Definition
Temperature Sensor is the small integrated circuit on a mining device that measures heat in real time, feeding the firmware the readings it uses to drive fans, throttle clocks, and shut a board down before it cooks. On an ASIC hashboard it usually appears as one or more LM75-class chips reporting the printed-circuit-board temperature alongside the silicon’s own die temperature.
Also known as: temp sensor, thermal sensor, board sensor (when referring to the PCB-mounted IC).
What the sensor actually measures
Most Antminer-class hashboards carry small LM75-pin-compatible parts such as TMP451, NCT218, or ADT7461. Each chip exposes two channels: a “local” reading for the board itself and a “remote” reading wired to the hottest corner of the silicon. The firmware reads these over an I2C bus, typically at addresses like 0x48 and 0x4C, sometimes routed through an on-board microcontroller or an I2C multiplexer when several boards share one bus.
Sensor counts vary by design. Modern direct-read boards may carry as few as two sensors, a typical air-cooled hashboard runs around four (intake and exhaust corners), and dense water- or immersion-cooled variants can carry seven or eight to track extra coolant rails. Separately, the ASIC chip itself has an internal temperature register, so firmware can cross-check the board sensor against the die reading and apply a fixed calibration offset (commonly around 15 degrees) to estimate junction temperature when the on-die reading is unavailable.
How firmware uses the readings
Temperature sensors are the input to a closed-loop control system. Firmware compares live readings to staged thresholds and acts at each tier: drive fans harder, step the device down to a lower power profile, or, at the danger limit, cut a board’s voltage entirely. In the firmware stacks studied across the ecosystem, board-PCB and chip-die domains are tracked in parallel, with the chip allowed a higher ceiling than the PCB because the silicon junction always runs hotter than the substrate around it.
Good firmware also treats sensors defensively. A jittery or NACKing sensor is filtered against its peers, dropped if it keeps disagreeing, and the device keeps running as long as at least one valid reading per board survives. A sensor read failure should always fail safe toward more cooling, never less. This is exactly the kind of reliability work the closed-beta, GPL-3.0 custom firmware tradition has pushed forward, standing on the shoulders of the open tuning stacks that came before it.
Why a home miner should care
If you run an ASIC in a basement, garage, or as a space heater, the temperature sensor is your first line of defense and your main tuning dial. Sustained chip temperatures well below the danger threshold are the single biggest lever on hardware longevity; cooler silicon and cooler power supplies simply last longer, which matters when you are buying used gear. Watching the spread between intake and exhaust sensors also tells you whether airflow is clean or whether dust and lint are choking a board.
For quiet residential setups, sensor accuracy is what lets you safely undervolt and underclock instead of blasting fans. A well-calibrated thermal loop can hold a target board temperature while keeping fan duty low, which is the difference between a whisper-quiet heater and a jet engine in your living room. When you shop for hardware on the miners catalog or troubleshoot a flaky unit through the ASIC troubleshooting guides, a dead or drifting temperature sensor is a common culprit behind unexplained shutdowns and throttling.
Common failure modes
A sensor can fail open (reads an impossibly low value, fooling firmware into spinning fans down) or fail short (reads a fixed high value, triggering needless throttling or shutdown). Loose I2C wiring, a cracked solder joint near the connector, or a damaged sensor IC will surface as “board temp” errors in the miner log. Because the firmware powers a board off at the danger threshold, a misreading sensor often masquerades as a hashboard fault, so checking the sensor is a smart early step before assuming the worst about the control board or the chips themselves.
Related terms: Hashboard, ASIC Chip, Control Board, Noctua Fan, Undervolting, Thermal Paste
In Simple Terms
Sensors monitoring chip and ambient temperatures to prevent overheating and control fan speeds.
Temperature Sensor is the small integrated circuit on a mining device that measures heat in real time, feeding the firmware the readings it uses to drive fans, throttle clocks, and shut a board down before it cooks. On an ASIC hashboard it usually appears as one or more LM75-class chips reporting the printed-circuit-board temperature alongside the silicon's own die temperature.
Also known as: temp sensor, thermal sensor, board sensor (when referring to the PCB-mounted IC).
What the sensor actually measures
Most Antminer-class hashboards carry small LM75-pin-compatible parts such as TMP451, NCT218, or ADT7461. Each chip exposes two channels: a "local" reading for the board itself and a "remote" reading wired to the hottest corner of the silicon. The firmware reads these over an I2C bus, typically at addresses like 0x48 and 0x4C, sometimes routed through an on-board microcontroller or an I2C multiplexer when several boards share one bus.
Sensor counts vary by design. Modern direct-read boards may carry as few as two sensors, a typical air-cooled hashboard runs around four (intake and exhaust corners), and dense water- or immersion-cooled variants can carry seven or eight to track extra coolant rails. Separately, the ASIC chip itself has an internal temperature register, so firmware can cross-check the board sensor against the die reading and apply a fixed calibration offset (commonly around 15 degrees) to estimate junction temperature when the on-die reading is unavailable.
How firmware uses the readings
Temperature sensors are the input to a closed-loop control system. Firmware compares live readings to staged thresholds and acts at each tier: drive fans harder, step the device down to a lower power profile, or, at the danger limit, cut a board's voltage entirely. In the firmware stacks studied across the ecosystem, board-PCB and chip-die domains are tracked in parallel, with the chip allowed a higher ceiling than the PCB because the silicon junction always runs hotter than the substrate around it.
Good firmware also treats sensors defensively. A jittery or NACKing sensor is filtered against its peers, dropped if it keeps disagreeing, and the device keeps running as long as at least one valid reading per board survives. A sensor read failure should always fail safe toward more cooling, never less. This is exactly the kind of reliability work the closed-beta, GPL-3.0 custom firmware tradition has pushed forward, standing on the shoulders of the open tuning stacks that came before it.
Why a home miner should care
If you run an ASIC in a basement, garage, or as a space heater, the temperature sensor is your first line of defense and your main tuning dial. Sustained chip temperatures well below the danger threshold are the single biggest lever on hardware longevity; cooler silicon and cooler power supplies simply last longer, which matters when you are buying used gear. Watching the spread between intake and exhaust sensors also tells you whether airflow is clean or whether dust and lint are choking a board.
For quiet residential setups, sensor accuracy is what lets you safely undervolt and underclock instead of blasting fans. A well-calibrated thermal loop can hold a target board temperature while keeping fan duty low, which is the difference between a whisper-quiet heater and a jet engine in your living room. When you shop for hardware on the miners catalog or troubleshoot a flaky unit through the ASIC troubleshooting guides, a dead or drifting temperature sensor is a common culprit behind unexplained shutdowns and throttling.
Common failure modes
A sensor can fail open (reads an impossibly low value, fooling firmware into spinning fans down) or fail short (reads a fixed high value, triggering needless throttling or shutdown). Loose I2C wiring, a cracked solder joint near the connector, or a damaged sensor IC will surface as "board temp" errors in the miner log. Because the firmware powers a board off at the danger threshold, a misreading sensor often masquerades as a hashboard fault, so checking the sensor is a smart early step before assuming the worst about the control board or the chips themselves.
Related terms: Hashboard, ASIC Chip, Control Board, Noctua Fan, Undervolting, Thermal Paste