Some initial speed tests and … ? I’m not sure where the bottleneck is, but there’s some sort of performance wonky.
I’ve got gigabit fiber, and speed test from the device is pretty decent:
Idle Latency: 19.69 ms (jitter: 0.37ms, low: 19.17ms, high: 19.99ms)
Download: 795.31 Mbps (data used: 889.9 MB)
21.07 ms (jitter: 1.63ms, low: 18.99ms, high: 32.69ms)
Upload: 722.11 Mbps (data used: 1.3 GB)
22.14 ms (jitter: 2.69ms, low: 16.86ms, high: 43.26ms)
Packet Loss: 0.0%
Speed test from a wired client, though… boy howdy:
Testing download speed................................................................................
Download: 351.29 Mbit/s
Testing upload speed......................................................................................................
Upload: 271.42 Mbit/s
I tried several times, and that wasn’t the best I got, but it wasn’t the worst either.
It looks like the a1300 is under some pretty severe load just trying to move packets across the interfaces.
If you’re just using it for a travel router, performance seems fairly decent:
OpenVPN, TCP, CHACHA20-POLY1305, run from A1300, displayed on web because the CLI display janked:
Speed test to same server from wired client was slower but still pretty respectable:
I suspect that some of the problem may be that I’m running iptables-nft (damn you Tailscale!) which may be responsible for some of the performance issues. I don’t know how much if any overhead that introduces over running nftables directly, but that would be the first place I’d look for some gains, if I cared. But largely I don’t.
WireGuard performance was weird too, getting 15mbps down and 200mbps up which seems… wrong? The upload speeds are in line with what you’d expect, so not sure where the D/L problem is. No time to debug at the present. I don’t recall having these problems running stock-ish on 21.02 with either the Slate AX or Brume 2 - I need to fire up an OG Brume and see what I get on performance there building a custom 21.02 and 22.03 build, since GL’s stock builds had some bad openssl options and suffered about a 40% performance penalty compared to a well optimized build.
So anyway… a mixed bag, maybe. I’m sure there are some things I could tweak a bit to get better performance, but this isn’t my main device, so my motivation is somewhat limited.