Whatsminer Error 2310-2350 – Low Hash Rate Warning

Cold-boot at the wall for 60 seconds. Pull the cord, wait for the PSU bulk caps to fully drain, reconnect. Some BTMiner builds hold a ghost `2310` across soft reboots; a full cold boot clears transient events where the hashrate monitor latched a bad sample during a brief network blip or a firmware init hiccup. Log the timestamp, monitor for `24 hours`. If the miner returns to nameplate and stays there, you caught a transient — set up a `GHSavg < nameplate × 0.95` alert and move on.

Measure intake ambient with an IR thermometer at the front grille. Target `≤ 30 °C` under load. A summer garage at `38 °C+` will trigger silent per-chip thermal derating on any M30S/M50/M60 class miner, which drags `GHSavg` below the `2310` floor without firing the `350-352` critical trip. If ambient is hot, fix airflow — open a door, redirect a duct, move the miner — before touching anything else. The miner isn't broken; the room is.

Clean the intake grille and front heatsink face. Shop-vac with a soft-brush attachment. Dust on the intake side drives the boundary-layer air warmer at the chips, which triggers the thermal derate described in step 2 even when the room air is fine. Run the vacuum across the full grille area for `60 seconds`. On M50/M60/M66 class, pop the front grille off if it's tool-less. This resolves a surprising number of `2310` tickets at D-Central's retail repair intake before any chassis work is needed.

Verify physical clearance: `≥ 30 cm` intake, `≥ 15 cm` exhaust. Multi-miner racks with less than `15 cm` between chassis recirculate hot exhaust into adjacent intakes, which triggers the same thermal derate. A single mis-placed miner in a row can drag its own hashrate and its neighbours'. Pull a tape measure, enforce the clearance, re-test. Free fix.

Pull the log via `WhatsminerTool` and confirm no other codes fire alongside `2310`. If you also see `350-352` (hashboard high-temp), `530-532` (chain not found), `550-552` (bad chips), `233-238` (PSU output protection), or `600/610` (ambient high), resolve those first — they are the upstream cause and `2310` is the downstream symptom. Fixing `2310` alone when a sibling code is also firing is chasing a symptom instead of the disease.

Measure the `12V` rail at the PSU output connector under full load. Multimeter on DC, probe the PSU-to-adapter connection while the chassis is hashing at nameplate. Expect `11.8-12.6V` sustained. Below `11.5V` means the PSU is sagging — tired unit, undersized household circuit, low line voltage, or a loose copper-bar connection. Swap in a known-good PSU of matching wattage class. If the rail recovers and `GHSavg` returns to nameplate, the original PSU is the fault. Most common bench finding on M50-class `2310` tickets where ambient is confirmed good.

Measure line voltage at the outlet under load. Expect `235-245V` on `240V` split-phase, `202-212V` on `208V` commercial, `215-225V` on `220V` industrial. Low line voltage forces the PSU to pull more current, which both stresses the PSU and reduces efficiency. The Whatsminer high-performance modes (M30S++ and later) require a strict `230V` minimum to hit nameplate; below that, BTMiner silently degrades to a lower power profile and `2310` fires. If line is low, the fix is electrical, not the miner.

Identify the dragging chain (if single-chain) or confirm system-wide drag. Dashboard → chain detail. Record per-chain TH/s. Healthy Whatsminer: three chains within `5%` of each other. A `10-30%` gap between one chain and the others is the dead-chip / voltage-domain signature. All three equally low is thermal or power-supply. The shape of the shortfall is the diagnostic.

Cold-boot and re-seat the ribbon cables and power harness on the dragging chain. Open the chassis at the PDU (power off at the wall first). Disconnect the ribbon at both ends, inspect for bent pins, oxidation, dust packing. Same discipline: negative first on insertion, signal last; reverse on disconnection. Reverse-order disconnects have damaged per-slot power-sequence MOSFETs on adapters and turned a `2310` into a `530` on the next boot. Careful, always.

