Bitcoin Forum

A little over a week ago ngzhang anounced his proposed production of the 'Avalon' ASIC with >= 60Ghs for about $1299 (about double the hashrate of BFL's Single 'SC', and Cablepair's bASIC)

Since then:
BFL have announced their Single 'SC' will now hash at 60Ghs (from 40Ghs) @ $1299
Cablepair has announced their bASIC will now hash at 54Ghs (from 27Ghs) @ $1069 (now selling 27Ghs version @ $599)

Were both BFL and Cablepair originally pricing their units at more than 100%+ markup of the cost of production and development?
Because as i see it they cannot afford to produce these now and make any profit.

Also how have they done it?
1) Are they just doubling the number of boards in each unit? (Only viable option if they have already started PCB production)
Or
2) Have they doubled the number of chips on each board? (Only viable if they haven't started production yet)

In either case i can not see any of these ever hitting the market before Christmas. My money is going firmly back in my wallet until some real devices hit the real world, because all i see at the moment is BS.

Welcome to bitcointalk!  Do you mind posting some links to all of these statements?  I haven't been able to keep up with all of the news lately.

Keep in mind all of the specs that were released were very preliminary.  However, I do hope you keep your money in your wallet b/c that will increase my profits over the next few months.

Adding chips would not increase the cost of each unit all that much, so that's an easy way to stay competitive.  However, I don't think anyone has released how they will accomplish this, b/c they haven't really released how they would accomplish the previous benchmarks either.

Getting excited for ASIC!

EDIT: https://forums.butterflylabs.com/showthread.php/16-Announcement-BFL-ASIC-Release-specifications

Not quite double, closer to 50% I think.

o.0

Strange... BF's mini rig SC is also up 50% 1000 -> 1500GH/s

http://www.butterflylabs.com/products/

They need more money... Something is not going to plan and more funds are required.

You can check their relevant posts here in Custom Hardware, and their respective websites.

Sure you can add more chips to the board, but that requires a complete redesign of the PCB. So no one is even at that stage of production yet (or they've lost a ton of money in R&D).

Granted their specs were 'preliminary', but that was months ago.
I can't see how any production can possibly have happend in the past months, therefore it seems unlikely any of them will hit their proposed October/December delivery dates if they are still 'Talking' and changing proposed specs.

And now we bare witness to the GH/s war.  :D

At this point, I don't care who says or promises what anymore.  As the OP puts it, I'm not giving anyone any of my money until the product exist in the real world.  Even if I will be paying the full price, I'm alright with that.

I think they are over blowing the spec's, to make a asic just takes money. To have someone make a PCB board to hold a FPGA is only 100 dollars. Send the fpga configuration to a ASIC maker and they will produce your chips. BFL was using a EP3SL150 which has 142,000 LE's you find a chip with more LE's you got a faster machine A AMD Radeon HD 6990 3072 ALUs just to put that in perspective. Now a EP4SE820 has about 813,000 LE's which is WAY WAY faster then the EP3sl150. After they have it configured to they send it to be made into Asics here is a preview of Altera's hard copy of the IV, Density 2.8M to 15M usable ASIC gates (not including transceiver hard IP, I/Os, PLLs, and built-in test logic)  15M imagine how fast a of a miner that would be lol. Now these chips are crazy expenisve, the IV sales for 11k through altera I Have not got any qoutes on this chip from china yet. The ep3sl150 is about 150 dollars a piece if you order 25+ from china.

Competition is the key and making a successful product for profit is the motivation.

Have to keep you clients happy. Common  business practice for successful product's.

Competition is great, i just have an issue with specs changing so near to their supposed delivery dates. Which makes me question how far along with development they really are.
When you've gone so far down a development path it becomes expensive to start changing things at such a late stage. Profits dwindle and companies go bust.

All we've seen so far from anyone is mockups, which makes me think that production and therefore delivery is still a very long way off.

With respect to these changes BFL may have been very conservative with their initial offering.

Mini SC units may have extra slots for more modules to buffer small issues. However only time will tell.

