ArmSoM
← All stories

Measuring 2.5GbE Throughput on an SBC: Avoiding Storage Bottlenecks

A negotiated 2.5GbE link is the start of a network test, not its result. Clients, switches, cables, software and storage can each limit transfer speed. Sige7 and Sige3 tests should record the complete path so the results can be reproduced.

ArmSoM Sige7

ArmSoM Sige7

Establish the link and the endpoint capacity

Check negotiated speed and errors on both endpoints. Confirm that the peer can generate the required traffic before testing the SBC. Separate network-only measurements from file transfers, then add storage and concurrent workloads. Keep direction, packet size and test duration with every reported number.

Add storage and protocol overhead

After the network baseline, test the actual file-sharing service with representative files. Large sequential files and many small files answer different questions. Observe disk activity and CPU use on both systems. Avoid presenting a cache-only copy as sustained drive performance. Sige7’s two 2.5GbE ports and Sige3’s mixed port arrangement should be tested according to the path actually used.

Report the bottleneck, not just the best number

Repeat measurements and include the environment: client, switch, cable, operating system, drive and protocol. Note whether the result is in bits or bytes per second. If one component limits the test, explain that limitation rather than implying the board cannot exceed the observed figure under any circumstances.

ArmSoM Sige3

ArmSoM Sige3

Match the choice to your requirements

Requirement or situation

What to evaluate

Decision detail

Network-only

Known-capable peer

Exclude disk effects.

File transfer

Real protocol and dataset

Include storage and metadata overhead.

Interpretation

Units and repeatability

Avoid cache and endpoint confusion.

Example design

Find the bottleneck in a 2.5GbE path

Start with a network-only test between known-capable endpoints, then add the file-sharing protocol and storage. If the first test is healthy but file copies are slow, investigate the disk, filesystem, protocol and workload before blaming the Ethernet port. Multiple clients and link aggregation introduce additional conditions; do not present their combined throughput as the speed of one ordinary file transfer.

Hardware and documentation

Sige7 · RK3588

RK3588, 6 TOPS NPU, dual 2.5GbE and M.2 PCIe 3.0 x4. Check the selected eMMC and wireless configuration.

Hardware documentation · Product details

Sige3 · RK3568B2

RK3568B2 with a 2.5GbE port and a separate Gigabit Ethernet port. Consult the revision-specific manual for expansion and boot support.

Hardware documentation · Product details

Frequently asked questions

Why can a 2.5GbE link copy files much more slowly?

The transfer may be constrained by storage, protocol, CPU, client performance or small-file overhead.

Product information and support

Explore Sige7

Contact ArmSoM

Sources

Written by ArmSoM using the documentation below. Example designs are proposals, not measured customer installations.

Related reading