NerdOctaxe – Chips 6-7 Not Enumerated on Boot

Cold-cycle three times. Pull both ends of the 12 V (PSU side and XT60 / barrel at the miner). Wait 30 seconds. Power back. Boot. Capture chain count. Repeat twice. If chain count is 8/8 on any of the three boots, your fault is intermittent - not fixed. Note the pattern (always 6/8? sometimes 6/8, sometimes 8/8?) - the pattern is diagnostic data: rock-solid points at silicon or trace; wandering points at cold-solder joint.

Re-seat the cooler over chips 5-8 with fresh thermal paste. Power off, pull the cooler covering the back half of the board, IPA-99% clean every chip top and the heatsink contact face, apply Arctic MX-6 (rice-grain dot per chip, spread by torquing the heatsink down evenly), reassemble to spec torque (four M3 screws, finger-tight plus a quarter turn). Cold-cycle and re-boot. About 25% of `chip 6/7 missing` units recover here because dry paste was lifting the cooler and BGA joints stress-fractured under thermal cycling.

Inspect the case and PCB for shipping damage. Drops bend boards. Bent boards crack BGA joints - and the back end of the board (chips 6/7 region) is the most likely place to crack because it is furthest from the chassis mounting points and acts as a lever arm. Look at the corners of the board, the screw mount points, and the chassis itself. If you see deflection, take photos before disassembling further - you may be filing a shipping insurance claim.

Confirm firmware is current and matches BM1368. In the NerdOS web UI System Info panel, verify the firmware build targets BM1368 (not BM1366, not BM1370). Cross-flashed firmware can produce false chain-break readings because the protocol framing differs between chip generations. Reflash the correct NerdOS build for NerdOctaxe / BM1368 if there is any doubt.

Update to the latest stable NerdOS release. Older builds had less detailed chain-init logging and occasionally false-positive chain-break detection at the back of the chain. Pull the current release from the BitMaker-hub repository, flash via OTA or the web flasher, factory-reset NVS, reconfigure WiFi and pool, retest. Pin the version that works for your board revision.

Capture the full boot log to a text file. USB-C, 115200 8N1, terminal of your choice (screen, PuTTY, minicom). Cold-cycle, capture the first 30 seconds. Look for the chain enumeration line (`BM1368 x N detected`), per-chip discovery messages, any `timeout` / `no response` entries, and which chip number first failed. Save the log - if you escalate to D-Central or to the open-source maintainers, this is the first artifact anyone wants.

IR-thermometer-walk the chip positions during a hash attempt. Power up, let the firmware try to hash for 60 seconds, point an IR thermometer at each BM1368 in sequence (through the heatsink fins works at lower resolution), log the temp. Healthy chips read 45-65 degC at 60s of attempted load; dead/unpowered chips read ambient (20-30 degC). The cold chips are your suspects. Cross-check against the boot log - chips 6 and 7 both cold is the canonical pattern.

Multimeter the local VCORE at chips 5, 6, 7. Power up, probe each chip's bypass cap pad. Expected: 1.05-1.15 V DC across all chips. A chip reading 0 V or below 0.9 V while neighbours read clean has a local power-rail fault - flag for Tier 3/4. Both chips 6 and 7 reading nominal VCORE while still silent points at silicon or joint failure at chip 6.

Re-seat the suspect chip's decoupling caps. Visible damage to caps near chip 6 is fixable with a hot-air station and replacement components - Murata or TDK MLCCs in the package size shown on the schematic. If you do not have the rework gear this becomes Tier 4. Do not try this with a soldering iron only on 0402 / 0201 packages; you will damage adjacent components.

Test on a different 12 V PSU (20 A+ rated). Borderline-spec PSUs can produce intermittent chain-break behaviour at the back of the chain. Swap to a known-good 12 V / 20 A+ / 240 W+ brick (HP DPS-1200FB or Dell 1100W server PSU with XT60 breakout) for 30 minutes of boot cycles. Log chain-init outcome each boot. If the chain comes up 8/8 clean on the swap PSU, your original PSU was sagging and breaking enumeration timing at the chips electrically furthest from the regulator.

