A guide to loopback testing without the “deadlock”
Usually to test at these speeds you need to spend thousands on specialist equipment but this is an easy way to get a fairly accurate results and to certify the quality of the link. We used two UGreen 5Gbps USB C adaptors for this setup.
You can also use two different latops but ensure the USB C ports and the CPU are capable of these speeds.
Because this method ideally uses a single machine, we prioritize UDP testing but it is the same for two laptops.
This bypasses the “loopback deadlock” often encountered with TCP protocols at these extreme speeds and the network config setup will only work for UDP optimised testing. If you want to do TCP testing which is not a good indicator of cable quality, you will need to revert back to defaults.
Phase 1: Preparation & System Tuning
Before testing, your hardware must be tuned to handle approximately 800,000 packets per second.
1. Physical Connections Plug both 5Gbps adapters into separate Thunderbolt 4 ports to ensure sufficient bandwidth, then connect your test cable directly between the two adapters. (Note: Auto installed drivers by Windows MUST be updated to manufacturer latest drivers)
2. IP Configuration Configure static IP addresses for each adapter to create a direct link:
- Adapter 1:
192.168.254.1(Subnet:255.255.255.0) - Adapter 2:
192.168.254.2(Subnet:255.255.255.0)
3. Advanced Driver Settings To prevent artificial throttling and reduce CPU load, modify the following settings in Device Manager (Right-click Adapter > Properties > Advanced) for both adapters:
- Jumbo Frame: 9KB (Crucial for CPU load management).
- Receive/Transmit Buffers: 256 (Set to Maximum).
- Flow Control: Disabled.
- Energy Efficient / Green Ethernet: Disabled.
- Interrupt Moderation: Disabled (Required for accurate stats).
Phase 2: Execution
We will use “iPerf Version 3.20” for Windows to run the certification. Open two instances of Command Prompt (CMD) as Administrator.
Window 1: The Server
Run this command to listen for incoming traffic:
DOS
iperf3.exe -s -B 192.168.254.1
Window 2: The Client (The Test)
Run this command to blast data across the cable:
DOS
iperf3.exe -c 192.168.254.1 -B 192.168.254.2 -u -b 4500M -t 30 -i 1 -w 8M
Command Breakdown:
-u: UDP mode to bypass TCP overhead.-b 4500M: Targets 4.5Gbps, the realistic ceiling for 5GbE.-w 8M: Massively increases the socket buffer to prevent crashes.
Phase 3: Interpreting the Results
Once the 30-second test concludes, check the final summary line on the Receiver side to diagnose the cable.
| Jitter (ms) | Packet Loss (%) | Rating | Diagnosis |
| < 0.050 | 0% – 0.05% | Elite | Perfect Cable. Fully certified for 5GBASE-T. Small loss is negligible OS noise. |
| 0.050 – 0.20 | 0.05% – 0.5% | Good | Stable Cable. Minor noise or interrupts. Perfectly fine for daily use. |
| 0.20 – 1.00 | 0.5% – 2.0% | Degraded | Signal Issue. Cable may be too long (e.g., Cat5e > 45m) or near power lines. Expect “micro-stutters”. |
| > 2.00 | > 5.0% | Failing | Physical Defect. Bad termination, kinks, or poor shielding. The link will likely downgrade to 1Gbps. |
| 0.000 | 9% – 15% | Limit | System Overwhelmed. If jitter is low but loss is high, your CPU/Bus is choked. Check that Jumbo Frames are enabled. |
Phase 4: Thermal Health Check
High-speed Realtek controllers are sensitive to heat. You can diagnose thermal health by monitoring the “Bitrate” column during the test.
- Stable Bitrate (4.50 Gbps flat): Adapters are operating within safe thermal limits.
- “Saw-tooth” Pattern: The adapter is struggling with interrupt moderation or approaching thermal saturation.
- Speed Collapse (Drops to 0.00): Thermal Throttling. The chip has likely hit ~85°C and cut the link. Stop testing and cool for 5 minutes.
Final Step: Reverting for Daily Use
Crucial: Once testing is complete, you must revert the Jumbo Frame setting to “Disabled” or “1500”. Most consumer routers cannot handle 9KB packets, and leaving this setting active will break your standard internet connection.
