Switch Port CRC Errors & Packet Drops: Layer 1 & 2 Diagnostic Guide
Switch port CRC (Cyclic Redundancy Check) and FCS (Frame Check Sequence) errors indicate physical Layer 1 signal degradation before packets ever reach Layer 3 routing. They are caused by damaged Cat6 terminations, electrical noise coupling from adjacent power lines, bent RJ45 pins, or dirty SFP optical endfaces. Discarded corrupted frames trigger TCP retransmissions, slashing effective throughput from 1,000 Mbps down to single digits.
- FCS / CRC Errors are Layer 1 Physical Faults: The 4-byte CRC-32 checksum calculated by the switch ASIC does not match the frame's header, indicating bit flipping caused by bad patch cables, punchdowns, or electrical noise.
- Runts (<64 Bytes): Frames smaller than the 64-byte IEEE 802.3 minimum that usually stem from early collisions in half-duplex links or faulty network interface cards (NICs).
- Giants (>1518/1522 Bytes): Frames exceeding maximum transmission unit boundaries, almost always caused by an MTU mismatch where an endpoint sends 9,000-byte Jumbo Frames while the switch port is locked at standard 1500-byte MTU.
- DOM/DDM for SFP Optics: Optical CRC errors on fiber links indicate dirty LC connectors or weak optical Rx power falling below the transceiver's sensitivity floor (e.g., < -14 dBm on 10G-LR).
Anatomy of Layer 1 & 2 Frame Corruption: CRC vs FCS
Every standard IEEE 802.3 Ethernet frame concludes with a 4-byte Frame Check Sequence (FCS) calculated using the CRC-32 (Cyclic Redundancy Check) polynomial algorithm across all header and payload bytes. When a frame arrives at an ingress switch port:
- The switch hardware ASIC computes a fresh CRC-32 checksum against the arriving bits.
- The switch compares its computed value against the 4-byte FCS appended to the tail of the frame.
- If the values match, the frame is forwarded. If even a single bit flipped during transit across the copper wire or optical fiber, the checksum fails, the switch increments its CRC / FCS Error counter, and the entire frame is discarded silently.
Because Ethernet does not retransmit lost frames at Layer 2, upper-layer transport protocols (TCP) must wait for retransmission timeouts (RTO), triggering catastrophic throughput drops, choppy video conferences, and disconnected SSH sessions.
Runts, Giants, and Late Collisions Explained
Beyond standard CRC errors, managed switch counters track several distinct frame anomalies:
- Runts (Undersized Frames < 64 Bytes): The IEEE 802.3 standard mandates that all Ethernet frames must be at least 64 bytes long (including headers and FCS) to ensure collision detection across maximum cable lengths. Frames shorter than 64 bytes with an invalid FCS are classified as Runts. They are typically produced when a collision aborts transmission prematurely or when a damaged NIC transceiver malfunctions.
- Giants (Oversized Frames > 1518/1522 Bytes): Frames exceeding standard maximum Ethernet transmission limits (1518 bytes for untagged frames, 1522 bytes for 802.1Q tagged frames). Giants almost universally stem from an MTU mismatch: a storage server (such as TrueNAS or Proxmox) has its NIC configured for 9000-byte Jumbo Frames, while the intermediate switch port is configured for standard 1500-byte MTU.
- Late Collisions: In full-duplex Gigabit and 10-Gigabit Ethernet, collisions should physically never occur because transmit and receive pairs are completely isolated. If you see late collisions accumulating, an auto-negotiation duplex mismatch has occurred—one side defaulted to Half-Duplex while the other operates at Full-Duplex.
Reading Switch Port Counters via CLI (Cisco, Aruba, MikroTik & UniFi)
To pinpoint the exact physical port dropping frames, inspect hardware interface counters using vendor CLI commands:
Cisco IOS-XE / Catalyst:
Switch# show interfaces GigabitEthernet1/0/1
# Look for:
# 1245 input errors, 1240 CRC, 0 frame, 5 overrun, 0 ignored
# 0 late collision, 0 deferred, 0 lost carrier
MikroTik RouterOS v7:
/interface ethernet monitor [find name=ether2] once
# Or check hardware packet drops:
/interface print stats-detail where name="ether2"
ArubaOS-CX:
switch# show interface 1/1/1 error-statistics
# Displays FCS Rx Errors, Align Errors, Runts, and Rx Drops
Ubiquiti UniFi Switch (via SSH):
# SSH into UniFi Switch:
swctrl port dump info 1
# Displays port link status, RX/TX bytes, RX packet errors, and FCS drops
SFP/SFP+ Optical Telemetry & DOM/DDM Diagnostics
When CRC errors occur across fiber optic uplinks (10GBASE-SR, 10GBASE-LR, or 25G SFP28), the root cause is almost never electromagnetic interference; it is optical power loss or contaminated fiber endfaces.
