NerdOctaxe – 200W+ PSU Undersized Shutdown

Retire any `12V` PSU rated under `240W` from NerdOctaxe service. This single fix clears 80% of reports. `12V` / `25A` (`300W`) is the recommended sizing; `12V` / `29A` / `350W` (Mean Well LRS-350-12) is the 'buy-once' pick for stock plus modest tuning; `12V` / `50A` / `600W` (Mean Well LRS-600-12) covers heavy overclocks or running two Octaxes off one supply. Do NOT 'get by' with a `200W` brick because the miner happens to light up - eight BM1370 dies starved of current don't crash cleanly, they cook slowly.

Do not reuse your NerdQAxe or NerdQAxe++ PSU. A `12V` / `10-15A` / `XT30` supply was correctly sized for a four-chip board pulling `130W`. It is drastically undersized for an eight-chip board pulling `208W`. If you previously ran a NerdQAxe, that PSU goes back in the drawer or powers another NerdQAxe - and you buy a new properly-sized supply for the Octaxe. Label your PSUs if you own both miners - this is the #1 most-common rookie mistake on the NerdOctaxe.

Use a PSU with an `XT60` output from the factory. Do NOT adapt `XT60` to `XT30` under any circumstances - the `XT30` housing cannot handle `18-20A` continuous without melting, and D-Central's setup guide calls this a fire hazard. If you must build your own pigtail, use genuine Amass `XT60` connectors and `14 AWG` silicone wire minimum (`12 AWG` if the run is longer than `2.5 ft`). Bitronics and D-Central both ship the NerdOctaxe with correctly-made `XT60` pigtails - start from theirs as reference.

Route the power cable away from the miner's air intake. PSU heat plus eight `BM1370` dies compounding in a closed cabinet is how PSUs sag faster. Keep the supply on a hard surface with clearance above and below; a small desk fan pointed at the brick drops steady-state temp `5-10 C`. PSUs live longer in airflow.

Leave a Kill-A-Watt inline permanently. `$25 CAD` of continuous wall-draw visibility is worth more than every other monitoring tool combined. Stock NerdOctaxe = `~208W` at the wall. A creeping wall-draw number over weeks is the earliest possible warning that your PSU is dying, the room temperature is drifting, or a chip has started to misbehave - days before NerdOS surfaces anything.

Measure the `12V` rail at the `XT60` under load. DC multimeter, probe on board-side `XT60` pins while the NerdOctaxe hashes at stock for 10+ minutes. Expected: `11.8-12.2V` sustained. Anything below `11.5V` = upstream sag. Compare to PSU-side measurement: if PSU is clean but board is sagging, connector/wire is resistive. If PSU itself is sagging, supply is undersized or failing. Write down both numbers before changing anything.

Replace the `XT60` pigtail with a genuine Amass part. Genuine Amass `XT60` has three metallic copper contacts visible down the barrel, yellow housing with the Amass logo, mechanical polarity keying. Counterfeits have muddy plating, loose shells, contacts that wiggle. `14 AWG` silicone wire minimum for a NerdOctaxe pigtail, `12 AWG` over `2.5 ft` or tuned past `10 TH/s`. Solder with flux and proper heat; cold joints cause most 'mystery XT60 heat'. Heat-shrink over solder, outer heat-shrink for strain relief.

Derate your PSU for continuous mining. A supply rated `85%` efficient at `50%` load is typically `78-80%` efficient at `95%` load. Running any PSU above `90%` of nameplate 24/7 = heat + reduced efficiency + accelerated cap aging. Size for `60-70%` of nameplate at miner full-load. That's why a NerdOctaxe at `208W` real draw wants a `300-350W` supply, not a `240W` one, for multi-year service.

Update NerdOS to the latest stable release. NerdOS is an ESP-Miner / AxeOS fork adapted for multi-chip boards. Current stable ships brownout-detection hardening, `TPS546` telemetry improvements, and per-chip `VCORE` dashboard fields. Flash via the NerdOctaxe web flasher or D-Central firmware update guide. Do not run pre-release / nightly builds on a production NerdOctaxe while chasing this bug - you won't know if the fix is firmware or PSU.

