LargeCoin Pricing Announced; Taking Pre-Orders

[casascius]

I am surprised that as a legitimate offering, this would even be offered for sale.

If I were in a position where I thought I'd be able to have a head start on manufacturing ASICs that blew all other miners out of the water, I'd be highly motivated to just keep it all to myself as I pretty much dominated mining.  I'd score 7200 BTC a day (with it halving soon, of course).

Or perhaps I could have my cake and eat it too.  I could sell rigs for a healthy $30k, knowing that I alone had the upper hand in deciding how productive those rigs were going to be.  If for every 10 rigs I made I ran nine and sold one, I'd effectively guarantee that all my customers were competing for 720 (360) BTC and nothing more.  The customers I sold it to would have 10x the expectations - in the end, 9x of their expected yield I'd be keeping to myself with my supply of "unsold" rigs that I was running for, um, "extensive burn-in testing" before being sold.  (What, you think I'd let them sit idle on a shelf in a warehouse?)

If this product results in centralization of mining, I bet there will be cries for Bitcoin to switch to a CPU-based algorithm!

What about the already existing network of approximately 10 Thash/s? And if they only have 25 units and they meet their expectations, they have 500 Ghash/s. Certainly a nice farm, but nothing that will "dominate mining". Just because they are more efficient doesn't automatically render existing miners unprofitable, especially considering miners that have already passed the break even point of their investment.

I completely disregarded the 25 as I don't see it as any meaningful limit, and my argument would apply to anyone doing ASIC mining, not just these guys in particular.  I assume that the only reason that they'd do 25 is for prototyping to keep the risk manageable in case there were any bugs to be worked out, and if everything works as planned, nothing's stopping them from making however many they want.

[notme]

I am surprised that as a legitimate offering, this would even be offered for sale.

If I were in a position where I thought I'd be able to have a head start on manufacturing ASICs that blew all other miners out of the water, I'd be highly motivated to just keep it all to myself as I pretty much dominated mining.  I'd score 7200 BTC a day (with it halving soon, of course).

Or perhaps I could have my cake and eat it too.  I could sell rigs for a healthy $30k, knowing that I alone had the upper hand in deciding how productive those rigs were going to be.  If for every 10 rigs I made I ran nine and sold one, I'd effectively guarantee that all my customers were competing for 720 (360) BTC and nothing more.  The customers I sold it to would have 10x the expectations - in the end, 9x of their expected yield I'd be keeping to myself with my supply of "unsold" rigs that I was running for, um, "extensive burn-in testing" before being sold.  (What, you think I'd let them sit idle on a shelf in a warehouse?)

If this product results in centralization of mining, I bet there will be cries for Bitcoin to switch to a CPU-based algorithm!

What about the already existing network of approximately 10 Thash/s? And if they only have 25 units and they meet their expectations, they have 500 Ghash/s. Certainly a nice farm, but nothing that will "dominate mining". Just because they are more efficient doesn't automatically render existing miners unprofitable, especially considering miners that have already passed the break even point of their investment.

I completely disregarded the 25 as I don't see it as any meaningful limit, and my argument would apply to anyone doing ASIC mining, not just these guys in particular.  I assume that the only reason that they'd do 25 is for prototyping to keep the risk manageable in case there were any bugs to be worked out, and if everything works as planned, nothing's stopping them from making however many they want.

Right.... if they have 25, the hard part is done.  Now, it's just set up the machine and press go.

[rjk]

Right.... if they have 25, the hard part is done.  Now, it's just set up the machine and press go.

Well if the initial production run were done on a MOSIS process (partial wafer with other users), they would need to tape out again for a full wafer, unless they wanted to spin the same wafer over again. The tape out is what costs the most, but having a proven-working design certainly helps things along when you want to graduate up to full wafers.

[farfiman]

I just read the whole 7 pages of this thread and sounds like a complete echo of the BFL thread at the start ( except we have BFL to compare to.)

If I had that kind of money to throw into bitcoin, I would buy the 6000 coins as stated a few messages above,  invest them in bitcoin ventures and have low risk ( accept maybe bitcoin price crashing....)

[defxor]

Also, if LargeCoin only sells to businesses the purchaser can get a VAT refund, at least, in Denmark (where I live), and I suspect also in most other EU countries. It seems almost like a waste of money to *not* purchase this as a business in the EU, where each member state is required to have a minimum VAT rate of 15%.

I'm assuming the prices quoted are ex VAT, thus there would be no refund. (25% VAT where I live)

However, the warranty needs to be sorted out if this is to be sold in Europe. In my country the manufacturer warranty period is three years, where the first 6 months all faults are considered to be from the manufacturing by default (after that the consumer has to show they're not due to wear or negligence).

