Antminer Z15 – Temperature Too High

Hard power-cycle at the breaker for 60 seconds. Not a soft restart — physically cut mains. Clears wedged driver state from recent firmware updates and forces a clean thermal-calibration pass on boot. Roughly 5% of `ERR_TEMP_HIGH` tickets clear at this step alone because the previous boot's temp loop never rezeroed cleanly.

Let the miner cool for 15 minutes before the next boot. With mains cut, chassis internal temperature drops from `80 °C+` to ambient at the rate the heatsinks dissipate into still air — about 15 minutes for a Z15. Don't rush the boot after a trip; the next boot will just trip again if you skip the cool-down.

Shop-vac the intake filter and front grille. Remove the filter if fitted; shop-vac both sides. Wipe the front grille with a dry microfibre. Verify the first `15 cm` in front of the intake is clear — no curtains, no cardboard, no other miner venting directly in. This single step resolves 40–50% of Z15 over-temp events in D-Central's intake, more than any chip-level work.

Verify ambient temperature with an IR thermometer at the intake grille. Target: ≤ `30 °C`. Not room-middle, not the hallway, not the thermostat reading from three metres away. Z15 thermal headroom is narrower than newer S19-class miners; `32 °C+` ambient pushes it outside its design envelope and no amount of paste refresh rescues that.

Check the fan RPMs in the web UI and listen at the chassis. All four fans should read `5,500–6,400 RPM` at full load, with no red-flag icons. Listen for chirping, clicking, or grinding — those are end-of-life bearing signatures even when the RPM looks nominal. Note the RPM readings for the facts file before moving on.

Deep chassis blow-out with compressed air. Remove the top cover. Blow every heatsink channel front-to-back with short pulses of `< 100 PSI` compressed air while holding fans stationary with a plastic stylus — don't let the fans spin under the air jet, or you'll over-speed the bearings and accelerate failure. Pay attention to the back corners of each heatsink where dust accumulates into felt-like mats. The chassis should show no visible dust when you're done.

Replace any failed or chirping fans. Spec: `12 V DC, PWM, 4-pin, 120 mm × 38 mm`, server-grade. Delta QFR1212GHE, Sanyo Denki 9GV1212P1J03, or equivalent high-static-pressure units. Don't substitute a quiet consumer fan — static pressure matters more than airflow number on a Z15 heatsink. Match all four if any one is replaced; mismatched fans create pressure imbalances and accelerate wear on the remaining originals.

Multimeter check on the PSU output under load. Probe the `6-pin PCIe` connectors while the miner is hashing at full load. Expect `11.8–12.4 V` sustained. Below `11.6 V` = PSU tired or circuit undersized; swap PSU with a known-good APW9 or APW12. Sagging rail forces chips to draw more current to hit target hashrate, which directly pushes heat up.

Measure line voltage at the panel under load. `235–245 V` on 240 V split-phase, `202–212 V` on 208 V commercial. Low line voltage = PSU sag = higher chip current = more heat dissipation = over-temp. If line voltage is low at peak hours (evening A/C peak in residential neighbourhoods is the classic signature), the fix is upstream — electrician or load-shifting, not miner-side.

Reposition the miner for better airflow. `30 cm+` clearance at the intake, `30 cm+` at the exhaust. No direct line-of-sight into another miner's intake or exhaust. If you're stacking Z15s on a rack, stagger them so nobody ingests another's exhaust. Heat-rising convection fights you here — mount stacks with the exhaust pointing into an open aisle, not against a wall.

Flash DCENT_OS — D-Central's open-source Antminer firmware. Landing: `https://d-central.tech/dcent-os/`. Source: `https://github.com/DCentralTech/DCENT_OS`. Gives per-chip temperature telemetry, true per-chip HW%, autotune, and stratum v2 — the Mining Hacker option, maintained in public, no licensing BS. Alternatives if you prefer: Braiins OS+, LuxOS, Vnish (verify Z15 platform support is current on the version you flash — the Z15 is EOL and third-party Z15 support has been maintenance-mode for years). Run 20 minutes to stabilise, then record per-chip temps across all three chains and note outliers.

Thermal paste refresh on all `BM1580` chips. Power off at the breaker, open the chassis, remove the suspect hashboard. Unbolt the heatsink from the board. Clean off old paste with isopropyl alcohol 99% and lint-free wipes — no cotton swabs, no paper towels. Apply a uniform thin layer of Arctic MX-6 or Thermal Grizzly Kryonaut to each `BM1580` package top (pea-sized blob centred, let heatsink pressure spread it). Reassemble, torque heatsink bolts in a cross-pattern to manufacturer spec. One hashboard at a time — don't mix board positions.

Thermal-pad replacement on PCH and voltage-domain ICs. The thermal pads on the non-`BM1580` components crumble worse than the chip-top paste. Replace with fresh silicone-based pads, matched thickness (measure the old pad with callipers before removal, match within `0.2 mm`). A too-thin pad leaves an air gap; a too-thick pad prevents heatsink seating. This is where most first-time Z15 paste jobs go wrong.

Heatsink cleaning. Once the heatsink is off the board, shop-vac the fins, then flush with isopropyl alcohol and let fully air-dry before reassembly. Residual dust inside heatsink fins reduces effective surface area by 20–30% and undoes half the benefit of the paste refresh. On 4-year-old Z15 heatsinks this step is non-optional.

Roll firmware to the last-known-good build for your specific hardware revision before accepting an over-temp diagnosis as silicon-level. Check your hardware revision label against Bitmain's table; the wrong firmware for a late-rev Z15 control board can mis-read the `PCB` sensor offsets and phantom-trip over-temp at normal operating temps. This is a known issue community-side with specific stock-firmware builds.

When to stop DIY. The same chain trips over-temp after a verified paste refresh + fan + airflow pass. The board's `PCB` sensor reads wildly different from the IR reading of the board surface (sensor drift). Visible component damage — bulged caps, discolored PMIC, burnt-smell — anywhere on the suspect board. A heatsink boss is stripped so you can't torque the heatsink flat. At any of those signals, stop bench work and ship to D-Central. `https://d-central.tech/services/asic-repair/` — 5–10 business-day turnaround, Canada/US/international accepted. The Z15 is EOL; Bitmain's RMA channel is effectively closed for this generation, so D-Central is your realistic Tier 4 path in 2026.

What D-Central does on the bench for a Z15 over-temp case. Full paste + pad refresh with measured pads, heatsink flatness check on a granite surface plate, sensor calibration cross-check against thermocouple reference, input-stage MOSFET and PMIC diode-test, per-chip isolation using community tools specific to `BM1580`, chip replacement with graded salvage silicon where needed, full board wash + reflow + reseal, then 24-hour burn-in at nameplate before the board ships back. Full chain-of-custody documentation so you know exactly what was done and which parts were swapped.

Ship the hashboards safely. Individual anti-static bags, double-boxed with at least `5 cm` of foam on every side. Include a note in the outer box with observed symptoms, firmware version string, exact kern.log trip lines, intake and heatsink IR readings, and your contact. That pre-packed diagnostic context saves the bench 15–30 minutes of intake time, which saves you repair cost. Insure the shipment for board replacement value (not repair value) in case the carrier eats it in transit.