Reflow chip 6 with preheat plus hot air. Disassemble. Mask adjacent chips with kapton tape. Flux chip 6's BGA from the side (no-clean flux wicks under the package via capillary action). Preheat the bottom of the board to ~150 degC to reduce thermal shock. Top-side hot air at 310-330 degC, slow circular motion, ~30 seconds total dwell time. Watch for the package to settle - you will see a tiny visible drop as solder balls reflow. Cool naturally on the preheat for ~3 minutes. Re-paste, reassemble, cold-cycle. About 35-40% of cracked-joint chain breaks at chip 6 recover here.

If reflow recovers the chain, log a 24-hour burn-in. A reflow that holds for 5 minutes and fails again is a chip on the way out. A reflow that survives 24 hours of stock-clock hashing has a real shot at sticking around for 6-18 months. Run a serial-log capture overnight, watch for chain re-breaks. If it survives, you have bought yourself meaningful service life for the price of a flux pen.

Scope the chain bus during enumeration. Trigger on the rising edge of the discovery command at chip 0's output. Step the scope probe through the chain - chip 0 output, chip 1 input, chip 1 output, and so on through chip 5 output and chip 6 input. The first point at which the signal disappears or degrades is the break. A weak edge or excessive ringing at chip 6's input suggests trace impedance issues or a partially-failing input stage on chip 6 - both flag-able for Tier 4.

Replace chip 6 with a graded BM1368. D-Central stocks salvaged-grade and new-old-stock BM1368 chips for exactly this repair; pricing per chip is ~$30 CAD plus rework labour. Chip removal: hot air at 330 degC, lift package with vacuum tool, IPA-clean residual solder. Chip placement: solder paste stencil, careful alignment to PCB pads, reflow with preheat plus hot air at ~330 degC. Bench-grade rework - if you are new to BGA work, practice on a Bitaxe Hex first.

Repair the chip-5-to-chip-6 trace if it is broken. With board out and powered down, confirm with continuity meter (Step 7 of the diagnostic tree). The fix is 30 AWG enamel wire from chip 5's output pad to chip 6's input pad, routed cleanly under the cooler clearance. Finicky but cheaper than replacing the board. On early NerdOctaxe revisions this is the canonical fix - the factory trace was marginal and a clean jumper outlasts the original.

When to stop DIY: you have done the cold-cycle, re-pasted the back-half cooler, scoped the bus, attempted a reflow on chip 6, and the chain still misses. OR you do not have hot-air rework gear and the issue is past Tier 2. OR you see PCB-level damage - burnt traces, lifted pads, missing or scorched caps. Stop. The NerdOctaxe is $500-$900 CAD hardware; sustained DIY beyond reflow can cost more in solder and ruined components than the bench fee. Book a D-Central NerdOctaxe repair slot.

D-Central bench process: test fixture with controlled 12 V rail and isolated VCORE monitoring per chip position. Per-chip enumeration verification with the open-source NerdOS test harness. Cracked-joint reflow with profiled preheat and hot-air at chip 6 or 7. Chip replacement from D-Central's graded BM1368 inventory at ~$30 CAD per chip plus labour. Trace repair with 30 AWG jumpers when needed. Post-repair 24-hour burn-in at stock clock. Typical turnaround: 5-10 business days Canada-wide.

Ship safely. Anti-static bag around the PCB. Bubble wrap around the bag. Rigid outer box with at least 5 cm of foam on every face. Include a one-page note: which chip count the boot log reports (6/8, 5/8, etc.), firmware version string, PSU spec, history of OC sessions / drops / liquid events, your contact info. Photograph the boot-log capture and include it on a USB stick or as a printed page. The clearer your write-up, the cheaper your repair - diagnostic time is the largest variable on small-board repairs.