IPQ4019 cake performance?

What is the approximate performance of the IPQ4019 devices like the Linksys EA8300 with cake traffic shaping? I’m currently using the Archer C7 and looking for an upgrade. I know Marvell is fast but I’m hoping for better WiFi stability with Qualcomm and don’t need the extra performance of the Netgear 7800 if the EA8300 suffices. Thank you.

My experience with the Archer C7v2 is that it couldn't push more than 100-200 Mbps with CAKE, being CPU bound between just handling the Ethernet interface (tops out around 400-500 Mbps with no other load) and the additional processing associated with CAKE.

The EA8300, or any of the ipq40xx-based devices, will be significantly better, between having multiple cores and any possible single-core speed improvements of the ARM architecture over the MIPS architecture.

What rates are you trying to achieve?

Thanks Jeff. Your posts and work on the EA8300 is what made me interested in the first place. I’m looking to shape around 150Mbps. I’ve found that the Archer C7 can’t do more than 100Mbps or so on WiFi. It might be ok purely wired and with a separate AP but I’d prefer a single all in one device

I can check it this weekend, but I have good confidence that the EA8300 or similar should be able to shape 200+ Mbps. I use my wireless devices as meshed, "dumb APs", but have an EA8300 "on the bench" as well. @slh may have already checked this on one of the devices he's worked with. There shouldn't be much difference in this kind of performance between any of the ipq40xx-based devices.

I can't speak to IPQ4019, but I was briefly doing builds for a GL.iNet B1300 (IPQ4028) and that thing could shape my 200Mbps cable with Cake just fine.

1 Like

Yep, I would find my C7's getting up to 140 or so, on ethernet only, Wifi was bringing that down to 115-125mbit, with cake shaping things..

Finally bit the bullet and bought a Zotac C327 box though, and that is barely 1/3d saturated with my up to 350mbit cable connection....

My NBG6617 showed absymal performance with SQM enabled. It might have something to so with the funkiness surrounding IPQ40xx switches - I have no idea. This was on a 50 Mbps cable downlink, went to ±25 Mbps. Bufferbloat did improve though :face_with_raised_eyebrow:

I tested SQM and it is perfect on 100 Mbps Ethernet.
It is not quote clear for me if I need it at all, that's another question.

Did bufferbloat improve? If so you could benefit from it.

Yes, it did improve on dslreports test from B to A+. But I don't feel the bufferbloat without SQM.