IceRiver Network Port No Lights (Yellow Steady, Green Flashing)

Substitute the Ethernet cable with a known-good Cat-5e or Cat-6 patch. Pull a working cable from another running device - laptop, another miner, your TV - and plug it into the IceRiver RJ-45 with the other end in the same switch port. Wait 5-10 seconds for link negotiation. If the yellow LED comes alive, your original cable was the failure. Throw it out - never keep maybe-bad cables on a mining bench. This single substitution clears 35-45% of dead-link tickets across all ASIC manufacturers.

Substitute the switch port. Move the cable's far-end to a different physical port on the same switch. If still dark, move it to a completely different switch entirely - an unmanaged 5-port Netgear / TP-Link / Trendnet off the shelf is the simplest diagnostic switch you can have on a bench, $20 retail. Skip every managed-switch configuration concern; you want the simplest physical-layer test possible. If LEDs come alive on the substitute switch, the previous switch port was dead, admin-shutdown, on the wrong VLAN, or speed-locked at a setting the IceRiver PHY can't negotiate.

Re-seat the cable on both ends. Pull the cable from both the IceRiver and the switch. Inspect both connector locking tabs - if a tab is broken, the connector is loose in the jack and the cable is effectively unplugged. Re-insert firmly; listen for the click. Wiggle test: the cable should not pull out under light tension. A broken locking tab is a $5 cable replacement, not a controller-side fault.

Hard power-cycle the miner. Rear rocker switch off, wait 30 seconds for caps to bleed and the network stack to fully unwind, power back on. The IceRiver's network stack initializes during boot; a soft restart sometimes leaves the PHY in a hung state. A full power-cycle rules out a transient firmware-side issue before you start blaming hardware. Wait a full 90 seconds after power-on before declaring the link still dead - the network stack initializes late in the boot sequence.

Confirm the far-end is alive. Look at the switch / router port LEDs for the cable's far end. If those LEDs are also dark on a port that should be active for any device plugged into it, the switch port (not the IceRiver) is the dead component. Test by plugging your laptop into the same switch port with the same cable - if the laptop also fails to link, the switch port is the failed component, not the miner.

Continuity-test the cable with a multimeter. Set the multimeter to continuity / beep mode. Probe pin-to-pin between the two RJ-45 ends of the suspect cable: pin 1 to pin 1, pin 2 to pin 2, ... pin 8 to pin 8. All eight pairs should beep with zero resistance. Then probe diagonally - pin 1 to pin 8, pin 2 to pin 5 - none of those should beep (no internal shorts). One missing pair = bad cable, replace. A dedicated cable certifier (`Klein VDV526-100` or equivalent) does this test in 5 seconds and reports which pair failed.

Visually inspect the IceRiver RJ-45 jack. Power off the miner at the rear rocker, wait 60 seconds for caps to bleed. Shine a phone flashlight at a slight angle into the jack. All 8 gold-plated pins should be visible, parallel, at the same height, shiny gold. Bent pin = try to recover with a thin dental pick or wooden toothpick - gentle, slow, no force. Missing pin / blackened pin / cracked plastic shroud / loose magnetics rattling inside = magnetics-package failure (Tier 3).

Force `100 Mbps full-duplex` on a managed switch port. If the miner is plugged into a UniFi / MikroTik / Cisco / Aruba managed switch, log into the switch admin and hard-set the IceRiver port to `100 Mbps Full Duplex`. IceRiver controllers ship `10/100` PHYs that do not negotiate gigabit cleanly with all switches. UniFi: Devices -> Switch -> Port -> Link Speed = 100 Mbps Full Duplex. MikroTik: `/interface ethernet set [find] auto-negotiation=no speed=100Mbps full-duplex=yes`. Re-test.

Disable PoE on the switch port. If the switch is PoE-capable, confirm PoE is administratively disabled on the IceRiver port. PoE-on a non-PoE device is usually safe (modern switches negotiate before injecting voltage), but a misconfigured forced-PoE port can damage non-PoE PHYs over time. UniFi: Port settings -> PoE Mode = Off. Cisco: `power inline never` on the interface. Re-test the link with PoE confirmed off.

Test for shorts on the controller's `3.3V` rail. Power off the miner, set the multimeter to resistance mode, probe the `3.3V` rail to ground at any decoupling cap on the controller PCB near the PHY chip. Resistance should be in the kilohm range (>1 kohm typical). A reading of 0-50 ohms = a shorted component on the rail, often a failed cap or a shorted PHY. This is a Tier 3 / Tier 4 escalation point - do NOT power on a board with a shorted rail. Verify-flag: exact PCB locations of decoupling caps vary by IceRiver controller revision; identify the rail by tracing back from the PHY's power pins.