Also, as far as I can see, the "DRM" issue is moot. Those still believing it to be a problem need to read ttul's posts in more detail (or if we could just get all that info merged into the top post)

My answer to the thread poll would unfortunately be "no". Not because of the price (it's likely not worse compared to keeping my money in the bank) but because of the warranty. Us europeans just shake our heads at the US "90 day warranty" madness.

[coblee]

I just read the whole 7 pages of this thread and sounds like a complete echo of the BFL thread at the start ( except we have BFL to compare to.)

Except BFL's initial price per performance target was outstanding and even with it's revised specs, it's still very good. LargeCoin's price per performance target is already not competitive. Who's to say they can even produce hardware up to these specs that their simulation showed.

[farfiman]

I just read the whole 7 pages of this thread and sounds like a complete echo of the BFL thread at the start ( except we have BFL to compare to.)

Except BFL's initial price per performance target was outstanding and even with it's revised specs, it's still very good. LargeCoin's price per performance target is already not competitive. Who's to say they can even produce hardware up to these specs that their simulation showed.

MAYBE if someone really wanted to mine somewhere with very high elec costs they could have an advantage with this product , but that person would real have to want to mine there badly...

[Brian DeLoach]

MAYBE if someone really wanted to mine somewhere with very high elec costs they could have an advantage with this product , but that person would real have to want to mine there badly...

It would have to be absurdly high. See here. It would take $.47 per kilowatt for five years before the C200 overtakes the Rig Box.

[defxor]

Hey there laptop manufacturer in your country, I went swimming with my laptop and it stopped working, since I bought it only 3 months ago I'd like to return it for a full refund under the warranty. 

"by default" means that if no obvious fault can be found the blame is on the manufacturer during the first 6 months.

If you turn it that laptop the manufacturer will examine it and kindle refuse your warranty claim since you've operated it provable outside the specifications.

[casascius]

Right.... if they have 25, the hard part is done.  Now, it's just set up the machine and press go.

Well if the initial production run were done on a MOSIS process (partial wafer with other users), they would need to tape out again for a full wafer, unless they wanted to spin the same wafer over again. The tape out is what costs the most, but having a proven-working design certainly helps things along when you want to graduate up to full wafers.

I assume that the company doing the multi client wafer would have a schedule and would be doing another one the next month or whatever period they run on.  Assuming their test worked, they could buy multiple slots on the next month's shared wafer and have 25*n chips, n being any number that suits their fancy.  And in that month while they waited for that production, they could put together all of those rigs, so by the time the chips themselves are ready, they can be dropped right into the rigs and fired up immediately.

[cablepair]

We will need some very specific details in order to take you seriously.

• What is the ASIC chip set you are using?
  Where are you physically located and will you allow members of the Bitcoin community to visit your place of business?
  Who or what is this escrow service and if it is not someone or some business we are already familiar with - why should we trust them?

We (Bitcoin Miners) are generally in a frantic FPGA mode right now but anyone would be stupid to fork over even a $4k deposit before we have some real specifics

[DeepBit]

Right.... if they have 25, the hard part is done.  Now, it's just set up the machine and press go.

Well if the initial production run were done on a MOSIS process (partial wafer with other users), they would need to tape out again for a full wafer, unless they wanted to spin the same wafer over again. The tape out is what costs the most, but having a proven-working design certainly helps things along when you want to graduate up to full wafers.

Wafer sharing is not the only method giving such low-quantity outputs.
It also may be a sASIC with some layers configured by ion-beam etching. Expensive, but literally single chip production is possible.

[fizzisist]

This is based on some kind of sASIC chips, right?

May I ask, just out of curiosity, how many chips are built into one of those rigs, or the other way round, how many MH/s you managed to cram into a single chip?

Any answer to this question?

Also, directed at pretty much everyone, why are we suddenly allowed to compare things to the BFL Rig Box as if it's real? Did we learn nothing from the "guarantees" of the Single? Compare this to products that exist now: BFL Single, the various Spartan-6 FPGA miners (see my sig for the best one!), GPUs.

Comparison to X6500s:

Code:

         X6500 (x50)   LargeCoin (x1)
Hash/s   20 G          20 G
Power    750-1000 W    100 W
Price    $25,750       $30,000

That's based on the bulk price of $515 each for the X6500.

Let's say the X6500s require 900 W more. That difference is less than 8 kWh per year. At $0.20/kWh, that's $1600. Plus, no DRM messiness. I'd still buy X6500s, thank you very much.

Still, in the long run, that power efficiency is amazing. Any proof that it's physically possible? Did you use a 28 nm process?

[defxor]

Where are you physically located and will you allow members of the Bitcoin community to visit your place of business?
Who or what is this escrow service and if it is not someone or some business we are already familiar with - why should we trust them?

I'm quite sure I've read that info in this thread already.

[DeepBit]

Comparison to X6500s:

Code:

         X6500 (x50)   LargeCoin (x1)
Hash/s   20 G          20 G
Power    750-1000 W    100 W
Price    $25,750       $30,000

That's based on the bulk price of $515 each for the X6500.

Something was missing from your table:

Code:

         Icarus (x53)  X6500 (x50)   LargeCoin (x1)
Hash/s  20G           20 G          20 G
Power   1000-1100 W   750-1000 W    100 W
Price   $24 600       $25,750       $30,000

Based on $464 per one Icarus unit (for batches of 30x, without separate PSUs. Next revision may be considerably cheaper).

[rjk]

Also, directed at pretty much everyone, why are we suddenly allowed to compare things to the BFL Rig Box as if it's real? Did we learn nothing from the "guarantees" of the Single?

They have already once adjusted the specs downward from 54.4 Ghash/s to 50.4 Ghash/s, and from 800 watts to 2500 watts. Although this is still vaporware, I'm pretty sure they learned from the Singles mistake and the current numbers are more accurate.

[ttul]

Of course if these are not the full blown type of ASIC but the structured ones or something hopefully there will also be R&D overhead for a full blown implementation also to take into account once the R&D that went into this one is accounted for.

This is quite true. The cost of fabrication is extensive, and much of that cost is for very very experienced engineers to validate the design before it is manufactured. Silicon wafers, and the literal fabrication process are only part of the cost. It's easy to understate the complexity of putting together an integrated circuit. We're talking about tens of millions of transistors, complex timing circuits, and little tiny wires that all have to fit together perfectly to result in a working product.

FPGAs are a great because they provide power efficiency within an order of magnitude of the ASIC, but without having to worry whether your simulation of the metal layout might not work just right with the lithography process. Unfortunately, FPGAs carry with them a couple of serious drawbacks versus ASICs:

1. FPGAs use a lot more logic to implement the same amount of logic. That logic eats up power.

2. FPGAs themselves are very complicated chips with high fabrication costs. That keeps the unit cost high even at high production volumes.

In our exploration last year of various options for increasing the efficiency of Bitcoin mining, we thoroughly evaluated the FPGA options, because of course using FPGAs instead of ASICs would have greatly sped things up and reduced our risk. But, what we discovered was that the power, space, and unit cost efficiencies of ASICs will always win out over FPGAs in the long run.

So, if you believe that Bitcoin is going to be around for many years to come, as we do, then you'll want to make a long term bet on ASICs rather than loading up on FPGAs that will be left in the dust as the difficulty factor is pushed higher by ASICs. Short term calculations based on today's difficulty factor and exchange rate are just not relevant. If your goal is short term profit and you fear the future of Bitcoin, then stick with FPGAs.

I am very pleased that there are others like Butterfly Labs who seem to also be working with ASICs. This is great news for Bitcoin - the more custom logic that gets pushed out there, the lower the risk that Bitcoin will be co-opted by botnets operated by nefarious groups. Because, last time I checked, your mom's PC didn't have a Bitcoin ASIC in it.

[bulanula]

Because, last time I checked, your mom's PC didn't have a Bitcoin ASIC in it.

Same is true with GPUs. Never seen a botnet of 5870s. Whoever is geeky enough to buy a 5870 knows how to be secure usually.

Quit the marketing and BS talking. You are not using custom ASIC as that requires millions to produce and nobody would take that risk.

Probably using sASIC in the best case ...

Also, this rig box and box that makes 50 ghash/s or 20 ghash/s will only centralize BTC to the few that can afford paying over your threshold amount of 10k+ which I doubt many "casual" miners are willing to pay. Stop trying to sound like you are making BTC more secure because this simply is not it.

ASIC will be devastating to BTC and make it 51% easily in the early true custom ASIC times until everybody gets same HW and drops GPUs.

ArtForz mined about 60k BTC in the days everybody was CPU mining with a single 4870 AFAIK. Someone with secret ASIC could do the same and 51% this easily taking all our coins.

[Proofer]

Because, last time I checked, your mom's PC didn't have a Bitcoin ASIC in it.

Same is true with GPUs. Never seen a botnet of 5870s. Whoever is geeky enough to buy a 5870 knows how to be secure usually.

Quit the marketing and BS talking. You are not using custom ASIC as that requires millions to produce and nobody would take that risk.

Probably using sASIC in the best case ...

As I read the ttul post it pretty much says that the current offering is based on sASIC.

[rjk]

some bullshit

Pipe down over there and let the adults talk in a reasoned manner please. If you don't agree with some claims, please present your arguments and questions in a less offensive format. Thank you.