Every ANTMINER carries its own thermal cut-off. The numbers below are the maximum on-board sensor temperatures Bitmain builds into each model’s protection program — the point at which the firmware throttles or shuts the machine down to keep the silicon alive. There are two of them per model: a PCB temperature (read by sensors on the hashboard) and, on some models, a chip temperature (read from inside the ASIC die). These are ceilings, not targets. A healthy miner should run with real headroom below them; if your live readings are creeping toward these figures, you have a cooling or dust problem to fix before the board fixes it for you.
This guide covers the per-model trip points. For the room and airflow conditions that keep you under them — intake temperature, humidity, altitude — see the companion environment guide linked at the end.
The two temperatures your miner actually measures
A common mistake is to treat “miner temperature” as one number. It isn’t. Each hashboard reports temperature to the control board — S9/S17/S19-class units typically carry three (Chain0, Chain1, Chain2), though other models differ — and the firmware watches two distinct readings per chain:
- PCB temperature — from temperature sensors soldered to the hashboard itself, typically near the air inlet and outlet. This tracks how hot the board substrate is getting. It is the value most protection ceilings in the table below are based on.
- Chip (junction) temperature — read from a sensor integrated into the ASIC die. On older units (S9, T9, Z9-class) the firmware reports the hottest chip directly; on newer S17/S19-class boards it is derived from the ASIC’s internal sensor. This is always higher than the PCB reading because it is measured at the heat source.
Because the chip die always runs hotter than the board under it, the chip ceiling (where published) is well above the PCB ceiling. Both are watched independently — whichever crosses its limit first triggers protection.
Maximum operating temperatures by ANTMINER model
Values are in degrees Celsius, as published in Bitmain’s protection tables. A dash (—) means the model does not expose a separate chip-temperature limit; protection runs off the PCB sensor alone.
| Model | Max PCB Temp (°C) | Max Chip Temp (°C) |
|---|---|---|
| E9 | 80 | — |
| L7 | 80 | — |
| S19 XP | 75 | — |
| S19a Pro | 80 | — |
| S19a | 80 | — |
| S19j | 85 | — |
| S19j Pro | 85 | — |
| S19 Pro | 80 | — |
| S19 | 80 | — |
| T19 | 80 | — |
| Z15 / Z15j / Z15e | 80 | — |
| S17e / T17e / S17+ | 80 | — |
| T17+ | 75 | — |
| S17 / S17 Pro / T17 | 75 | — |
| S15 | 80 | — |
| T15 | 80 | — |
| S11 | 95 | — |
| Z9 Mini | 80 | 125 |
| Z9 | 80 | — |
| V9 | 80 | — |
| S9 | 90 | 135 |
| S9i | 85 | — |
| S9j / S9 Hydro | 95 | — |
| T9 | 90 | 135 |
| T9+ | 95 | 125 |
| S9 SE / S9k | 85 | — |
| B7 | 75 | — |
| K5 | 80 | — |
| D5 | 85 | — |
| DR3 / DR5 | 85 | — |
| E3 | 65 | — |
| L3+ / L3++ | 90 | — |
| D3 | 85 | — |
| A3 | 85 | 115 |
| X3 | 80 | 127 |
| B3 | 80 | 127 |
Reading the table: the newer, denser boards (S19 XP, T17+, S17/S17 Pro/T17) sit at the lowest PCB ceiling — 75°C — not the highest, while the lower-density S9/S11-class boards carry limits of 90–95°C. The pattern is consistent with tighter chip pitch and higher power density leaving less thermal margin on the denser boards, though the exact reasoning behind each trip point is Bitmain’s. Either way, do not assume a “bigger” miner runs hot safely; the opposite is usually true.
Normal running vs. the trip point
The table lists ceilings, not where a healthy machine lives. Under steady load in a well-ventilated space, PCB sensors should report comfortably below their limit, leaving margin for a hot afternoon, a clogging filter, or a fading fan. The gap between your live reading and the ceiling is your thermal headroom, and it is the single most useful number to watch over time. Headroom that shrinks month over month in the same room is the classic signature of dust on the heatsinks and fans losing RPM — a slow failure that ends in a shutdown on the first genuinely warm day. Track the trend, not just today’s snapshot.
What the miner does when it reaches the ceiling
Bitmain firmware defends the ceiling in stages rather than an instant kill:
- Fans ramp to 100%. The first response to rising board temps is maximum fan speed. If your fans are pinned at full RPM at idle-normal hashrate, the machine is already fighting heat — investigate before it escalates.
- Frequency throttles (firmware-dependent). Some firmware — newer stock releases and most third-party builds — will lower chip frequency as temperatures keep climbing to cut heat output, which shows up as reduced hashrate. Classic stock Bitmain firmware often skips this intermediate step and goes straight from full fans to shutdown.
- Chain or unit shutdown. If a board still crosses its PCB or chip limit, protection halts the affected hashboard — or the whole miner — and logs a high-temperature event. The unit will not resume hashing until it reads a safe temperature again.
A machine that repeatedly hits the shutdown stage is not “just hot” — it is telling you a fan, a filter, the ambient air, or the thermal paste under a heatsink has failed. Chronic over-temperature shutdowns cook solder joints and shorten chip life; fix the cause, don’t just power-cycle.
How to read your live temperatures
- Web dashboard. The miner’s status page lists a temperature per chain. Most Bitmain firmware shows the PCB reading; some also expose the chip/derived value. Check every chain — one board running hotter than the others points to that board’s airflow or heatsink.
- Monitoring API. The cgminer-style API (and any fleet dashboard reading it) returns per-chain
tempand, where available, chip temperature, so you can log and alert on the trend instead of eyeballing it. - Set alerts below the ceiling. Alarm on a value with margin under the trip point — that gives you time to act while the miner is still hashing, instead of finding out after it has already shut down.
Common mistakes
- Confusing intake air temperature with these numbers. The table is board and chip sensor temperature, not room temperature. A 30°C room can still push a dusty S19 XP board past 75°C. Ambient limits are a separate spec — see the environment guide.
- Ignoring one hot chain. Averaging the boards hides a single failing hashboard. Always look per-chain.
- Blaming the sensor. A board reading implausibly high or low is more often a real thermal or airflow fault (or a dislodged heatsink) than a bad sensor. Rule out the physical cause first.
- Silencing the alarm instead of the heat. Repeated high-temp shutdowns are a symptom. Clear the dust, replace the fan, or cool the room — don’t just reset and walk away.
When to stop and get help
If a board keeps tripping its limit after you have cleaned it, confirmed all fans spin at full RPM, and brought the intake air into range, the fault is on the hardware — degraded thermal interface material, a lifted heatsink, or a failing temperature sensor. At that point the board needs bench diagnosis, not another reboot. Persistent single-chain over-temperature is one of the most common reasons a hashboard lands on the repair bench.
Related: pair this with our guide to optimal environmental conditions for operating ANTMINERs for the intake-air and humidity side of the equation, walk through fixing an ANTMINER “temperature too high” fault when a board is tripping, and use the ASIC fault finder to isolate a bad hashboard or start a repair when the heat won’t come down. For the opposite problem — sub-zero intake air, cold starts and condensation — see setting up and maintaining miners in cold climates.