Swap the dragging hashboard into a known-good slot. Label slots `0/1/2` with tape, rotate the suspect board into a working slot, rotate a working board into the empty slot. Cold-boot. Re-record per-chain TH/s. If the shortfall follows the board → board-level issue (Tier 3 or Tier 4). If it stays in the original slot → adapter or control-board side (see [Whatsminer Error 530 — Hashboard Not Found](https://d-central.tech/asic-troubleshooting/whatsminer-error-530-hashboard-not-found/) for the adapter-level path).

Reset the firmware profile to stock (no over-tune, no under-tune). `WhatsminerTool` → Remote Ctrl → Power Fast Boot / reset config. Observe `30 minutes` at stock. If `GHSavg` returns to nameplate, your tuning was the cause — rebuild the profile slowly, `50-100 MHz` per step, `10-minute` stability window between steps, stop at the step before `HW error` crosses `2%`. That's this miner's silicon-lottery ceiling; it varies per miner because chip bins vary per miner.

Thermal-image the suspect hashboard under load. FLIR or Seek camera, side-on shot through an open chassis (fans removed temporarily, miner hashing at reduced power). Look for cold spots — chips or voltage-domain regions that should be radiating at `60-75 °C` but read `30-40 °C`. A cold region means that domain isn't drawing current, which means its chips aren't hashing. Map the cold region to the board schematic (MicroBT doesn't publish these officially; Zeus Mining and Thanos Mining both publish repair guides that include the domain layouts on M30/M50/M60 class boards).

Apply fresh thermal paste to the hashboard's chips. Arctic MX-6 or Thermal Grizzly Kryonaut. Uniform thin layer, don't glop. Pay attention to the PCH and voltage-domain ICs — dried pads there are a frequent cause of drift. A board that was marginal before fresh paste often recovers fully after; boards that don't recover are usually into dead-chip territory.

Inspect voltage-domain capacitors and MLCCs under magnification. USB microscope or `30x` loupe. Bulging electrolytics, cracked MLCCs near a buck-boost inductor, dry-solder joints on the domain ICs — all of these cause domain undervoltage, which drops chip count, which drives `GHSavg` below the `2310` floor. This is a soldering-iron + hot-air repair, not a reflow job.

Re-flash stock BTMiner firmware via `WhatsminerTool`. Verify hardware revision against MicroBT's compatibility table before flashing. The wrong firmware for a hydro board on an air chassis bricks the control board. Never flash over wireless; wired only; never while the miner is hashing. If the reflash restores nameplate, firmware corruption was the cause (rare but real).

Run a `get_miner_status` query over the `4028` API for `30 minutes` at 1-minute intervals. Log the `GHSavg`, `GHS5s`, per-chain TH/s, per-chip temps. Analyze the time-series after. If `GHS5s` is consistently healthy but `GHSavg` is stuck low, the miner is derating or the averaging window is catching transient drops — add thermal diagnostic. If both are low and stable, the hashboard has a real capacity loss (dead chips or dead domain) and it's bench-bound.

When to stop DIY. Any of: (a) thermal imaging confirms a cold voltage-domain region that didn't recover with fresh paste and a reseat, (b) visible damage on a hashboard (lifted pads, burnt traces, swollen caps, cracked MLCCs near the PMICs), (c) the dragging chain follows the board through two slot-swaps and firmware reflash doesn't recover it, (d) you've run out of diagnostic tooling (no thermal camera, no hot-air rework, no serial-console access). You're now in component-level bench territory. [Book a D-Central ASIC Repair slot.](https://d-central.tech/services/asic-repair/)

What D-Central does at the bench. Per-chain test on a programmable load with known-good control board and known-good PSU. Per-chip voltage and frequency sweep under a calibrated ADC reference. Chip replacement from matched-bin NOS (new-old-stock) or salvaged-grade inventory for the M30-class and newer BM-class silicon. Voltage-domain rework — FET replacement, cap bank refresh, PMIC swap where indicated. Full BTMiner reflash against the correct hardware-revision build. Then a `24-hour burn-in` at nameplate in a `25 °C` controlled environment. Turnaround is `5-10 business days` Canada-wide; US and international welcome.