Antminer S17 – Temperature Too High

Power off at the breaker or unplug the miner for 15 full minutes. Do not hot-restart — you will trip the same shutdown, waste another thermal cycle on tired pads, and stress the capacitors. Let the aluminum heatsink bar, chips, and PCB cool completely before any further action. The S17 holds heat in its heatsink longer than most ASIC operators expect.

Verify intake air temperature with a thermometer placed at the front grille of the miner — not room center, not ceiling. The S17's rated intake limit is 30 C. Above that, the firmware cannot keep the chips under threshold. If ambient is too high, relocate the miner, add room ventilation, or restrict operation to cooler hours until you solve the environmental problem. Log the reading before changing anything else.

Check the dashboard for pending firmware updates. Some S17 stock firmware builds from 2019-2020 shipped with a latched-fault bug where ERR_TEMP_HIGH would persist after the chip cooled. If you are on an older image, a controlled firmware refresh via SD card can clear a ghost fault. Check support.bitmain.com/downloads for the latest stock image matching your hardware revision before flashing.

Confirm the miner's physical orientation and airflow path. S17s are designed for horizontal airflow only. Stacking them vertically, laying them on their side, blocking the intake with furniture or fabric, or deploying them in a sealed closet within 30 cm of another miner will trigger ERR_TEMP_HIGH within 10-15 minutes of runtime — even on a mechanically perfect unit.

Shop-vac and compressed-air the heatsink fins on every hashboard. Pay attention to the leading intake edge where most dust collects. Work from the intake side outward — blowing dust the other way packs it deeper between the heatsink bar and the PCB where you cannot reach without full disassembly. Budget 30-45 minutes. On S17s older than 6 months, expect surprising dust volume even in clean rooms.

Inspect all four axial fans. Chassis open, power off: rotate each fan by hand and feel for bearing grit, resistance, or wobble. Power briefly and confirm all four fans reach nameplate RPM on the dashboard. A single bad fan causes one chain to overheat while others stay fine. Replace with a known-good 12 V DC 120 mm axial (Delta or Sanyo Denki industrial preferred over generic replacements).

Pull each hashboard one at a time and run the miner on the remaining two for 10 minutes. If ERR_TEMP_HIGH follows a specific board, you have isolated the fault to that board. If the error only appears with all three boards installed, you are looking at PSU capacity, intake airflow, or ambient, not a single-board issue. Record which chain carries the fault before proceeding.

Under full mining load, multimeter the 12 V rail at the PSU-to-board connector. Expect at least 11.8 V sustained on a healthy APW7, at least 12.0 V on an APW9. Sag below 11.5 V sustained indicates a tired PSU; the DC-DCs compensate by pulling more current, which presents as a thermal fault. Swap PSU with a known-good unit and retest before committing to pad work or chip-level repair.

Verify the heatsink bar on each board is mechanically solid. With gloves on and power off, press the aluminum bar firmly with your thumb at each end and in the middle. A rock-solid bar rules out pad failure. A bar that flexes, tilts, or lifts confirms pad degradation — you need a Tier 3 pad job. Slide a business card between the bar and the chips along the length of the board — if it slips in at any point, confirm separation.

With an IR thermometer or thermal camera, scan the heatsink bar during a 5-minute run. Expect even heat distribution along the bar (plus or minus 5 C end-to-end). A cold spot indicates a dead chip underneath (no power dissipation). A hot spot indicates pad failure or a dying chip at that position. Mark anomalies on a printed board diagram for later reference.

Replace the thermal pads on the suspect board — the signature S17 repair. Unscrew the four heatsink retention screws, apply gentle even upward force (expect resistance from remaining good pads). Scrape old pad material from chips and bar with a plastic spudger, never metal. Clean with isopropyl alcohol 99 percent and a lint-free wipe. Fit Laird Tflex HD80 or T-Global H48-2 W/mK pads, 1.5 mm thickness, cut to size. Re-seat at roughly 0.6 Nm — snug, not aggressive.

Re-apply thermal paste to the backplate if your S17 variant uses backplate-assist cooling (S17+ and some S17e revisions). Arctic MX-6 or Thermal Grizzly Kryonaut are both appropriate choices. Uniform thin layer, never glop — over-application on the large backplate contact area causes mounting pressure issues that make the thermal problem worse, not better. Reassemble with clean fasteners and torque evenly in a diagonal pattern.

Flash DCENT_OS — D-Central's own open-source Antminer firmware, built for exactly this kind of diagnostic work. All the per-chip HW% and per-chip temperature visibility you need, fully open-source, maintained by D-Central for the plebs. Alternatives if you prefer: Vnish or community S17 forks of LuxOS. Stock Bitmain firmware does not expose per-chip temperature on the S17; without per-chip visibility you are fixing this error blind. Run 20 minutes and record the thermal gradient across all 48 chip positions per chain.

If per-chip diagnostics identify a hot chip with consistent HW% errors, reflow it. The BM1397 BGA package tolerates a preheat-plus-hot-air cycle: bottom preheat 150 C, top-side hot air 310-330 C for roughly 30 seconds, controlled cool-down. Lowest-risk chip-level intervention on an S17. If HW% clears and temp normalizes after reflow, re-pad the section and reassemble. Document before and after readings.

Inspect voltage-domain capacitors, especially the bulk electrolytics near the PSU input. S17-era electrolytics are undersized for their thermal environment; bulging or leaking caps are a known failure mode that presents as random thermal shutdowns. Replace any questionable cap with a same-value 105 C low-ESR equivalent before committing to more expensive chip-level work. This is a soldering-iron plus hot-air job, not a reflow job.

Verify the PIC temperature-sensor chip is reporting correctly. If the dashboard shows suspicious temperature values (e.g. -40 C or 255 C on any sensor), you may have a corrupt PIC reading root cause — see the related PIC chip error pages. A bad PIC can cause a false ERR_TEMP_HIGH that no amount of cooling work will fix. PIC replacement and reflash is an advanced bench procedure.

Stop DIY when: pad replacement on both suspect boards fails to drop temps by 10 C after a 20-minute burn-in; per-chip thermal imaging identifies three or more dead chips on one board; you observe visible cap or PMIC damage; or the fault pattern moves between boards suggesting control-board or PSU anomaly. You are now in test-fixture territory. Book a D-Central ASIC Repair slot at https://d-central.tech/services/asic-repair/ .

D-Central bench process on incoming S17: test fixture with programmable load, official Bitmain BM1397 test binaries for per-chip isolation, salvaged-grade or NOS BM1397 chip replacement, full heatsink re-bond with pro-grade pads, cap and PMIC replacement as needed, 24-hour nameplate burn-in with continuous thermal logging. Expect 5-10 business-day turnaround from receipt. Canada-wide shipping, US and international welcomed.

Ship safely: S17 hashboards in anti-static bags, double-boxed with at least 5 cm of foam on every side. Include a note with observed symptoms, firmware version, which chains faulted, timestamps of first and most recent faults, and your contact. Clear notes save diagnostic time, which directly reduces your repair cost. Do not ship the PSU unless it is part of the fault.

Consider end-of-life repurposing for tired S17 units. The S17 was obsolete on the 2024 hashrate curve; a board needing three or more chip replacements is rarely economic at current BTC-difficulty levels. D-Central converts retired S17s into Bitcoin Space Heaters for Canadian winter deployment, where thermal output is the primary revenue. A mechanically sound S17 with fresh pads makes an excellent 2800 W garage heater that pays you in sats.