Check your settings and connections. Ensure the device is properly paired and there are no interference issues.
Check your settings and connections. Ensure the device is properly paired and there are no interference issues.
I've received a PCIe card for my NAS (Ubuntu 24.04) but when testing with iperf3 from Windows 11, which uses 2.5G Ethernet, I only get 1.35Gbps in one direction and 1.6Gbps in the other. The speed remains steady and consistent throughout the test. The connection between the computers is a TrendNET TEG-S350 unmanaged 2.5G switch. Both systems use RealTek adapters—RTL8125 on an Ubit 2.5G PCIe x1 adapter and RTL8125B on an MSI B550 Gaming Edge WiFi. All cables are CAT 5E, under 6ft in length. If this isn't a 2.5G connection, it's likely a TP-Link AX1500 router. Despite the lower speeds, data appears to bypass the router entirely. Where should I begin troubleshooting?
I'm curious about whether the Windows operating system includes a Quality of Service feature that restricts performance limits.
Use iperf to verify Linux to Linux connectivity, then test Windows to Windows if needed to confirm stability. Ensure Windows 8125 comes with Realtek drivers installed directly, as previously it only supported 700Mbps for the 8111 model until the official update was released.
The reported speed matches what both NICs indicate, confirming the connection is a 2.5Gb link.
I plan to verify with a Linux Live USB soon to confirm the operating system. I’ll also attempt to reinstall the Windows drivers, typically sourcing them from the motherboard site, though it’s safe to try. Both devices are showing 2500Mbps. The switch displays a 2.5G connection for both units using their indicator lights. I’ve also checked that the NAS’s NIC is reporting the right PCIe generation—Gen 2 at 5.0GT/s (about 500MB/s)—and that the drives are connected via SATA II 3.0 Gbps, not 1.5Gbps, which seems to be the intended configuration.
Consider using cat6 cables; perhaps your cat5e connections aren't supporting speeds above 1Gb/s.
I don't actually possess two CAT 6 cables, but I attempted rearranging all the available ones (they're all CAT 5E) and the outcomes remain identical regardless of the arrangement. For more information, here are the iperf results: first from the Windows machine to the Linux one, then in reverse. These findings stay consistent – I obtain the same by configuring Windows as the server and Linux as the client. Switching between Windows to Linux consistently yields 1.6Gbps, while Linux to Windows gives about 1.35Gbps. Spoiler: F: Downloads/iperf-3.17.1-win64. I'll run the command from the Windows machine to the Linux host and vice versa.
I managed to install a Linux LiveUSB successfully. It seems the main computer needed GPT formatting for UEFI, likely because Secure Boot was enabled for Windows 11—this process took longer than expected, requiring multiple attempts before resolving the issue. The transfer speeds were consistent: 1.6Gbps from main to NAS and 1.35Gbps in the opposite direction. I reinstalled network drivers on Windows 11 via MSI but got no improvement. Using RealTek drivers directly from their site also didn’t help. I then installed the latest RealTek drivers into Ubuntu’s kernel, yet results remained unchanged. It appears the problem isn’t with the operating system or drivers themselves, but possibly with the cables—despite differences in length, source, and purchase timing. All adapters seem identical. Should I accept that one or both might be faulty, and is there a way to identify which adapter is causing the issue?
I finally had the chance to execute this test. The outcomes matched perfectly. Below are the iperf3 data from the Windows 11 machine using two different cables, with both devices directly linked:
- Spoiler: Downloaded from F:\Downloads\iperf-3.17.1-win64, executed via .\iperf3.exe
- Results show identical performance across both connections.
Here are the key findings:
- Initial connection to 192.168.0.25 on port 5201 succeeded.
- Transfer speeds remained consistent regardless of cable used.
- Interval and bitrate metrics stayed stable throughout the test.
If you need further details, let me know.