TP-Link Archer C7 vs Home Hub 5

I have LEDE installed on both an Archer C7 and a Home Hub 5 but the performance is wildly different.

Its my understand they both use the same 802.11ac chipset so why am I able to comfortably get up to 55MiB/s on the Archer C7 but the Home Hub 5 struggles to hit 20MiB/s, when tested in the same location?

The HH5 also is performing identical at 40Mhz and 80Mhz channel width, although it does halve accordingly if I use 20Mhz.

It is showing as using VHT modes but the link rates are much lower than I would normally expect. I have seen it occasionally show 400Mbit but generally a lot lower. The Archer C7 comfortably lingers around 720Mbit with it not unusual to hit the full 866Mbit for 2x2 MIMO.

I was under the impression the HH5a was supposed to have better amplifiers not worse? Could leaving the wires I used to access the UART interface be causing signal disruption? Is it likely I will ever need them again or should I be safe to remove them?

There are differences in the WLAN chipsets.

TP-Link -

WLAN Hardware:
Qualcomm Atheros QCA9558, Qualcomm Atheros QCA9880-BR4A

Home Hub 5 -

WLAN Hardware:
Atheros AR9227, Qualcomm Atheros QCA9880

Sorry I should have said its an Archer C7 v2 which uses the QCA9880-BR4A for 5Ghz, just like the HH5.

Archer C7 v2:

    inactive time:  4 ms
    rx bytes:       2113260
    rx packets:     22228
    tx bytes:       729641174
    tx packets:     476135
    tx retries:     0
    tx failed:      1
    rx drop misc:   0
    signal:         -65 [-70, -69, -69] dBm
    signal avg:     -64 [-69, -68, -68] dBm
    tx bitrate:     6.0 MBit/s
    rx bitrate:     650.0 MBit/s VHT-MCS 7 80MHz short GI VHT-NSS 2
    authorized:     yes
    authenticated:  yes
    associated:     yes
    preamble:       long
    WMM/WME:        yes
    MFP:            no
    TDLS peer:      no
    DTIM period:    2
    beacon interval:100
    short slot time:yes
    connected time: 95 seconds

HH5:

    inactive time:  0 ms
    rx bytes:       2245010
    rx packets:     24589
    tx bytes:       453190024
    tx packets:     295543
    tx retries:     0
    tx failed:      1
    rx drop misc:   0
    signal:         -57 dBm
    signal avg:     -56 dBm
    tx bitrate:     6.0 MBit/s
    rx bitrate:     780.0 MBit/s VHT-MCS 8 80MHz short GI VHT-NSS 2
    rx duration:    1687736 us
    authorized:     yes
    authenticated:  yes
    associated:     yes
    preamble:       long
    WMM/WME:        yes
    MFP:            no
    TDLS peer:      no
    DTIM period:    2
    beacon interval:100
    short slot time:yes
    connected time: 63 seconds

The Archer C7 despite having a slower link rate is doing a real-world transfer rate of double the HH5.

This is confirmed with iperf3, scp and samba transfers.

The TP-Link example I posted is for the Archer C7 v2.

As posted above, the HH5 does not use the QCA9880-BR4A...look up the tech data for each from the TOH page.

If the tech data page is incorrect, it should be updated.

Seems pretty clear to me it does:

HH5:

[ 11.213370] ath10k_pci 0000:02:00.0: qca988x hw2.0 target 0x4100016c chip_id 0x043202ff sub 0000:0000
[ 11.221235] ath10k_pci 0000:02:00.0: kconfig debug 0 debugfs 1 tracing 0 dfs 1 testmode 1

[ 11.236578] ath10k_pci 0000:02:00.0: firmware ver 10.2.4-1.0-00016 api 5 features no-p2p,raw-mode,mfp crc32 0c5668f8

Archer C7 v2:

[ 20.051206] ath10k_pci 0000:01:00.0: qca988x hw2.0 target 0x4100016c chip_id 0x043202ff sub 0000:0000
[ 20.060608] ath10k_pci 0000:01:00.0: kconfig debug 0 debugfs 0 tracing 0 dfs 1 testmode 1
[ 20.073672] ath10k_pci 0000:01:00.0: firmware ver 10.2.4-1.0-00029 api 5 features no-p2p,raw-mode,mfp,allows-mesh-bcast crc32 88595bb8

Yes I have tried copying over the firmware-5.bin from the Archer C7 to the HH5, it doesn't improve anything.

I wonder if "debugfs 1" could be a hint at something that might be slowing it down?

https://lede-project.org/toh/hwdata/tp-link/tp-link_archer_c7_ac1750_v2.0

https://lede-project.org/toh/hwdata/bt/bt_homehub_5_type_a

It seems to me that https://wiki.openwrt.org/toh/bt/homehub_v5a is correct, as dmesg confirms they are EXACTLY the same target and chip_id, which surely wouldn't be the case if they weren't the same chip?

