IceRiver Fan Abnormal on Web Interface: Full Diagnosis

Reach the web UI from your LAN at http://<miner-ip> in any browser. Default login is `admin` / `12345678` - change it the first time you log in. If you cannot reach the miner from your browser, fix that first (network, DHCP lease, or static IP) - trying to diagnose a fan abnormality from a dashboard you can't see is a waste of time. IceRiver also publishes a desktop discovery tool on iceriver.io if scan-by-IP fails for you. Bookmark the dashboard URL once you have it; you'll be on this page weekly if you operate KS-series at scale.

Read the Status page slowly, top-to-bottom, before clicking anything. Both fan RPMs. Both fan PWM duty values. Temp1. Temp2. Per-board hashrate. Any banner, flag, or red highlight. Write all of it down on paper. The web UI is a snapshot - it changes second-to-second - and a written baseline lets you tell whether the next refresh is improving or deteriorating. Skipping this step is the most common reason operators misdiagnose: they read one value, panic, open the chassis, and never establish a true baseline.

Refresh the page every 30 seconds for 3 minutes. Watch each value's trajectory. RPM stable or wobbling? PWM duty steady or ramping? Temps climbing, stable, or settling? A miner mid-failure shows different signatures than one already failed: mid-failure is dynamic (Temp2 climbing 1 C per refresh, PWM ramping 60% -> 80% -> 100%), already-failed is locked (PWM stuck at 100%, temps high but not climbing because firmware throttled the hashboards). Trajectory tells you which side of the cliff you're on.

Power-cycle from the rear rocker, not from the web UI reboot button. A web UI reboot is a soft restart - it doesn't clear the IceRiver controller's latched fan-governor state. Kill power at the rear rocker, wait 60 seconds, restart. Wait for the boot self-test to complete (about 90 seconds on most KS-series), then re-read the web UI Status page. About 1 in 6 web UI fan-abnormal warnings clear here and don't return - they were latched on a transient tach glitch.

Roll firmware back to last-known-good if all sensors look corrupted. If Temp1 reads -10 C or RPM reads 65535 or PWM duty reads 255%, you have a firmware regression, not a hardware fault. Pull the prior image from https://www.iceriver.io/firmware-download/, match the variant exactly to your model (KS3M / KS3L / KS5L / KS5M - wrong variant bricks the controller), flash via the web UI's firmware upgrade page. If the web UI is too unstable to flash through, escalate to SD-card recovery (Tier 3 Step 14).

Confirm the web UI diagnosis at the chassis. Bring the laptop or phone with the web UI open to the miner. Power off, remove the top cover (six Phillips #2 screws standard, two extra on KS5L / KS5M rear bracket). Power back on briefly. Watch the fan the web UI flagged - does it spin? Does it grind, tick, or rattle? Does it fail to start at all? The chassis-side observation must agree with the web UI reading. If web UI says fan dead but you see it spinning, harness or tach issue. If web UI says fan healthy but you see it not spinning, sensor fault.

Blow out both fans with canned air while the chassis is open. Even if web UI reads what looks like a hard fault, dust loading is the #1 cause of premature 12038 bearing failure in home-mining setups. Hold each impeller still with a plastic probe while blasting; canned air backspin induces back-EMF that stresses an already-tired bearing. Blow out the heatsink fins downstream. Blow out the front grille filter if installed. Re-power, refresh the web UI Status page, recheck both RPM columns. Sometimes a 60-second clean takes a flagged fan back into spec.

Reseat both fan connectors at the controller. Power off, wait 60 s for caps to bleed. Locate the two 4-pin headers on the controller - silkscreen typically reads `FAN_F` (front) and `FAN_R` (rear), positions vary across KS3M / KS3L / KS5L / KS5M revisions. Disconnect each, visually inspect for bent pins or crushed latches, reseat firmly until you hear the positive click. Apply a trace of dielectric grease. Re-power, re-read the web UI. Vibration backs the connector out one notch over 6-18 months - and the web UI is sometimes the first place you'll see the symptom.

Multimeter the fan rail under load with the chassis open and the web UI showing live RPM. DC volts, probe V+ to GND on each fan header. Expect 11.8-12.2 V steady on both with the controller commanding >0% PWM. Below 11 V or zero on the suspect side while the web UI shows 0 RPM and PWM >0% confirms a controller-side rail fault, go to Tier 3. Healthy 12 V on the rail with 0 RPM measured = fan is dead, ignore the rest of Tier 2 and order a 12038 replacement (Step 11).

Run the front-rear fan swap test with the web UI open. Four-minute test, decisive. Power off. Unplug both fans. Swap their physical positions on the controller - front fan to rear header, rear fan to front header. Power on, wait for boot, refresh the web UI Status page. If the abnormal reading migrates with the fan (front-fan-now-on-rear-header is the one reading low RPM), the fan is dead and you ordered the right $15 part. If the abnormal reading stays on the same header regardless of which fan is plugged in, the controller-side rail is the problem - Tier 3.

