ASIC - Whats the big deal?

[os2sam]

I used to be a field service technician for a vertical market manufacturer.  As such I was only peripherally involved with the actual manufacturing process.

When a new product was designed the first run usually used FPGA's for the proprietary circuitry.  When enough units were sold and the design was thought to be solid then they would mask the silicon and have ASIC's made which are allot cheaper than FPGA's.

So I'm at a bit of a loss as to what the big deal is about using ASIC's and the expected huge increase in performance over FPGA's?  Can someone please enlighten me?

It would seem to me that the design would be implemented in FPGA's and then when the design is solid and quantities sold warranted it then move to ASIC's for the lower per chip cost.  But that doesn't seem to be the consensus here.

Thanks,
Sam

[1715239096]

1715239096

[1715239096]

1715239096

[1715239096]

1715239096

[1715239096]

1715239096

[1715239096]

1715239096

[Gabi]

Here the design is the bitcoin mining. And it is solid, we have been mining for 3 years and more!

The same more or less happened here. First people mined using their GPU, since they already had them. Then they noticed that bitcoin mining is here to stay and specialized and invested in FPGA. Using FPGA imply buying them (and they aren't cheap) and making the bitstream but once you do that, they use much much much less energy than a GPU. So it's more efficient for mining, moar profit!

Then some ppl started to invest into making an ASIC. But not for lower per chip cost. Well, this is a reason, but not the main one. The main ones are speed and efficiency (aka energy used). An ASIC is a chip specialized for mining, you put every transistor as you wish to mine faster as possible. An FPGA is a chip made by someone wich well, can do everything, not only mining, it's much much slower.

ASIC is FAST, very fast. The bfl asic "single" will do 40Gh/s for 1299$. For that price you can get what, a 3Gh/s computer? Or a 1.6Gh/s FPGA.

[os2sam]

Here the design is the bitcoin mining. And it is solid, we have been mining for 3 years and more!

The same more or less happened here. First people mined using their GPU, since they already had them. Then they noticed that bitcoin mining is here to stay and specialized and invested in FPGA. Using FPGA imply buying them (and they aren't cheap) and making the bitstream but once you do that, they use much much much less energy than a GPU. So it's more efficient for mining, moar profit!

Then some ppl started to invest into making an ASIC. But not for lower per chip cost. Well, this is a reason, but not the main one. The main ones are speed and efficiency (aka energy used). An ASIC is a chip specialized for mining, you put every transistor as you wish to mine faster as possible. An FPGA is a chip made by someone wich well, can do everything, not only mining, it's much much slower.

ASIC is FAST, very fast. The bfl asic "single" will do 40Gh/s for 1299$. For that price you can get what, a 3Gh/s computer? Or a 1.6Gh/s FPGA.

Hmm, so ASIC is just faster than FPGA.  I kind of figured that must be the reason but wasn't really sure.  In my past experience an FPGA was used kind of like a breadboard for field testing a circuit before it was made into the ASIC and performance, like we would need in Bitcoin, wasn't a need or concern for the most part.

Thanks for the reply and explanation.  That helps.
Sam

[Dexter770221]

It's not about to be FAST. It's about to be efficient. Best FPGA results so far is something 25MH/W (GPU is 3-5MH/s). ASIC will bring 1000+ MH/W. No heat, no noise and 40GH/s under the hood. Just profit... Well at least for first difficulty change. After that, no heat, no noise, no profit

[crazyates]

It's not about to be FAST.

Yes, yes it is. I bought ASICs because they're fast. End of strory. 

[Chrstian]

Here the design is the bitcoin mining. And it is solid, we have been mining for 3 years and more!

The same more or less happened here. First people mined using their GPU, since they already had them. Then they noticed that bitcoin mining is here to stay and specialized and invested in FPGA. Using FPGA imply buying them (and they aren't cheap) and making the bitstream but once you do that, they use much much much less energy than a GPU. So it's more efficient for mining, moar profit!

Then some ppl started to invest into making an ASIC. But not for lower per chip cost. Well, this is a reason, but not the main one. The main ones are speed and efficiency (aka energy used). An ASIC is a chip specialized for mining, you put every transistor as you wish to mine faster as possible. An FPGA is a chip made by someone wich well, can do everything, not only mining, it's much much slower.

ASIC is FAST, very fast. The bfl asic "single" will do 40Gh/s for 1299$. For that price you can get what, a 3Gh/s computer? Or a 1.6Gh/s FPGA.

Get me a 3 Gh/s computer ill pay you 130 BTC!!!! please!! (ill use escrow)

[Gabi]

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

Have fun.

BitForce Jalapeno  3.5 GH/s  -  $149

Yes i know, they are still to be released

[flynn]

It would seem to me that the design would be implemented in FPGA's and then when the design is solid and quantities sold warranted it then move to ASIC's for the lower per chip cost.  But that doesn't seem to be the consensus here.

FPGA waste a lot of internal gates/latches/buffers/whatever to implement a design. So ASICs can do what a FPGA does, but not the way around in most cases.
For mining, you want to use every gate available to perform massively parallel computations. So the design of the ASIC has to be made specially for that ASIC and will not just be a port of what was done for the FPGA.

The other side of the (bit)coin being : designing the function for the ASIC is 10% of the work. simulation and most specifically Thermal response simulation is the hard part. You pretty well can come up with a design that behave very poorly thermically speaking once you have the final chip, and which finally runs at 1/10th of the originally designed speed.

And redoing it is a no-go, since it costs all that money again.

So, ..., big surprises are expected ...