Mining Firmware

An ASIC is a specialized computation chip, but the chip alone does not know which pool should receive the device’s results. Firmware handles that and other tasks. Braiins OS, for example, lets the operator configure a pool, the worker name identifying the device in the pool account, a power target and cooling. Martina can therefore run the same software version with a different operating profile. The version identifies the software; the profile determines how she is using it. An accessible administration page does not prove that work is being accepted. She also needs to check the pool’s records. A pool usually records partial results called shares. They demonstrate computational work, but not every accepted result is a new Bitcoin block. [1] [8]

Before uploading anything, Martina checks the documentation against her particular device. The control board, chip-bearing hashboards and power supply unit, or PSU, all matter. In Braiins documentation, E2 suggests probably unsupported hardware, while P5 identifies an incompatible power supply model. These are diagnostic clues for this product, not universal miner codes. Moving a control board with software already installed does not establish support for another machine. Compatibility concerns the complete assembly, not merely a file that uploaded successfully. [2]

BITMAIN’s guide requires firmware for the correct model. It separately explains Keep Settings, the option to preserve configuration. With it enabled, the documented update retains pool and network settings; otherwise defaults are restored. Martina therefore saves the necessary configuration beforehand. Resetting network settings can change the IP, the device’s address on the local network. Its old page may stop responding without the machine being damaged. Keeping settings is not a backup of the previous software or a promise of an easy rollback. [3]

For supported setups, Braiins describes booting from a removable SD memory card while leaving the original software in internal storage untouched. Removing the card on those setups lets the original firmware and its settings boot again. Installation into NAND, a type of internal memory, instead writes software inside the device. The distinction cannot be generalized to every model: available procedures depend on the board and version. Martina checks the installation method and recovery options before changing anything, rather than after the device stops responding. [4]

In Braiins OS, Target Temperature is the temperature around which fan operation is regulated. The higher Hot Temperature threshold can trigger Dynamic Performance Scaling, which lowers the configured power-consumption or hashrate target when overheating, but only if that feature is enabled. Hashrate here means the number of hashing attempts per second. Dangerous Temperature is a further critical limit at which mining stops. These are not interchangeable names for a single thermostat. When Martina sees performance fall, she also checks for a thermal cause. Disabling protection cannot fix insufficient cooling; a better screen reading may conceal a bigger problem. [1]

Pool groups in Braiins allow computing power to be divided. A Quota ratio of 2:1 targets 2/3 of the work to the first group and 1/3 to the second. Alternatively, fixed proportions must add up to 100%. If every pool in a group becomes unavailable, allocation is recalculated without that group. The ratio describes intended routing, not a guaranteed share of revenue over any short period. Martina checks availability and accepted work: a configured percentage alone cannot tell her what happened during an outage. [1]

For version 23.04 and newer, Braiins OS documentation describes license renewal every 60 minutes and an acquired license’s validity of 12 hours. A lost connection therefore need not cause an immediate restriction. If a license expires or cannot be acquired, however, the software reduces hashing performance according to its license terms. That is behaviour of a particular product, not an automatic invalidation of Bitcoin blocks. Martina also distinguishes the dev fee, payment to the software provider, from the pool fee. Any relationship between them needs checking against current terms. [5]

In advice updated in 2022, BITMAIN suggests an initial sample of 10–30 devices observed for at least 24 hours. The point is to catch problems before changing the whole farm, not to provide a universal guarantee or require a small operation to own that many machines. Martina compares accepted work, errors, temperatures and restarts under comparable conditions. J/TH means joules per terahash: the energy needed for a trillion hashing attempts. Comparisons must cover the same device components and the same measurement period. Higher local speed may not be an improvement if power consumption, rejected work or downtime increases. [6] [1] [7]