Simplified version of an FPGA bitcoin miner. Although not developed to pull data blocks from a blockchain pool, the server.py script can easily be modified to do so. At the moment, it generates random blocks to feed the FPGA via a TCP socket connection.

Essentially, the FPGA is responsible for finding the nonce corresponding to an integer, which is concatenated with the blockchain block and then hashed using SHA-256, that results in an hash starting with a given number of zeros, also known as difficulty.

• ≈350K hashes per sec: With word input of 1024 bits (2 blocks of 512 bytes)
• ≈2.8s per block in the FPGA: With network interaction and validations (difficulty = 5)
• ≈14.4s per block in a CPU: CPU Intel i5-7500 3.5GHz, 8GB RAM (difficulty = 5)
• ≈5x faster: 5x more bitcoins

The miner is essentially based on the following IP cores.

This interface is in charge of receiving and storing the Block Header and the processed mining difficulty mask. It receives 21 32-bit words (20 - block header | 1 - zeros mask) and enables the Mining Core.

This module is in charge of generating a 256 bit hash, using the SHA-2 algorithm.

This component is in charge of the search for the nonce that produces an hash with the predefined difficulty. Each time the nonce is incremented, updates the block header and starts the SHA-256 component to generate the hash.

This module is in charge sending back to the Microblaze the resultant hash and the respective nonce.

This module is in charge of controlling the refresh rate, brightness and the displays and respective values of the 7 segment displays.

This component acts like animating the board LEDs making them look like a loading bar. It stores the values and counter which is incremented each 2 seconds (or up until 128 times less) which controls the speed of the animation.

Through software, a TCP socket is open to communicate with the mining pool exchanging block headers. Parses generic blockchain headers and communicates with the various auxiliary IPs to calculate the nonce.

The first step is to run to compile and run the files on the board project folder on a compatible FPGA (Xilinx), via the SDK app.

By default, the board will be configured to listen a TCP connection on 192.168.1.10:7. Then run python3 server.py to start feeding the board with blocks so it can compute SHA-256 hashes over those blocks and return the proof of work nonce.

Diogo Ferreira @ UA
Pedro Martins @ UA
2018