If you have decided on a product then its best to trust that you made a good call.  Hindsight is 20/20!

In regards to the bASIC the plan was to have 2 PCB boards 27gH/s and 54gH/s for $1k/$2k respectively. However he found that price wise he could offer the $2k board at $1k after getting further along in the development (though I am sure this would not have happened if it were not for the competition that avalon brought).

Who are you again?

I'll tell you what sets off my alarm bells...multiple brand new posters bashing ASICs mining equipment companies.

Capitalism FTW :)

Someone with some common sense and a fresh perspective.

Asking said ASIC mining companies to explain their sudden shift in offerings is not bashing. It should be straight forward for them to answer.

It'll be interesting to see if they shift again should someone come out offereing a 100Ghs @ $1000 solution.

the thing about asking questions is that the answers will lead to more questions until one of two conditions are met.

1. are you satisfied with the answers.
2. are you satisfied with the working result as evidence.

Not too long ago, all of the ASICs for preorder were ~30GHs/1kUSD, give or take 10%.

Avalon preorders come along, and ups that to ~46GHs/1kUSD. Everyone wants some! After the preorders, the Avalon's price would go up to $1999, which is ~30GHs/1kUSD, right in line with everyone else at the time.

Tom's bASIC sales weren't exactly extraordinary, so he had to do something to increase sales. Adding a 2nd chip to double the speed at the same price makes it much more attractive! Note that he didn't push his hardware beyond advertise specs, he just doubled the number of chips. Also, this put the bASIC ahead of the Avalon, at ~49GHs/1kUSD.

Meanwhile, BFL is still ~30GHs/1kUSD, which ain't lookin too good. However, their numbers were always conservative because they didn't want to do what they did last time, and advertise unobtainable specs. They push their hardware closer to their maximum specs, and now BFL's Single and MR are between 46-50GHs/1kUSD. Bam, competitive again.

I don't see any BS in any of this.

How did they obtain this increase?

If it's through using conservative numbers and pushing the hardware in testing with stable results ... OK.
If they had to go back to the drawing board means that there will be delays in production and delivery dates.  ???

If anyone was as far along with their design as they are saying, they wouldn't be able to change it.

Exactly, they would most likely have to start from scratch and redo the development of the chip.  Such an increase is extremely fishy since they advertise that they will be able to ship next month.

Watch as the largest scam in bitcoin 2012 unfolds.

All ASICs are using multichip architecture in their PCB so you would need to redo the PCB layout, but the actual chip architecture/design.

Yet their prices are the same..

A multichip architecture may be true, but the entire board would have to be revised. You can't just slap another chip on and have everything else communicate properly. Not to mention Power Distribution. I hypothesize that the numbers originally stated were low estimates, which falls in line with something that Sonny said in an interview once (IIRC). The chip clocks were likely lower at first, then raised to a new level. Likely, the clocks would be lower as a marketing ploy. If they can release another product with the same chip 3 months later at 2x the efficiency/hashrate why not?

Their prices are the same for the above reason.

The possibilities are endless, but in the end only one path, one result.

Didn't BFL say that they'd already planned to increase hash power in the future via a firmware update?  That would imply that the units were built to handle the higher specs from the outset and they were keeping that increased power in reserve in case a competitor emerged with better specs/lower price.  In fact, had no such competitor emerged, they could have not given the firmware update to the initial purchasers and simply released a "new" rig with the higher specs for no additional cost to them when sales started slowing down.

This. I've heard similar team from the DB Reclaimer team.

Firmware for a ~50% increase in speed... At first BFL said heat will not be an issue. I wonder if this increase in speed will change that.

If you read this announcement here you will understand
https://bitcointalk.org/index.php?topic=79637.msg1229735#msg1229735

Also I can see why your logic would lead you to believe this but it's not your fault - you just dont know some specific things about this industry and this process

1) All the costs for making an asic chip are in research and development - once you have that part of the process completed - making the chips themselves costs a few dollars each.
2) PCB Design / Fabrication is not any kind of complicated or time consuming process and it's no big deal to change things on the design.

