exception of the top lid screws) so IMHO it would be a shame if us existing Avalon owners could no longer use them.

Ngzhang, are you guys going to produce your own modules, or are you going to abandon the backwards compatibility idea?

[QualitySeeds]

Thanks for your opensource project :-)

And do u guys still work on this project ?

[Lucko]

If there is some interest for sure... We do looking into way to put this in Avalon casing...

[QualitySeeds]

If there is some interest for sure... We do looking into way to put this in Avalon casing...

I'am and there should be more...;-) there are only a few opensource topic's and surely people would like to know more about that....I'm getting into the whole pcb design and frankly i love it there is sow much potential

A friend have's allot of equipment i can use to make a board and test it and if i all get it prob buy that for my own really
just getting started sow i do not know if i can help you with anything....

What kind of software where u using making the pcb design ? a'm using Kicad

( can order 80 or more A3256-Q48 ) hope its legit do

[ngzhang]

Ngzhang, are you guys going to produce your own modules, or are you going to abandon the backwards compatibility idea?

We will release a few (very few) Avalon Gen2 Mini prototype on our shop in 1-2 days, this prototype will NOT mass production.

http://avalon-asics.com/shop/

(not put up yet)

It's basically based on our open-sourced design, 96 chips running at 1V/1.5GH, 144G overall speed. most of the parts are the same with Gen1 mini.

I think this is a start, it's easy to build upgrade modules for Gen1 machines.

But we will NOT produce compatible modules our self in large number, a lot of people can do it faster I think.