Hacker News Reader

Comment by mutagen | Hacker News Reader

With tongue firmly in cheek, I wonder how long before it becomes cost effective to mine Bitcoin with raw 74xx logic?

Bitcoin miners are already using custom ASIC's on expensive, process nodes. Anyone using discrete logic would be running at a snail's pace compared to them.

What about mining ethereum on FPGA, would it be suited?

I don't know specifics on Ethereum. Basically, it has to be within the hardware constraints of FPGA's where CPU's, GPU's, and ASIC's don't convey an advantage.

Do you know how bitcoin miners use FPGAs? They're used for communication somehow. The Antminer S7 uses an SoC FPGA which allegedly was being used with ASICBoost.

You can implement many forms of I/O on FPGA's that let you bridge things that otherwise can't communicate for whatever reason. The FPGA can also be used to accelerate I/O vs a software implementation. For example, many network devices offload TCP/IP from the host. Finally, the connecting systems might have the necessary I/O built-in but FPGA accelerates some computational part of it. You see that in compression, TLS or IPsec engines.

I don't know what Bitcoin people are doing. The above are common use cases, though.

That's (sort of) asked and answered in the comments :)

"it would be theoretically possible. But if you calculate that... it isn't. For example: a DE0-Nano board ($79) mines at 12 MH/s. It has 22320 logic elements. Assuming the slices are equivalent, you would have to have 22320 of these boards. Each containing 12 7400 chips."

http://www.potatosemi.com/potatosemi2010/products_74.html

You'll want fast 74-series logic!

Reply on news.ycombinator.com