Network-only benchmarking
iPerf3 10GbE Test Guide: Isolate Network Speed Correctly
File copies mix network, filesystem and storage performance, so they are poor first-line tests when you need to know whether 10GbE itself is healthy. iPerf3 creates controlled network traffic without making the disk path the headline variable.
Quick answer
Start by isolating the layer, not by buying hardware
File copies mix network, filesystem and storage performance, so they are poor first-line tests when you need to know whether 10GbE itself is healthy. iPerf3 creates controlled network traffic without making the disk path the headline variable.
Current Amazon listings
Multi-gig hardware matched by exact model
Technical decision
The troubleshooting decision
Use a single-stream baseline first, then add parallel streams only if needed to test aggregate behavior. Compare iPerf3 with file-copy speed: a large gap usually means the next investigation belongs outside the raw Ethernet path.
Interactive troubleshooting tool
iPerf3 10GbE Result Interpreter
Enter measured values where possible. The result is a diagnostic priority, not a substitute for vendor support, cable certification, security policy or controlled testing.
Multi-gig checklist
Four checks before changing or replacing anything
Find the slowest link
End-to-end throughput cannot exceed the slowest NIC, switch port, uplink, cable path, host bus or storage endpoint in the flow.
Separate line rate from file speed
10GbE is 10 gigabits per second at the Ethernet layer, not a promise that every SMB copy will show 1.25 GB/s.
Test the network without storage
Use iPerf3 or an equivalent network-only benchmark before concluding that SMB, RAID or a disk is the reason the Ethernet path is slow.
Measure storage as well as network
A single HDD, small RAID, SATA SSD, NVMe pool and RAM cache can produce very different ceilings even on the same 10GbE link.
Use current iPerf3 builds
ESnet released iPerf3 3.21 in April 2026. Use a current version on both endpoints when possible so platform fixes and features are not confused with network problems.
For iPerf3 10GbE Test Guide: Isolate Network Speed Correctly, checkpoint 1 records use current iperf3 builds. Keep iPerf3 10GbE Test Guide: Isolate Network Speed Correctly conditions fixed. Change one variable inside iPerf3 10GbE Test Guide: Isolate Network Speed Correctly, then remeasure use current iperf3 builds. If use current iperf3 builds moves, iPerf3 10GbE Test Guide: Isolate Network Speed Correctly stays on this layer. If use current iperf3 builds remains stable, iPerf3 10GbE Test Guide: Isolate Network Speed Correctly advances. That iPerf3 10GbE Test Guide: Isolate Network Speed Correctly discipline protects iPerf3 10GbE Test Guide: Isolate Network Speed Correctly from unrelated hardware changes.
Choose the server and client deliberately
Run the server on the NAS when supported or on a fast host attached to the same switch path. The client should use the exact NIC and route involved in the real workload.
Start with a simple single-stream test
A basic TCP test provides the cleanest baseline and is easy to repeat. Record negotiated link speed and CPU usage alongside the result so the number has context.
For iPerf3 10GbE Test Guide: Isolate Network Speed Correctly, checkpoint 3 records start with a simple single-stream test. Keep iPerf3 10GbE Test Guide: Isolate Network Speed Correctly conditions fixed. Change one variable inside iPerf3 10GbE Test Guide: Isolate Network Speed Correctly, then remeasure start with a simple single-stream test. If start with a simple single-stream test moves, iPerf3 10GbE Test Guide: Isolate Network Speed Correctly stays on this layer. If start with a simple single-stream test remains stable, iPerf3 10GbE Test Guide: Isolate Network Speed Correctly advances. That iPerf3 10GbE Test Guide: Isolate Network Speed Correctly discipline protects iPerf3 10GbE Test Guide: Isolate Network Speed Correctly from unrelated hardware changes.
Use reverse mode to test the other direction
Read and write paths can behave differently because of drivers, offloads or host CPU behavior. A reverse test helps reveal an asymmetric problem without changing the physical topology.
Add parallel streams only with a purpose
Multiple streams can better utilize some hosts or reveal aggregate capacity, but they can also mask a single-flow CPU or tuning problem. Keep the single-stream result as the reference point.
Interpret a healthy 10GbE result realistically
QNAP currently uses about 9.4Gbps, roughly 1.17GB/s, as a healthy 10GbE iPerf3 reference in its troubleshooting material. Treat that as a useful benchmark range, not a promise for every system.
For iPerf3 10GbE Test Guide: Isolate Network Speed Correctly, checkpoint 6 records interpret a healthy 10gbe result realistically. Keep iPerf3 10GbE Test Guide: Isolate Network Speed Correctly conditions fixed. Change one variable inside iPerf3 10GbE Test Guide: Isolate Network Speed Correctly, then remeasure interpret a healthy 10gbe result realistically. If interpret a healthy 10gbe result realistically moves, iPerf3 10GbE Test Guide: Isolate Network Speed Correctly stays on this layer. If interpret a healthy 10gbe result realistically remains stable, iPerf3 10GbE Test Guide: Isolate Network Speed Correctly advances. That iPerf3 10GbE Test Guide: Isolate Network Speed Correctly discipline protects iPerf3 10GbE Test Guide: Isolate Network Speed Correctly from unrelated hardware changes.
Watch CPU utilization during the test
If one endpoint reaches high CPU or a single core saturates, the test may be host-limited rather than network-limited. RSS, interrupts and virtualization can change how work is distributed.