Replace the failed 12038 fan. IceRiver KS-series fans are 120 x 120 x 38 mm, 12 V, 4-pin PWM, dual-ball-bearing axial - DF1203812B2UN (~6,000 RPM, ~140 CFM) is the OEM-equivalent. Do NOT substitute a 120 x 120 x 25 mm consumer PC fan; static pressure is wrong. Swap, reconnect, power on, refresh the web UI. The newly-installed fan should appear in the dashboard at a healthy RPM curve within 60 seconds of power-on. If it does not, recheck connector seating before assuming the new fan is also dead. Verify-flag: confirm exact part number against your specific KS-series revision before ordering.

SMD fuse rework on the fan rail. If Step 9 confirmed +12 V dead on one fan header but the other rail is healthy, an SMD fuse popped - typically a 1206-package fast-blow in the 2-4 A range on the affected rail. Identify by following the trace from the dead fan header back with a continuity probe in fuse-test mode. Hot-air rework station at 290 C, Kapton-tape adjacent plastics, lift the failed fuse, place a matching replacement, reflow. Power on with no fan attached, multimeter the rail, expect 12 V. Then attach the fan, re-read the web UI - RPM should now climb to nominal.

PWM driver MOSFET replacement. Each fan header is fed by a small MOSFET that the controller toggles at 25 kHz for PWM duty. A failed MOSFET reads either always-on rail (web UI shows fan at 100% PWM with steady RPM regardless of commanded duty) or dead rail (web UI shows PWM commanded >0%, fan reads 0 RPM, and Step 9 confirmed dead rail). Hot-air remove, clean pads with solder braid, reflow a matching replacement. Verify-flag: IceRiver does not publish controller schematics - D-Central identifies these MOSFETs by cross-reference against salvaged KS-series control boards. Without a salvaged reference unit, escalate to Tier 4.

Re-flash via SD-card recovery if the web UI is unreachable. If the controller is too far gone to serve the web UI, but power and Ethernet light up normally, SD-card recovery is the last DIY shot. Pull the correct image from https://www.iceriver.io/firmware-download/, write to a microSD with BalenaEtcher, insert into the controller's SD slot, power-cycle, watch for the recovery LED pattern (varies by model - D1 / D2 / D3 / D4 LED sequences are documented in IceRiver's per-model operation manuals). After successful recovery, the web UI should return - log in, re-read Status, confirm sensors are sane. DCENT_OS does not apply here - IceRiver runs 1004LV100 Kaspa kHeavyHash silicon; DCENT_OS is Bitmain-Antminer-only.

Replace thermal pads if Temp2 stays high after the fan is fixed. If you've replaced the fan and the web UI now shows healthy RPM and PWM commanded at 100% but Temp2 is still pinned high, the underlying cooling stack is compromised. Lift the heatsink, clean residue with isopropyl 99% and lint-free wipes, apply a fresh 0.5 mm thermal pad (or Arctic MX-6 paste depending on the original mount style), reassemble. Refresh the web UI - Temp2 should drop 5-15 C within 10 minutes of the rebuild. If it doesn't, you likely have cascaded silicon damage on the 1004LV100 chips - Tier 4.

Stop DIY when any of these are true: web UI reading garbage on all sensors after a clean firmware reflash (sensor subsystem hardware fault), +12 V dead on both fan headers (controller power-tree fault), web UI showing Fan Abnormal paired with chip-count-mismatch errors (cascaded silicon damage), you attempted SMD rework and lifted a pad, two web-UI-flagged fan abnormalities within 30 days even after parts replacement, miner ran with Temp2 above 80 C for hours before catch (1004LV100 silicon damage possible), or web UI completely unreachable after SD-card recovery. That's D-Central repair-bench territory. Book a slot at https://d-central.tech/services/asic-repair/.

What D-Central does at the bench. Diagnosis against a known-good KS-series reference rig with continuous web-UI logging during burn-in. Component-level controller repair including SMD fuse and PWM-MOSFET replacement. Fan harness remake with dielectric-greased connectors. Full hashboard thermal service if overheat events cascaded from the fan fault. 1004LV100 chip-level rework if cooling failure damaged the silicon. Firmware reflash and validation against the IceRiver official archive. Final verification: 24-hour nameplate burn-in (KS3M ~6 TH/s @ ~3,400 W; KS5L ~12 TH/s @ ~3,400 W; KS5M ~15 TH/s @ ~3,400 W) with web-UI fan and temp readings logged every 60 seconds. We ship the unit back with the burn-in log included.

Pack and ship the whole miner. Double-box the chassis. Remove and separately wrap the hashboards in anti-static bags. Include a printed note with the dashboard error codes (if any), web UI screenshots showing the abnormal Status page, the firmware version, observed symptoms, ambient operating conditions, and your contact info. The web UI screenshots in particular save 30+ minutes of bench reproduction time per unit - they're cheap to capture and they directly reduce your repair bill. Canadian customers ship to our Quebec bench; turnaround 5-10 business days. US and international welcomed - the repair desk handles international shipping paperwork.