Route10 firmware 1.5f – L3 throughput / hardware acceleration issue

Routers
,
1

Hi Folks,

I’m experiencing what appears to be a port-specific hardware acceleration/forwarding issue on my Route10, which became apparent immediately after the router updated to firmware 1.5f in the early hours of 21st August.

Issue

LAN port 4 (L3) is now limited to approximately 900–1,000 Mbps WAN download throughput, despite negotiating at 2.5 Gbps.

LAN ports 2 (L1) and 3 (L2), using the same devices, cables and Speedtest server, achieve approximately 1.6–1.7 Gbps.

The problem follows physical port 4, rather than the connected device.

Testing performed

My internet connection delivers approximately 1.7 Gbps downstream / 105 Mbps upstream.

Route10’s internal Speedtest:

  • Download: ~1,695 Mbps

  • Upload: ~105 Mbps

Mac connected to a known-good LAN port:

  • Download: ~1,682 Mbps

  • Upload: ~106 Mbps

When the same Mac and Ethernet connection are moved to L4:

  • Download: ~895 Mbps

  • Upload remains approximately 100 Mbps

I initially discovered the problem through my VMware ESXi server.

ESXi is connected using an Intel I225-LM 2.5GbE NIC and negotiates at 2,500 Mbps full duplex.

With ESXi connected to L4:

  • Ubuntu VM: ~940–1,000 Mbps

  • Debian VM: ~940 Mbps

Moving the same ESXi server/cable to a known-good Route10 LAN port immediately results in:

  • Ubuntu VM: ~1,688 Mbps

No changes were made to ESXi or the VM between these tests.

I have also tested L3, which behaves normally like L2 and achieves full throughput. The issue therefore appears isolated to L4.

Route10 CPU behaviour

There is a significant difference in Route10 CPU utilisation depending on which physical LAN port carries the traffic.

During a Speedtest through L4 at approximately 900 Mbps:

  • Software IRQ (sirq): ~44%

  • ksoftirqd/0: ~16%

  • ksoftirqd/1: ~10%

  • ksoftirqd/2: ~13%

  • Suricata: ~26%

  • CPU idle: ~29%

During the equivalent test through L2 at approximately 1.7 Gbps:

  • Software IRQ: ~1%

  • ksoftirqd: approximately 0%

  • Suricata: ~1%

  • CPU idle: ~97%

This makes it appear that traffic through L4 may be falling back to software packet processing rather than using the hardware-accelerated forwarding path.

Port configuration

L2, L3 and L4 are configured equivalently.

L4 reports a 2.5 Gbps link, so this does not appear to be a 1 GbE autonegotiation issue.

There are no configured:

  • Client bandwidth limits

  • WAN download/upload limits

  • CAKE bandwidth limits

  • ESXi traffic-shaping policies

I have also:

  • Reset/reapplied the configuration for L4

  • Rebooted the Route10

Neither changed the behaviour.

Firmware

The issue became apparent immediately following the automatic update to 1.5f.

Given that L2 and L3 continue to achieve approximately 1.7 Gbps with very little CPU utilisation, while L4 is limited to approximately 900 Mbps and generates very high software-interrupt/packet-processing load, could you please investigate whether there is a known issue in 1.5f affecting hardware acceleration/flow offloading on individual LAN ports?

For the time being I have moved my ESXi server from L4 to L3, where full ~1.7 Gbps WAN performance has been restored.

Thanks

2

Same issue here!

I only have a gigabit network at the moment, but the speed dropped significantly on the same port. I am getting a download speed around 238 Mbps and upload around 32 Mbps. Tested with my Thinkpad T14 Gen1 attached via UGREEN 2.5Gb USB-c adapter and also on an M2 Macbook Pro. :grimacing:

3

Did anyone encounter the same issue just we did?

4

I am hoping once an Alta tech can see the post, they can confirm whether this is a firmware fault. It’s unusual that you are experiencing the same behaviour on the same port, and only since the 1.5f update. The only reason I noticed was that I run a speedtest Docker container that monitors my internet speed, and I got a lot of alerts because it was falling well below what I usually get; that’s when I started troubleshooting and narrowed it down to that specific port.

5

I took 5-6 speedtests because I thought that my UGREEN adapter has a driver issue. Afterwards, I realised that IDS/IPS is enabled but that would not slow down dramatically the main bandwidth. Then I was thinking..there was a firmware update and that could caused an issue. For the CPU throttle on the Route10, same issue is happening.

6

Ask support to rollback to 1.5e, there’s no way that a script can fix this.

7

We’ve resolved the issue with the LAN4 port not getting accelerated. Please reach out if you’d like to test it before release.

8

Hi @Alta-Jeff , Thanks for replying. I am working ok on another port, just glad you have gotten to the bottom of the issue. I can wait until the next firmware release

9

Looks like I’m experiencing this issue based on reading this post and how my net is behaving. I’m only seeing 300mbps on anything downstream on LAN4 and 40-50% CPU util with that load. However, speed test out from the Route10 itself to the WAN is normal. Any way I could get this test firmware to see if tha resolves my issue?