Pages:
Author

Topic: community asic project - page 2. (Read 3538 times)

donator
Activity: 290
Merit: 250
April 14, 2013, 11:51:52 PM
#17
I'm not clear on where the difficulty lies: is it the chip design? The cost of getting the chips created? The board layout?
I'm honestly surprised that ASICS aren't more common.

btw...Virtex7 anyone?

hero member
Activity: 756
Merit: 500
April 13, 2013, 11:29:28 AM
#16
I am also interested in contributing to this.

I think we can get a much smaller quantity of asic chips by using what is essentially a cybershuttle program:

http://www.tsmc.com/english/dedicatedFoundry/services/cyberShuttle.htm

Taiwan semiconductor offers this, and the original 4000 chip run for avalon run 1 was done using this service as far as I can tell.  

That should bring the cost down to $20,000-$40,000 if the FPGA conversion can done on the cybershuttle service.  Thoughts?

I'll have the final cost numbers on doing it through this program Tuesday/Wednesday, there are several solutions that offer a lower NRE like this, I'm compiling the data from them all will, different solutions will produce different results (hash speed/heat/power consumption) and different cost

hero member
Activity: 756
Merit: 500
April 13, 2013, 11:27:19 AM
#15
As a guy who works on ASICs/FPGAs for a living, I wish you luck. It sounds to me like you're being overly optimistic on the amount of time and money required to produce a working chip.

Yes, SHA256 is easy. What isn't easy is all the cell placement, wire routing/sizing, silicon doping parameters, etc etc which all feed in to your final achievable performance.


Not overly optimistic - for a high speed asic, it is more costly and more timely - for an "avalon" clone - it can be done quicker and cheaper using a different approach - but would an avalon clone be worth the effort compared to an avalon killer?

Since you work on asic's, you know what it takes to go from 60Ghash/s to 600Ghash/s and what the technical changes requirements are.
hero member
Activity: 756
Merit: 500
April 13, 2013, 11:24:59 AM
#14
To get the performance hash rate I mentioned, it would be a custom asic - depending on the batch size the cost could be from $1500 to $4500 for a 600Ghash/s to 1.2THash/s - yes its a multi-million dollar project - so it wouldn't be taking the bAsic approach after all - think about it - ASIC's round 2, it would blow the doors anything in the market place today

I've run the numbers and while we could get a "avalon" like unit out the door in 3 to 6 months, why not invest in something that will blow it away but at double the time to market....


sr. member
Activity: 401
Merit: 250
April 13, 2013, 11:23:18 AM
#13
As a guy who works on ASICs/FPGAs for a living, I wish you luck. It sounds to me like you're being overly optimistic on the amount of time and money required to produce a working chip.

Yes, SHA256 is easy. What isn't easy is all the cell placement, wire routing/sizing, silicon doping parameters, etc etc which all feed in to your final achievable performance.
sr. member
Activity: 392
Merit: 250
http://casinobitco.in/ A+ customer support
April 13, 2013, 02:56:15 AM
#12
ill be the cheerleader

seriously this is great guys
member
Activity: 87
Merit: 12
April 13, 2013, 01:10:19 AM
#11
I am also interested in contributing to this.

I think we can get a much smaller quantity of asic chips by using what is essentially a cybershuttle program:

http://www.tsmc.com/english/dedicatedFoundry/services/cyberShuttle.htm

Taiwan semiconductor offers this, and the original 4000 chip run for avalon run 1 was done using this service as far as I can tell.  

That should bring the cost down to $20,000-$40,000 if the FPGA conversion can done on the cybershuttle service.  Thoughts?
full member
Activity: 163
Merit: 100
April 13, 2013, 12:31:06 AM
#10
Does sound awesome. The problem is getting enough backing on kickstarter and enough community trust in order to get funds together to make it work. The bulk of those that would be interested would be the people already fairly well in-the-know about the current ASIC situation.

With the trials and tribulations (and downright misleading information) that has happened over the last 9 months (wow, has it been that long?) there is going to be even MORE skepticism the second time around.

I think bASIC was wanting a similar idea to have a lot of things 'taken care of' by other companies but it apparently became a nightmare and Tom imploded under it all.