Reflow the RJ-45 magnetics package. Hot-air rework station, preheat the controller bottom-side to ~150 C, top-side hot air at 280-310 C aimed at the magnetics' through-hole legs for 30-45 seconds. Goal: re-melt cracked solder joints without damaging adjacent components. Let it cool naturally - do not move the board for 60 seconds after removing heat. This recovers magnetics whose solder joints have cracked from mechanical stress (the most common board-level Ethernet failure mode) without needing to source a replacement part. Re-test link with a known-good cable.

Replace the RJ-45 magnetics package. If reflow fails or the magnetics are visibly cracked, source a replacement integrated jack with magnetics - common parts: `Pulse Electronics J0011D21BNL`, `Bel Stewart SI-60035-F` series, or equivalent `10/100 BASE-TX` integrated MagJack. Verify-flag: exact part number depends on the IceRiver controller revision; cross-reference markings on the existing magnetics package against a Pulse / Bel datasheet before ordering. Hot-air desolder, clean pads with solder wick + flux, reflow new part. Verify under bench magnification (4x minimum) before re-powering.

Replace the Ethernet PHY chip. If magnetics are confirmed clean and the PHY is suspected (rail sag, hotspot, electrical short), the PHY transceiver IC needs replacement. IceRiver controllers historically use `Realtek RTL8201` / `Microchip KSZ8051` / `Texas Instruments DP83848` family `100 Mbps` PHYs depending on revision - verify-flag against the actual chip markings on your specific board. Source a matched replacement, hot-air desolder, clean pads, place new chip with paste + wick + fresh flux, reflow. Fine-pitch QFN-32 or QFP-48 rework, benefits from a PCB preheat plate.

Inspect and replace electrolytic / MLCC caps near the PHY and magnetics. Bulged electrolytics, cracked MLCCs, or visibly-discolored caps within 3 cm of the PHY / magnetics need replacement before re-powering. Use 105 C-rated electrolytic capacitors with the correct voltage rating; X7R MLCCs in matching values. A bench DMM with low-ohms / capacitance mode helps verify replacement parts before installation. This step is often combined with magnetics or PHY replacement - if the failure was caused by an upstream cap, replacing only the downstream component will lead to repeat failure.

Re-flash IceRiver firmware after a controller-side hardware repair. Once magnetics / PHY repair is complete and the rails are confirmed clean, fresh-flash the controller firmware via SD card recovery from `iceriver.io/firmware-download/` to ensure the network-stack init is starting from a known-good image. Wrong-model image bricks the controller - confirm model SKU exactly (KS0 / KS0 Pro / KS0 Ultra / KS1 / KS2 / KS3 / KS3L / KS3M / KS5 / KS5L / KS5M). Note: DCENT_OS is NOT applicable here - IceRiver runs `1004LV100`-class kHeavyHash silicon, completely different from Bitmain hardware. Stick with native IceRiver firmware.

Stop DIY and ship to D-Central. Book a repair slot at https://d-central.tech/services/asic-repair/ when any of these are true: both LEDs remain dark after substitute cable + substitute switch port + power cycle, visible damage to the RJ-45 jack (cracked shroud, missing pins, blackening, loose magnetics), multimeter detects a short on the `3.3V` rail, reflow / replacement of magnetics was attempted and link still doesn't come up, you don't own a hot-air rework station, or two or more KS-series units in the same shipment exhibit identical dead-link symptoms (batch defect / shared shipping-impact event).

What D-Central does at the bench for IceRiver Ethernet failures. Controller diagnosis against a known-good KS-series reference rig with an oscilloscope on the PHY's TX/RX pairs to confirm signal integrity, programmable bench load on controller power rails to rule out brownout-induced PHY hangs, hot-air rework of the RJ-45 magnetics package and PHY chip as needed with bench-grade replacements, full DHCP / static-IP regression test on the bench LAN, and 24-hour reachability burn-in (continuous ping, hashrate verification at the SKU's nameplate spec) before we ship the unit back. Verify-flag: exact bench process is repair-ticket-driven and may vary by failure mode and SKU.

Ship the miner safely. Pack the IceRiver in original foam if you have it, or double-box with at least 5 cm of foam on every side - magnetics packages are vulnerable to shipping impact, so do not under-pack. Wrap the controller separately in anti-static if you've already removed it from the chassis. Include a printed note with: model SKU, serial, observed symptoms (both LEDs dark / Detect IP empty / which Tier 1-3 steps you tried), firmware version if known, and your contact info. The note saves us 30-60 minutes of bench diagnostic time, which directly saves you repair dollars. Canada-wide / US / international welcomed.