I can get an original PCB designed and fabricated in very little time, especially if you want to spend a little extra money.

The last PCB batch I had fabricated for the ModMiner Quad, I had 200 Backplane PCB boards made, and 800 Spartan-6 PCB plugin cards made
the entire process took about 12 days and that includes shipping from China to NY

I am not sure about BFL's speed increase, they would have to answer that - its possible they are just clocking their chips higher, they created a better firmware - or doing what I did (changing the amount of chips per board) or something else all together

but to answer for the bASIC

I had two board designs made - one with twice the amount of chips as the other, originally I was pricing the 27Gh/s board at 1069.99

but, After looking at the financials and talking it over with my engineers I realized it made much more sense just to price the "double  chips" board at what I had originally been selling them at because It would not be right to do to my current customers (to release a faster unit after they had already bought one) instead they got a free upgrade from 27gh/s to 54gh/s and we decided to lower the price of the 27Gh/s board.

hope this clears up any confusion you might have.

thanks! :)
Tom
BTCFPGA.com

It seems that some people don't understand the development cycle at all.

The initial design of the hardware (chip or board) takes time and wages. this is a relatively small investment. Once this is done you can start development of the firmware.

The largest costs for an ASIC run isn't the run itself - it's getting the manufacturing in setup to make it. That's a huge front-loaded cost... followed by a relatively tiny price per unit.

That's just the chip. The board production is much much much cheaper. It does have a front loaded cost, but when you're talking about 10s of millions it's trivial.

Now if you're designing a such a product - it makes sense to make it modular in the sense that some small number of chips goes on each board and that board is designed to be linked together (either physically or by production of larger boards with more modules on them).

The reason we've got 27/54 as a number is because he's done exactly that, added a unit in the same enclosure or made a double board. I assume he's got harder numbers than BFL had because of the existing sha3 bASIC products that exist...

BFL on the other hand, going completely custom had very little hard numbers to base estimates on --- and given what happened with their fpga initial releases (instability, design/version changes, heat issues, etc). I'd expect them to take a best guess based on workbench schematic testing (or whatever) , and then lowball it... by a lot. If for no other reason then to prevent a customer service nightmare if they had to ship a product at 50% of the estimated spec. It also has the advantage of leaving them 'firmware' updates that could keep their prices up for a longer period of time as competition develops.

The fact remains, there are still ways to refine the machines themselves once they're released - if they're well designed. for example of any of them have the top of the chips as the highest points on the pcb then some other company (or the makers themselves) could design a custom enclosure/heatpipe/cooling system that could provide for better hash rates. Just like some people are replacing BFL's stock heatsink and fan with a 120 mm fan and massive piped CPU block (or water cooling).

But I imagine effeciency of the hardware hasn't been a primary concern for anyone at this point - this is a race to get a product out first, at a price people will pay.

~~~

That being said, I would expect faster firmware / enclosure refinements / changes in board design to be what happens after they've released and get initial orders shipped.

And of course, there's always potential competition from some as of yet unknown entity. What if AMD jumped into the market --- how would existing companies compete with deep pockets + production side being owned in house + 20 years GPU production experience + much more compact chips?

Only time will tell.

It's possible you haven't seen anything yet. As R&D costs are covered, they could be much cheaper. I can imagine one day maybe 70% lower cost than you see for asic now. But, hopefully, the overall competition does not cause this to happen too fast. Otherhwise, all the investments made will equate to a lot of money invested for soon to be low cost products.

Anyway, over time we should see even better deals. Competition starts dropping prices, perhaps, one dropping first, and then the other matching the offer later. And so on, until they have an unwritten (hopefully) line in the sand that they both feel is low enough (and each hopes the other will not attempt to cross). Then perhaps the best way to proceed is to make a better product people will want to buy. Yet, this makes a hazardous  assumption that they have similar or proportionate up front (development) costs.

Who knows how it will all unfold though. :P

||bit