A 2.5GbE link that falls to 100Mbps only after sleep or reboot usually points to power-state renegotiation, driver behavior, or a marginal physical pair.
This is a different problem from a port that can never negotiate above 1GbE. First prove the link reaches 2.5GbE from a cold working state, then reproduce the drop after sleep or reboot while keeping the same cable and switch port. Compare advertised modes, power-saving settings, driver state, and error counters before replacing the whole network path.
Prove the 100Mbps Drop Happens Only After a Power Transition
Record the negotiated speed immediately after a clean cold boot, after several minutes of traffic, after system sleep and resume, and after a normal reboot. Use the NIC and switch port status rather than internet speed-test results.
A Realtek troubleshooting write-up describes a deep-sleep link-state problem and shows why the transition itself can be the useful discriminator.
If the link starts at only 100Mbps even from cold boot, return to the existing generic negotiation article. If it reliably begins at 2.5GbE and only falls after sleep or reboot, preserve that state transition for the next tests.
Check NIC Power-Saving Settings Before Changing the Cable
Inspect the adapterโs power-management and advanced-driver settings for Energy Efficient Ethernet, Green Ethernet, Reduce Speed on Power Down, Wake-on-LAN link speed, and permission for the operating system to power down the NIC.
A current Ethernet tuning guide notes that power management can disable adapters independently from ordinary speed and duplex configuration.
Disable or change one relevant power option at a time and repeat the same sleep-resume test. Do not disable every offload and performance feature together, because a successful result would not reveal which setting actually mattered.
Recheck the Driver and Chipset After Resume
Record the exact NIC chipset, active driver version, firmware where applicable, and advertised link modes before and after the drop. A resume path can reload the same device with different driver state even though the interface name remains unchanged.
PLANEXโs 2.5GbE adapter breakdown shows that 2.5GbE drivers expose power controls alongside ordinary Ethernet settings.
If updating or rolling back the NIC driver removes the post-resume fallback while the cable and switch stay fixed, document that version. Avoid forcing a static speed as the first fix because that can hide a broken auto-negotiation or resume path.
Rule Out a Marginal Cable Pair That Fails on Renegotiation
A cable can work at 2.5GbE until the link renegotiates and then settle at a lower mode if one pair, termination, coupler, or connector is marginal. Reseat both ends and compare with one short known-good cable on the same ports.
A recent Ethernet troubleshooting article explains that 100Mbps can indicate pair failure even when the visible cable category looks adequate.
If the short cable survives repeated sleep and reboot cycles at 2.5GbE, test the original patch leads, wall jacks, couplers, and in-wall run separately. Fix the physical path rather than leaving the NIC forced to a higher mode.
Check USB or Docked 2.5GbE Adapters Separately
If the server or client uses a USB-C or docked 2.5GbE adapter, include the USB controller and dock power state in the test. Resume can reset the USB device before Ethernet auto-negotiation begins.
A detailed 2.5GbE adapter test found that USB adapters depend on drivers and host-side enumeration are part of the path, not just the RJ45 cable.
Test the same adapter directly on the host without the dock, then through the dock. If only the docked path falls back after resume, keep the fix at USB power and dock firmware rather than the NAS switch.
Verify 2.5GbE Survives Multiple Sleep and Reboot Cycles
After changing the confirmed cause, run at least two sleep-resume cycles, one warm reboot, and one sustained LAN transfer while watching negotiated speed, link errors, and renegotiation events on both ends.
A networking article on multi-gig performance reminds readers that link speed differs from throughput so the final test should validate both stable negotiation and useful traffic.
The repair is complete when the same 2.5GbE link survives power transitions without dropping to 100Mbps. The related ZimaSpace article on a NAS port negotiating at 1GbE remains the correct path when the problem is no longer specific to sleep or reboot.
Support & Tips
More to Read

Can Plex Share a GPU With Another Docker Container?
Plex and another container can often access the same GPU, but you must test driver support, device mapping, video-engine load, memory, and recovery behavior.

How to Tell Whether a Plex Error Comes From the Client or Server
Reproduce the same item on another client, compare the session path, then collect server evidence only after scope tells you where the failure actually...

How to Configure Plex Cache and Transcode Temporary Storage
Protect persistent Plex state while placing transcode temp files on suitable local storage, then verify cleanup, free space, and restart behavior.