Look at retransmits and errors, not only Gbps
A high average throughput number can hide an unstable path. Interface errors, retransmissions and large variability point back toward cabling, transceivers, switch ports or host networking.
Keep storage out of the conclusion
iPerf3 does not prove that RAID or SMB can sustain the same rate. Its job is narrower: establish whether the IP network can carry the traffic before you diagnose the file-service layers.
For iPerf3 10GbE Test Guide: Isolate Network Speed Correctly, checkpoint 9 records keep storage out of the conclusion. Keep iPerf3 10GbE Test Guide: Isolate Network Speed Correctly conditions fixed. Change one variable inside iPerf3 10GbE Test Guide: Isolate Network Speed Correctly, then remeasure keep storage out of the conclusion. If keep storage out of the conclusion moves, iPerf3 10GbE Test Guide: Isolate Network Speed Correctly stays on this layer. If keep storage out of the conclusion remains stable, iPerf3 10GbE Test Guide: Isolate Network Speed Correctly advances. That iPerf3 10GbE Test Guide: Isolate Network Speed Correctly discipline protects iPerf3 10GbE Test Guide: Isolate Network Speed Correctly from unrelated hardware changes.
Compare iPerf3 with the file-copy result
If iPerf3 is near line-rate but SMB is 300MB/s, investigate disks, protocol features and CPU. If iPerf3 itself is near 1Gbps, investigate negotiation and the physical path first.
Retest after every meaningful network change
Changing a cable, switch port, NIC driver or MTU should be followed by the same command and duration. Comparable tests make improvement measurable instead of subjective.
Save a small evidence bundle
Record command line, result, link rate, endpoint models, MTU and timestamp. That information is useful when asking a vendor for support or comparing future firmware and driver changes.
For iPerf3 10GbE Test Guide: Isolate Network Speed Correctly, checkpoint 12 records save a small evidence bundle. Keep iPerf3 10GbE Test Guide: Isolate Network Speed Correctly conditions fixed. Change one variable inside iPerf3 10GbE Test Guide: Isolate Network Speed Correctly, then remeasure save a small evidence bundle. If save a small evidence bundle moves, iPerf3 10GbE Test Guide: Isolate Network Speed Correctly stays on this layer. If save a small evidence bundle remains stable, iPerf3 10GbE Test Guide: Isolate Network Speed Correctly advances. That iPerf3 10GbE Test Guide: Isolate Network Speed Correctly discipline protects iPerf3 10GbE Test Guide: Isolate Network Speed Correctly from unrelated hardware changes.
Questions people ask
Common questions about iPerf3 for 10GbE
What is a good iPerf3 result on 10GbE?
QNAP currently cites about 9.4Gbps as a healthy 10GbE reference in its troubleshooting guidance. Hardware and software differences can move the exact value.
Does iPerf3 test my SSD or RAID?
No. It is specifically useful because normal operation does not depend on reading and writing your file workload.
Should I use multiple streams?
Start with one stream. Add parallel streams to test aggregate behavior or overcome a known single-flow limit, but keep both results.
What does reverse mode do?
It makes the server send toward the client, letting you compare the opposite traffic direction without rearranging the network.
Why is iPerf3 fast but SMB slow?
That usually means the raw network path is healthy and the next bottleneck is storage, SMB, CPU or filesystem behavior.
Why is iPerf3 only around 940 Mbps on a 10GbE system?
That strongly suggests a one-gigabit segment or equivalent limit somewhere in the path. Check negotiated speeds first.
Do I need jumbo frames for iPerf3?
No. A standard MTU baseline is valid and often preferable for initial troubleshooting.
Can CPU limit iPerf3?
Yes. Packet processing, virtualized networking and poorly distributed interrupts can limit a host before the wire.
Should I test through the same switch as the NAS?
Yes, when diagnosing the production path. The point is to reproduce the network topology while removing normal storage I/O.
Which iPerf3 version should I use?
Use a current ESnet build on both ends when practical; the project released version 3.21 in April 2026.
Primary references and methodology
Verify the exact port, cable, host and storage path
Use a single-stream baseline first, then add parallel streams only if needed to test aggregate behavior. Compare iPerf3 with file-copy speed: a large gap usually means the next investigation belongs outside the raw Ethernet path. In iPerf3 10GbE Test Guide: Isolate Network Speed Correctly, use current iperf3 builds is measured before add parallel streams only with a purpose. iPerf3 10GbE Test Guide: Isolate Network Speed Correctly then checks keep storage out of the conclusion before save a small evidence bundle. The iPerf3 10GbE Result Interpreter keeps iPerf3 10GbE Test Guide: Isolate Network Speed Correctly measurements ordered. In iPerf3 10GbE Test Guide: Isolate Network Speed Correctly, a passing layer is left untouched. A replacement purchase enters iPerf3 10GbE Test Guide: Isolate Network Speed Correctly only after the measured diagnosis points there.
- ESnet iperf3 project news
- QNAP: troubleshoot dramatic NAS transfer slowdowns
- Red Hat: network troubleshooting and performance tuning
As an Amazon Associate, Cloudzat may earn from qualifying purchases. Ethernet line rate is not the same as application throughput. Product revisions, chipsets, firmware, operating-system drivers, cable quality, thermals, switch configuration and storage performance can change results; verify the exact hardware revision before purchase.