Clean the `XT60` contacts. Isopropyl `99%` and a small nylon brush. Repeated heating events oxidize contacts and bury clean metal under a thin insulating layer. Clean male and female halves, let dry fully, reconnect with a firm click. Combined with a re-crimp this clears ~70% of resistive-connector reports that haven't already gone past Tier 2.

Scope the `12V` rail at the `XT60` under load. DC-coupled oscilloscope, `20 MHz` bandwidth limit, probe on board-side pin with ground on chassis. Expected: `11.8-12.2V` DC with under `100 mV` peak-to-peak ripple. Ripple above `300 mV` = output caps dying, undersized, or supply is an unregulated unit masquerading as regulated. `TPS546` VRMs cover for this up to a point but sustained noise degrades `BM1370` silicon over months. Replace the PSU.

Scope `VCORE` at one of the `TPS546` output caps. DC-coupled, short ground lead, probe on a ceramic cap near a `BM1370` die. Expected: `1.15-1.20V` DC with under `20 mV` ripple during stock hashing. Large `VCORE` excursions tracking `12V` ripple = `TPS546` losing regulation because input rail is too noisy. Real fix is upstream - size and clean up the PSU.

Inspect the on-board `XT60` socket. Look for discolouration, lifted solder joints on through-hole pads, or wobble in the socket body. If the socket ran hot (melted housing, board near-burn, discoloured solder joints), the copper pour underneath may be damaged even if surface looks OK. Reflow the joints or replace the socket. `12 AWG` tinned wire into proper through-hole, `60/40` or equivalent lead-free solder with a station that delivers enough heat.

Check each `TPS546` regulator for heat damage. Run the miner 15 minutes at stock, then IR-scan each of the eight `TPS546` positions. Healthy: `55-70 C` under load, roughly even across all eight. Suspect: above `85 C` or wildly uneven. A cooked `TPS546` will fault `Device ID mismatch`, fail init handshake, or silently run out of regulation - replace the part (`TPS546D24A` / `D24S` from Digi-Key or Mouser) with hot-air rework; not a soldering-iron job.

If a `BM1370` is suspected damaged from sustained undervoltage stress, reflow first. Profile same as Bitaxe Gamma / Antminer S21 Pro family: `150 C` bottom-side preheat, `310-330 C` top-side hot air for ~`30 s`, let cool naturally, re-apply thermal paste, re-seat Thermalright AXP90-X53 cooler. Reflow before replacement because `BM1370` dies are not cheap and reflow recovers a surprising number of 'dead' chips whose solder joints were merely stressed by thermal cycling. If reflow fails, replace with a graded die.

Stop DIY when: the on-board `XT60` socket pads are lifted or burnt, the `12V` bus copper is discoloured or damaged, a `TPS546` has visibly failed with collateral board damage, a `BM1370` is non-responsive after reflow, you've replaced the PSU twice and the `XT60` pigtail once and the miner still sags, or two or more chips are dark in positions that don't respond to heatsink re-seating. You're in bench territory. Book D-Central NerdOctaxe repair.

D-Central bench process: full inspection under magnification, power-path probe with scope (wall -> PSU -> `XT60` -> `12V` bus -> each of 8 `TPS546`s -> each of 8 `BM1370`s), `TPS546` replacement where needed, `BM1370` replacement from graded stock if reflow fails, `XT60` socket reflow or replacement, Thermalright AXP90-X53 cooler re-seat with fresh paste and pads, post-repair 24-hour burn-in at stock `9.6-9.8 TH/s`. We track NerdOctaxe failure patterns across inventory so every repair feeds back into our recommended PSU and `XT60` BOM.

Ship safely. Anti-static bag for the board, bubble wrap, rigid outer box with at least `5 cm` of foam on every side. Include a one-page note: PSU brand + rating, how long in service, symptoms observed, stock vs tuned / ECO, whether `XT60` was stock / re-crimped / replaced, ambient room temp range. That note saves us diagnostic time and you money. Canada-wide and worldwide shipping welcomed - the NerdOctaxe community is small and global, and we've bench-repaired units from four continents.