Link the kickstarter project and let people make their choices. It may be that the most funding will come after BFL ships when there are more ASIC owners looking for something else to 'invest' in Smiley
hero member
Activity: 798
Merit: 1000
www.DonateMedia.org
April 13, 2013, 12:18:38 AM
#9
ASIC costs literally millions of dollars to develop, I wish you the best of luck though you will have a pretty long road to get there.
full member
Activity: 158
Merit: 100
April 12, 2013, 09:08:59 PM
#8
Yes, but at what price point?

While doing your design work, think about a scrypt-based miner too (yes, I know you need a bit of fast memory).
hero member
Activity: 756
Merit: 500
April 12, 2013, 01:49:46 PM
#7
I have reached a key point in this potential project - it is definitely possible that this can be done and be brought to market within 6-9months (or less) - and it could be 10x faster than avalons current line up

however I would need some additional people with some specific skillsets
 

this will be on kickstarter - need to decide best business plan as well, either hosted, cloud, or just selling miners....


I'm looking at units at 600Ghash/s to 800Ghash/s EACH - little longer to develop and a little more costly but the wait would be worth it -
hero member
Activity: 756
Merit: 500
April 11, 2013, 02:08:30 PM
#6

 This would be a great idea if it was on Kickstarter.


doing that now



also just some info - the goal is not to do a full custom asic, but rather leverage an existing core/technology or SoC and put in the development effort to adapt it for mining

member
Activity: 78
Merit: 10
April 11, 2013, 02:05:12 PM
#5

 This would be a great idea if it was on Kickstarter.
hero member
Activity: 756
Merit: 500
April 11, 2013, 01:56:13 PM
#4
now before you start saying an asic is too hard, google is your friend, 99% of the work has been done and the startup costs are not out of control

What do you estimate those costs to be? Have any sort of preliminary plan on how you would go about this? Manufacturing? Chip design? Business plan?

"If it were easy, everyone would be doing it!" -Un Known

For the costs...MOSIS starts at $50k, goes up from there, I have spent a lot of time sorting through all the requirements and existing designs and limitations and also what is required to make a FASTER unit than avalon that scales


hero member
Activity: 756
Merit: 500
April 11, 2013, 01:54:52 PM
#3
now before you start saying an asic is too hard, google is your friend, 99% of the work has been done and the startup costs are not out of control

What do you estimate those costs to be? Have any sort of preliminary plan on how you would go about this? Manufacturing? Chip design? Business plan?

"If it were easy, everyone would be doing it!" -Un Known


Yes I do have a plan, but I also am looking to see about a community driven effort, at a bare minimum, I will put together a plan for a prototype that is close to the avalon in terms of speed/hash

a lot of places do fpga to asic, so I will be leveraging the work that has already been done, and there are several High speed asic designs available, but they need to be changed to work as a miner (double sha-256)

I will be contacting UMC & TSMC, identifing the missing key points needed.  Locating resources to fill in the gaps, and in tandem start looking at the financials... avalon's "Batch" processing actually is very cost-effect - MOSIS offers low qty chips, so we could get a design done and put down into a chip for testing pretty quick
member
Activity: 112
Merit: 10
April 11, 2013, 01:47:10 PM
#2
now before you start saying an asic is too hard, google is your friend, 99% of the work has been done and the startup costs are not out of control

What do you estimate those costs to be? Have any sort of preliminary plan on how you would go about this? Manufacturing? Chip design? Business plan?

"If it were easy, everyone would be doing it!" -Un Known
hero member
Activity: 756
Merit: 500
April 11, 2013, 01:34:49 PM
#1

okay, before everybody yells SCAM, I'm just throwing down an idea.. open-source community asic miner - now I bought pre-orders from bfl & and bAsic, and not avalon (agh), but avalon isn't even shipping... alot of miners just lost a lot of money


now before you start saying an asic is too hard, google is your friend, 99% of the work has been done and the startup costs are not out of control


please read http://cryptography.gmu.edu/athena/NIST/2010_11/ASIC_benchmarking_Patrick_Schaumont.pdf -  some pricing data (yes outs of date)




there are also numerous sha-256 cores available, are we going to let avalon continue the monopoly?
Pages:
Jump to: