ASIC = The end of decentralized mining

[P4man]

Using this IP core a 10x10mm die @130nm (200kgates/mm2) would give about 3Ghash/s.

Hmm.. not sure where you got the 130nm figures from. I only see 180 and 90nm if you click the asic button.
On 90nm, one core is 50,800um2 or 0.0508 mm2
One core is 7.75Mbps / MHz. On 90nm they project 500 Mhz, so thats 3.8 Gbps or ~7.5 bitcoin MH.
a 100mm2 chip would therefore yield 15 GH. A single 200mm wafer over 4 TH.

But as DnT pointed out, thats unlikely to be anywhere near optimal for bitcoin.

edit: revised math

[1714038854]

1714038854

[1714038854]

1714038854

[1714038854]

1714038854

[1714038854]

1714038854

[mrb]

It is my opinion that FPGA will be #1 and ASIC is not and I am willing to bet on it.

However you are right its impossible to quantify in those terms

can anyone think of a way we can quantify the bet?

With eldentyrell's TML Spartan6 bitstream requesting mining jobs from centralized servers, he should be able to tell exactly how many Ghash/s of Spartan6 there are

[spiccioli]

Well,

there is at least one thing we're not considering right now: BFL could start selling ASICs which are underclocked so that while being better than the average FPGA they don't give out the maximum available hashing power.

Doing so they can, next year, sell their ASIC/2 board, with higher clock and same price just like GPU vendors do all the time.

I mean what makes you think that the first board will be 100x when a 10-20x underclocked or otherwise "crippled" board can be sold as well?

spiccioli.

[mrb]

we use the same Spartan6LX150 FPGA chip that all the other manufacturers are , also our unit is the cheapest quad unit currently available

Ztex offers large discounts when buying their 1.15y boards in quantity, beating your unit price: 719 EUR (~$900) in qty 25-49.
Does it mean you would match their prices?

[DeathAndTaxes]

Well,

there is at least one thing we're not considering right now: BFL could start selling ASICs which are underclocked so that while being better than the average FPGA they don't give out the maximum available hashing power.

Doing so they can, next year, sell their ASIC/2 board, with higher clock and same price just like GPU vendors do all the time.

I mean what makes you think that the first board will be 100x when a 10-20x underclocked or otherwise "crippled" board can be sold as well?

spiccioli.

If BFL has a monopoly they likely will.  It maximizes their profit and allows them to pay down the NRE faster.  Anytime sales get slow they can just increase the MH/$ metric. 

If 2+ companies have ASIC based products then competition likely will drive prices down much faster and that will be disruptive in the short term.  The economics of ASICs (high fixed cost very very low per unit cost) almost guarantee a pricing war.  I mean if you can sell more units by cutting your price from a 1000% markup to "only" a 900% markup wouldn't you?

[Dexter770221]

From your link http://rijndael.ece.vt.edu/sha3/publications/DSD11SHA3.pdf I compute 400 Mhash/J for a 130nm ASIC. So probably ~1000 Mhash/J at 40nm. This is 50x better than a 45nm FPGA (Spartan6 = 20 Mhash/J).

Care to share your math? I suspect you forgot a bitcoin hash consists of two SHA256 hashes.

I took this into account:
1.51 Gbps / 8 (bits/bytes) / 64 (bytes/block) / 2 (SHA-256 blocks per Bitcoin hash) / 0.0037 (Watt) = 398.5 Mhash/Joule

Should be 68(bytes/block) its then 395 MH/Joule. But we need to consider that is a very good design maked by professionals. BFL design is only 450MH/35W=12.85MH/W. They can produce ASIC so quickly becuse they propably using HDL code used in Singles. Moving from FPGA to ASIC (with same design) should improve efficency no more than 10-20 times. Thats why I think 200MH/W we will see from that ASICs.

[P4man]

Well,

there is at least one thing we're not considering right now: BFL could start selling ASICs which are underclocked so that while being better than the average FPGA they don't give out the maximum available hashing power.

Doing so they can, next year, sell their ASIC/2 board, with higher clock and same price just like GPU vendors do all the time.

I mean what makes you think that the first board will be 100x when a 10-20x underclocked or otherwise "crippled" board can be sold as well?

spiccioli.

They might, but then again, its no different as selling the same chip for $1000 initially and dropping prices to $100, $50 etc over time.  The net effect is the same; per GH price will likely start close to FPGA prices and eventually drop by somewhere between one and two orders of magnitude.

Crippling does have the potential benefit of being able to sell "upgrades". Flash the bios or something and double, quadruple or ten fold your hashing power.  

[DeathAndTaxes]

A SHA-256 block is always 64 bytes (512 bits).  Even if the block is partially empty it will be a zero padded 512 bit block.

Bit Thunderdome: 512 bits enter, 256 bits leave.

[Syke]

Look at how long it took BFL to release products based on regular FPGAs. There's no way they'll release a "full custom ASIC" before the block reward halves. ASICs like that are a long way off. There's plenty of time for FPGA profits.

[crazyates]

when will bitcoin asic's be available for certain?

2016 

[Garr255]

when will bitcoin asic's be available for certain?

2016 

+/- 4 years

[rjk]

when will bitcoin asic's be available for certain?

If you believe BFL's words, Friday the 15th.

[BR0KK]

Didn't the same thing happened when gpus entered the bitcoin mining area ......

What about bitstream and Stream developers. They simply can improve their code. As for my understanding Asics are xxgh for their lifetime?

[lame.duck]

Saying: FPGA manufacturers are enjoying sales in the many tens of thousand
...is talking about unit quantity sales and not money, if you meant money then you need to learn to phrase your thoughts right.

There are Xilinx marketing paper about ASIC vs. FPGA comparisation that operate with per piece cost costumers get buying the FPGAs in 10 k quantities. Needless to say that Xilinx FPGAs win

[Transisto]

...that operate with per piece cost costumers get buying the...

 

If BFL were to offer lifetime price protection it would greatly reduce risk to miners while also motivating enough initial funding.

[jjshabadoo]

I also could have clarified some points and stated that I do not believe ASIC technology is BAD for bitcoin. Allowing something like ASIC to be rolled out under traditional free market principles is going to be bad for bitcoin in my opinion.

High Initial cost+low production cost+low marginal cost is not going to be good for the people who buy the first units.

I guess many here don't give a shet about those people, but I happen to believe its unfair and could have backlash.

In the long run ASIC will be great for bitcoin and bitcoin has to get there eventually. Maybe the only path is a painful one.

Maybe an OpenASIC involving a bunch of motivated bitcoin investors could spread the cost amongst those who want to put up money and they can have access for a period of time and then let the floodgates roar open.

[cablepair]

When they come, I'll be ready
I hear their voices inside
The stars in the heavens are moving
Soon they will align

Gavin, god of Bitcoin
Let me die with a FPGA in my hand
Raise your hands, GPUs in the wind
Brothers of Bitcoin together again
With blood in our voices we ride
We'll fight till we win or we'll fight until we die  

[lemonz]

when will bitcoin asic's be available for certain?

If you believe BFL's words, Friday the 15th.

Let's be fair rjk, BFL said:

Announcements regarding abbreviated shipping schedules and future product (BitForce SC) release will be made on June 15th.

Friday will be an announcement regarding the ASICs, not that they will be available.  Best case scenario that I'm hoping for is that they release performance information and start accepting pre-orders.

[BR0KK]

bfl preorder more a prepaid sollution

[rjk]

Friday will be an announcement regarding the ASICs, not that they will be available.  Best case scenario that I'm hoping for is that they release performance information and start accepting pre-orders.

Sure, this is the most likely scenario. Although I had a small glimmer of hope because they made various statements to the effect that SC production was entirely separate from Single/MR production, and wouldn't suffer from similar delays, and etc etc etc. Personally, I'm doubting it, but hey! maybe they can still turn over a new leaf.