* absolutely terrible wireguard performance on a 4ghz box - need tuning help
@ 2022-10-25 6:58 Kyle Sanderson
2022-11-04 20:10 ` Daryl Richards
2022-11-04 20:55 ` Metin Evrim Ulu
0 siblings, 2 replies; 3+ messages in thread
From: Kyle Sanderson @ 2022-10-25 6:58 UTC (permalink / raw)
To: wireguard
hi wireguard,
APU2C4 - AMD Embedded G series GX-412TC, 1 GHz quad Jaguar core with
64 bit and AES-NI support
There has to be a way to improve this. I'm getting 20-40MB/s (causing
a system load of 6 on this poor box) where SSHFS in comparison is
still able to fly. Traffic not traversing the tunnel is also,
strangely, impacted as well. I have the Intel i210AT controller, which
doesn't appear to have any offloading to it when wireguard is present.
ksoftirqd/0 spins at around 50%, ksoftirqd/1 spins around 20%, and the
other 12(!) wg-crypt threads hover around 20% respectively.
If I disable wg I'm able to get line-rate through acceleration.
The other 2 ksoftirqds sit around 5% respectively. kworker events
jumping between 10-25%.
Regrettably openwrt doesn't appear to have perf available...
K.
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: absolutely terrible wireguard performance on a 4ghz box - need tuning help
2022-10-25 6:58 absolutely terrible wireguard performance on a 4ghz box - need tuning help Kyle Sanderson
@ 2022-11-04 20:10 ` Daryl Richards
2022-11-04 20:55 ` Metin Evrim Ulu
1 sibling, 0 replies; 3+ messages in thread
From: Daryl Richards @ 2022-11-04 20:10 UTC (permalink / raw)
To: wireguard
On 2022-10-25 2:58 a.m., Kyle Sanderson wrote:
> hi wireguard,
>
> APU2C4 - AMD Embedded G series GX-412TC, 1 GHz quad Jaguar core with
> 64 bit and AES-NI support
>
> There has to be a way to improve this. I'm getting 20-40MB/s (causing
> a system load of 6 on this poor box) where SSHFS in comparison is
> still able to fly. Traffic not traversing the tunnel is also,
> strangely, impacted as well. I have the Intel i210AT controller, which
> doesn't appear to have any offloading to it when wireguard is present.
>
> ksoftirqd/0 spins at around 50%, ksoftirqd/1 spins around 20%, and the
> other 12(!) wg-crypt threads hover around 20% respectively.
> If I disable wg I'm able to get line-rate through acceleration.
> The other 2 ksoftirqds sit around 5% respectively. kworker events
> jumping between 10-25%.
>
> Regrettably openwrt doesn't appear to have perf available...
Perhaps this is an OpenWRT issue? I have a number of these APUs with
astlinux installed, doing wireguard, and can easily do 300-400mbps, and
that's across the internet so perhaps faster if on a test bench.
The APU isn't the problem..
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: absolutely terrible wireguard performance on a 4ghz box - need tuning help
2022-10-25 6:58 absolutely terrible wireguard performance on a 4ghz box - need tuning help Kyle Sanderson
2022-11-04 20:10 ` Daryl Richards
@ 2022-11-04 20:55 ` Metin Evrim Ulu
1 sibling, 0 replies; 3+ messages in thread
From: Metin Evrim Ulu @ 2022-11-04 20:55 UTC (permalink / raw)
To: wireguard
Kyle Sanderson <kyle.leet@gmail.com> writes:
> hi wireguard,
>
> APU2C4 - AMD Embedded G series GX-412TC, 1 GHz quad Jaguar core with
> 64 bit and AES-NI support
>
Um, I don't recall any Rijndael/AES in WG implementation, any particular
reason for mentioning AES-NI?
best,
evrim.
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2022-11-04 20:59 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-10-25 6:58 absolutely terrible wireguard performance on a 4ghz box - need tuning help Kyle Sanderson
2022-11-04 20:10 ` Daryl Richards
2022-11-04 20:55 ` Metin Evrim Ulu
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).