Repeater Mode (connected to 2.4GHz wireless network)
Client Connection: 5GHz
Speed: 84Mbps
NordVPN enabled: 8Mbps
Repeater Mode (connected to 5GHz wireless network)
Client Connection: 5GHz
Speed: 34Mbps
NordVPN enabled: 8Mbps
Ethernet Mode (wired directly to switch)
Client Connection: 5GHz
Speed: 330Mbps
NordVPN enabled: 8Mbps
I disabled the VPN on the router and then downloaded the NordVPN client for my computer and ran a speed test. Speed: 268Mbps. That tells me that NordVPN servers are not impacting performance more than 10%.
I’ve performed multiple resets, both with holding the physical switch and doing a Firmware reset in the admin console.
Speed is unusable with the VPN enabled, but it is also seems like extremely poor performance in Repeater mode. I would expect a 50% hit, but it’s much worse than that, and in 5G, it’s close to a 90% performance impact.
No other configuration changes. Pulled from the box, upgraded firmware, attached to wireless network, set timezone, setup VPN.
What kind of rates do you get going using pure Ethernet cabling? I’ve noticed consider impact combining Wi-Fi w/ VPN, even with WireGuard, as the unneeded variable of client device radios just so vastly range in quality it can be rather disturbing. It’s better to remove it fr the equation, IMO.
Also, I don’t use Nord but t it may be worth a look. WG is advertised as 91 Mbps vs 21 Mbps on your Beryl:
Pure Ethernet (from Beryl to home router and from computer to Beryl) is roughly the same with VPN enabled. It’s 10.6Mbps. If I disable the VPN, it’s 922Mbps.
If I enable NordVPN on the my computer (disabling on the Beryl), I get 637Mbps.
I’m now wondering if this isn’t a MTU issue. Can you get such a setting from the PC’s Nord client?
Failing that all I can suggest is pulling the CL VPN Client conf files fr the GL device, then reflashing via U-boot… wholly overwriting the NAND flash as if the previous firmware(s) were never installed.
I believe what I’m seeing is a lack of hardware flow offloading. The issue referenced in the video below is almost exactly what I’m seeing. I can’t run htop but I’m seeing almost exactly 25% cpu usage when running a speedtest. That suspiciously looks like I’m running 1 core at 100%. (4 core cpu onboard)
In the video, he enables the hardware flow offloading in the command line. I went through the Luci console and checked and confirmed it is enabled both in the interface and from the command line.
root@GL-MT1300:~# uci show | grep 'offloading'
firewall.@defaults[0].flow_offloading='1'
firewall.@defaults[0].flow_offloading_hw='1'
Ah, thanks for the package… it just wasn’t loaded.
After running htop, I can see a single core go to 100% during a speed test. I even back-rev’d the firmware to that last 3.x build and I see the same thing.
The striking thing about Van Tech’s test is @ 9.46: he’s not connected via WG or OVPN when running a H/W offloaded test. What kind of VPN rates to do you get when offload is disabled?
I think it’s also worth mentioning that Nord puts their own ‘flavour’ on WireGuard… & call it NordLynx. It’s enough of a reason I stay away fr 'em. You might want to consider a 30 day period of a ‘pure WG’ provider just to remove that variable but as always, YMMV:
Significant improvement in using a WireGuard VPN solution. (used your suggestion and found a free one to test with) That got me to 130Mbps which is more than sufficient for my needs.
Brilliant! That’s a real nice rate you’ve got there… well over the GL advertised max speed. Don’t forget to mark the ‘Solution’ to let others know who may stumble upon a similar issue.
