OpenWrt Forum Archive

Topic: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

The content of this topic has been archived between 19 Apr 2018 and 6 May 2018. Unfortunately there are posts – most likely complete pages – missing.

I didn't take any readings and have replaced my old router now so don't want to mess around too much.

I'll post an update if I do find some time to mess with it, but getting back to stock sounds like a bit of a hassle.

Update with LEDE 17.01 stable: I get up to 500 Mbit/s throughput with both a rMBP 2016 on MacOS and a Lenovo T460s on Linux right in front of the router from WiFi to LAN (a NAS on the same network, no NAT). I did not try stock firmware, but I assume that at least the rMBP 2016 should be able to achieve 800 MBit/s throughput.

(Last edited by makruiten on 26 Feb 2017, 23:17)

Yeah, I also did the upgrade to LEDE but didn't see any improvement at all.

I think I've been maxing out at ~280 on either firmware.

I got 200...300Mbit tcp LAN to LAN performance with snapshot openwrt. With dd-wrt r29095 little bit more.
It was tested with different cables and network cards. I think there is a problem with the ethernet driver. Polling? DMA? PLL config?

The performance problem was also seen in a WR1043v4. There are little bit more CPU clock frequency, but little bit slower than Archer C7, with very similar hardware.

It looks like a platform default driver problem.

LAN to LAN or WiFi to LAN or NAT LAN to LAN?

LAN to LAN doesn't make sense, as its only over the switch.
NAT LAN to LAN that seems about right, any faster than that is only via stock with the hardware NAT support (although that potentially creates its own compatibility issues).

Also, were you running iperf on the router itself, as that is not recommended as it WILL be CPU bound.  eg on my Archer C7 running iperf on the router only achieves 148Mbit max but from a PC on WiFi to LAN it does up to 500Mbit.

Speed tests should always be PC to PC where there is unlikely to be a CPU bottleneck.  I also would recommend iperf3 rather than iperf, as I have had iperf results when testing over the Internet that were completely impossible - faster than my broadband speed.

makruiten wrote:

Update with LEDE 17.01 stable: I get up to 500 Mbit/s throughput with both a rMBP 2016 on MacOS and a Lenovo T460s on Linux right in front of the router from WiFi to LAN (a NAS on the same network, no NAT). I did not try stock firmware, but I assume that at least the rMBP 2016 should be able to achieve 800 MBit/s throughput.

Thought ~500Mbps actual throughput is about right for a 3x3 setup.

Cheers.

(Last edited by loyukfai on 8 Mar 2017, 18:27)

Not at all, I can get ~500Mbit over 2x2.

Have anyone tried alternative ath10k firmware from Candela Technologies (candelatech.com/ath10k-10.4.php)? I'm using it on my Archer C7 and it seems pretty realiable, however I can't run mastabog's tests. In my setup the bottleneck are 200Mbps ehternet powerline adapters (and I have no access to the back of the router without breaking stuff), so actually there's no point to test higher WiFi rates.

The discussion might have continued from here.