Digital Optical Monitoring (DOM / DDM) allows the switch to query the transceiver's onboard micro-sensors in real time:
Switch# show interfaces transceiver detail
# Optical Rx Power: -16.2 dBm [Warning Low: -14.4 dBm] <-- LOW RX SENSITIVITY FAULT!
# Optical Tx Power: -2.1 dBm [Normal Range: -8.2 to 0.5 dBm]
If Optical Rx Power drops below the receiver's sensitivity threshold (typically -14.4 dBm for 10GBASE-LR), the photodiode receives insufficient photonic energy, flipping bits and generating thousands of CRC errors per second. Remediation requires inspecting the fiber ferrule with an optical scope, cleaning with an IBC one-click cleaner, and testing patch cables with a visual fault locator (VFL).
Switch Interface Error Counter Diagnostic Matrix
| Counter Name | Primary Root Cause | Layer | Detection Command | Remediation Action |
|---|---|---|---|---|
| CRC / FCS Errors | Bad Cat6 crimp, bent pins, EMI noise | Layer 1 (Physical) | show int counters errors | Replace patch cable; re-punch keystone jack |
| Runts (< 64B) | Half-duplex collision, failing NIC driver | Layer 1 / 2 | show int [id] | inc runts | Force auto-negotiation; update NIC driver |
| Giants (> 1518B) | Host sends Jumbo Frames on 1500B switch | Layer 2 (Data Link) | show int [id] | inc giants | Enable 9216B Jumbo Frames on switch port |
| Late Collisions | Duplex mismatch (Half vs Full) | Layer 1 / 2 | show int [id] | inc collision | Ensure both switch and host use Auto-Neg |
| Optical Rx Power Low | Dirty LC fiber connector, micro-bend | Layer 1 (Optical) | show int transceiver detail | Clean fiber endface; replace patch cable |
| Ignored / Overrun | Switch buffer exhaustion under burst load | Layer 2 (ASIC Buffer) | show int [id] | inc overrun | Enable 802.3x Flow Control; upgrade uplink |
Physical Port Isolation & Troubleshooting Checklist
- Clear switch port statistical counters (
clear counters [interface]) to verify whether errors are historical or actively accumulating in real time. - Swap the RJ45 patch cable connecting the switch port to the wall plate with a factory-tested Cat6 or Cat6a patch cord.
- Inspect the RJ45 port receptacle using a flashlight for bent, corroded, or debris-covered copper pins.
- Verify that both the switch port and the connected device's network card are configured for Full Duplex and Auto-Negotiation.
- For SFP/SFP+ fiber uplinks, check real-time Optical Rx Power via DOM CLI to ensure light levels are at least 3 dB above the sensitivity floor.
- If errors persist across multiple patch cables, re-punch the punchdown block or keystone jack using an impact punch-down tool.
Frequently Asked Questions
Do switch port CRC error counters automatically reset when the switch reboots?
On most enterprise switches (Cisco, Aruba, Juniper), counter registers reset to zero on reboot or power-cycle. However, on web-managed consumer switches, historical cumulative error counts may persist in flash until explicitly wiped using a Clear Counters command or web UI button.
Can a malfunctioning PC or server NIC cause CRC errors on the switch port?
Yes. A failing physical Ethernet controller on a PC, server, or IP camera can transmit malformed electrical signals with inaccurate voltages, invalid preambles, or corrupted timing that trigger CRC checksum failures at the switch ASIC.
What is an acceptable baseline number of CRC errors on a switch port?
On a healthy wired network, the acceptable rate of CRC errors is exactly zero. A link should be capable of transmitting tens of millions of frames without a single bit error. If CRC errors increase continuously during data transfers, physical remediation is required immediately.
Why do late collisions cause catastrophic packet loss compared to normal collisions?
In legacy half-duplex CSMA/CD, normal collisions occur within the first 64 bytes (the collision window) and the hardware immediately schedules a retransmission. A late collision occurs after 64 bytes have already been sent, meaning the hardware considers the frame transmitted and aborts Layer 2 retransmission, forcing higher-layer TCP timeouts that stall network traffic.
Related Switch Troubleshooting & Engineering Guides
- 10GbE SFP+ DAC vs RJ45 (10GBASE-T) Switch Guide — Compare 10GbE SFP+ Direct Attach Copper (DAC) cables, OM3/OM4 LC fiber transceivers, and 10GBASE-T RJ45 modules for low-power switch, router, and NAS links.
- Cat6 Keystone Jack & Patch Panel Wiring (T568A vs T568B) — Wire Cat6 and Cat6a keystone jacks and patch panels using TIA/EIA T568B vs T568A color codes, maintain <13mm pair twists, and pass 1–8 pin continuity tests.
- SFP & SFP+ Transceiver Not Working: Link Down Diagnostic Guide — Fix SFP and SFP+ transceiver link down faults, fiber polarity inversions, vendor EEPROM lockouts, auto-negotiation mismatches, and dirty LC connectors.
- How to Test Local LAN & WiFi Speed with iperf3 — Benchmark real local LAN, Cat6a switch, and WiFi throughput using iperf3 server and client commands to isolate cable bottlenecks without ISP speed limits.