OPNSense on Protectli max WAN speed of 600Mbps

Hi, everyone!

Great stuff here, and thanks, Tom for all your videos.

I’ve just upgraded my Spectrum service to 1Gig and had a tech come out and confirm I’m getting better than 940 through my Motorola cable modem.

The problem I’m having is max speedtest results on a Windows 10 box connected directly to the same Unifi 1G switch as the Protectli LAN port. I get up to 650, occasionally a bit higher, but nowhere near the 940 I get connected directly to the modem.

I tried running speedtest-cli directly on the firewall, but the results were even more abysmal - around 340.

Anybody else have a similar issue getting full speed through the firewall?

My Protectli is about a year old, four ports with an i3 chip - CPU and memory are generally below 10%.

Thanks in advance, and apologies if this was discussed and I didn’t find the thread in my search.


Correction - The vault I have is Intel Quad Core (Atom E3845) - not i3.

Thanks again!


Do you use PPPOE? If so, that can cause only a single core to be used. Setting the isr handling to be deferred per this doc can help, but I also has other suggestions that may, or may not apply to your hardware APU2 1Gbit throughput on pfSense (configuration instructions)


I’ll take a look.


Yes. I followed a similar instruction sheet on their site for OPNSense:

I didn’t see a huge difference, but in reading the link you sent, that version talks about getting the full gigabit using two connections - I was just trying a single speed test.

I’m going to test with two connections and see what happens.

Thanks agian.

With two clients running speedtest the router is still getting 660Mbps max.

Does your Protectli box have Realtek or Intel NICs? Sometimes Realtek NICs on a lower frequency CPU are not able to reach gigabit speeds (900+ Mbits).

It would also worth to have a look via the CLI to see if any core is maxed out during the speed test [especially soft IRQs].

Thanks, Adrian.

The Protectli box I have says they’re Intel NICs. I’ll monitor the cores to see if it’s pegging one of them during the test. Great idea. Thanks again!