1-2 years ago I bought a DEC740. I usually run custom builds, but I wanted to support the project. I thought it would last me for a while, and this seemed like a (for me) future-proof purchase.
Today, I decided to plug my Mikrotik CRS305 into one of the SFP ports and run an iperf test from my Linux workstation to the DEC740.
This was unfortunately rather disappointing, at barely over 2 Gbits/sec. Following (fully) https://docs.opnsense.org/troubleshooting/performance.html and (loosely) https://binaryimpulse.com/2022/11/opnsense-performance-tuning-for-multi-gigabit-internet/ got me to a reasonably stable 2.58 Gbits/sec (both in iperf server and client mode). DEC740 'CPU Usage Total Host' is at 96.28 during the test. Interface speed is shown at 10Gbps in the GUI and shell.
I first thought this was a single-thread iperf problem, but I got the same speed via an SFTP upload (NVMe to NVMe).
All other physical machines/VMs connected to each other via the CRS305 operate at the expected ~10Gbps.
Could this be a compatibility problem with my Mikrotik Active Optics Direct Attach Cable, am I missing some setting, or is the DEC740 simply too weak to drive full 10Gbps speeds?
I don't think the routers own NVMe would be best target nor source.. Or generally any single NVMe for that matter... If you wanna test the networking capabilities of the router, do it with, say, 2 hosts using SPF+ each and whatever perf test you manage to do that just throws data to /dev/null in practice... Otherwise you're bot testing networking performance quite at all...
Do not test Iperf between a host and the GW, test it across.
PC1 (iperf) - OPNsense - PC2 (iperf)
As well upgrade your BIOS
https://forum.opnsense.org/index.php?topic=46672.15
Regards,
S.