Major speed regression with 21.02-rc3 on TP-Link Archer C7 v2

For Developers
1

Hello. I have two TP-Link Archer C7 v2 routers, both overclocked to 1 GHz. Because of a topic brought up on Reddit, I performed some benchmarks of overclocked routers.

The result is that the 21.02-rc3 release candidate has a serious performance regression as compared to 19.07.7. These are the results from one of the units (1 GHz CPU, 760 MHz DDR, 333 MHz AHB), as measured by iperf3 from a Linux laptop in LAN against a Linux PC in WAN, over a wired connection.

Test Version Download Upload
No SQM, no flow offloading 19.07.7 459 Mbit/s 432 Mbit/s
21.02-rc3 421 Mbit/s 301 Mbit/s
Cake + Piece of Cake set to 500 Mbps, no flow offloading 19.07.7 278 Mbit/s 221 Mbit/s
21.02-rc3 205 Mbit/s 176 Mbit/s
Cake + Piece of Cake set to 500 Mbit/s, software flow offloading 19.07.7 345 Mbit/s 291 Mbit/s
21.02-rc3 276 Mbit/s 300 Mbit/s

Any other TP-Link Archer C7 owners, with or without the overclock: can you confirm the regression?

Developers: is this known/expected?

2

Testing results might be skewed further by the fact you could be comparing ar71xx to ath79. Were you running 19.07 ath79 builds or not?

21.02 and up will be ath79 only.

1 Like
3

I tested ath79 vs ath79.

1 Like
4

just a general observation...

threads regarding this device have abound over the last 1.5-2yrs... and while I would not rule out some sort of quirk or need to revise oob packages or something...

seems pretty obvious to me... that alot of these issues are related to 'hardware capability decay'... in other words... what might be easy to label as a regression... can simply be put down to newer code on older hardware and the limitations that reveals...

2c

1 Like