Anyway I think I found the problem:
CPU: 8% usr 3% sys 0% nic 2% idle 0% io 0% irq 84% sirq

I used to have the same problem on the Archer C7 v2 in the early days of OpenWRT support, so it seems the HH5 build may need a little TLC.

I do wonder about that debugfs flag, as IME nothing called debug ends well when it comes to CPU consumption. :wink:

[ 11.221235] ath10k_pci 0000:02:00.0: kconfig debug 0 debugfs 1 tracing 0 dfs 1 testmode 1

I'm informed by 'mkresin', the 'debug 0' flag disables ALL debugging and so the 'debugfs 1' flag has no effect.

He added actual throughput is limited to wired ethernet speed of 180mbps, because LEDE has no support for the xrx200's undocumented features called PPA, and directpath mode.

See also this old thread from a year ago:
https://openwrt.ebilan.co.uk/viewtopic.php?f=7&t=149

WikiDevi pages agree with the LEDE pages...

https://wikidevi.com/wiki/TP-LINK_Archer_C7_v2.x

https://wikidevi.com/wiki/BT_Home_Hub_5A

The QCA9880-BR4A is one of three reference designs of the QCA9880 (Peregrine) chipset...

QCA9880-AR1A v1
QCA9880-2R4E
QCA9880-BR4A v2

It may report the same, but whether it performs the same is another question.

Is SMP still disabled on the current builds? Do we know if the problems with that enabled were ever resolved?

I have tried searching the forums but the results are really unhelpful. Is there someone specific who pioneered support for this device?

For the specific case of the bt hh5a, smp should be enabled in master snapshots - other devices with FXS port do disable SMP (the second core is dedicated to the vmmc there, doing the voice processing).

So not in the stable build?

A casual look at the kernel configs of the lede-17.01 branch suggests that SMP is disabled there.

I have found since moving off if .01 or .02 I can't remember exactly which, my hh5a is much slower with WiFi, I'm on UK fibre an I get the full 74 mb downloads which is pretty good and on .01 I got this on wifi, now I struggle on .04 to get above 55mb

fwiw, I don't have 80/20 FTTC but here are some test results I have compiled between my two Windows laptops (gigabit port and 2x2 stream wifi card) running Filezilla FTP client and Server apps, connected to a HH5a:

FTP client - LAN - HH5a - LAN - FTP server
17.01.0 - 113.2 MiB/sec (905 mbps)
17.01.2 - 113.2 MiB/sec (905 mbps)
17.01.4 - 113.2 MiB/sec (905 mbps)

FTP client - 5 GHz wifi (20MHz) - HH5a - LAN - FTP server
17.01.4 - 11 MiB/sec (88 mbps)

FTP client - 5 GHz wifi (40MHz) - HH5a - LAN - FTP server
17.01.4 - 16.9 MiB/sec (135.2 mbps)

FTP client - LAN - HH5a - WAN - FTP server
17.01.0 - 8.2 MiB/sec (65.6 mbps)
17.01.2 - 8.5 MiB/sec (68 mbps)
17.01.4 - 13.3 MiB/sec (106.4 mbps)
r5597 - 10 MiB/sec (80 mbps) – snapshot, SMP-enabled

FTP client - 5 GHz wifi (20MHz) - HH5a - WAN - FTP server
17.01.0 - 6.6 MiB/sec (52.8 mbps)
17.01.2 - 6.5 MiB/sec (52 mbps)
17.01.4 - 6.5 MiB/sec (52 mbps)
r5597 - 7.6 MiB/sec (60.8 mbps) – snapshot, SMP-enabled

FTP client - 5 GHz wifi (40MHz) - HH5a - WAN - FTP server
17.01.0 - 6.5 MiB/sec (52 mbps)
17.01.2 - 6.8 MiB/sec (54.4 mbps)
17.01.4 - 6.5 MiB/sec (52 mbps)

The FTP download speeds via 5 GHz wifi do appear to max out around 52 mbps during my tests when routed through WAN interface, which would seem to support your own observations. SMP seems to offer a bit of a boost to 60+ mbps.

Strangely, my Apple TV 3 gets full speed on 5ghz, my Samsung S7 edge doesn't, my hp laptop doesn't
I am.using the ookla speedtest app on all of them using the Vodafone server

I tend not to trust speedtest.net on Android as I have seen it show both vastly slower and vastly faster than real-world performance.

That said, I do think Android tends to struggle to reach my full speed compared to PCs, but then how often do you NEED 80Mbit on a phone?

I agree, it's all over the place at times, but generally it's consistent if I use the same server on the devices.
It's just strange that I used to get full bandwidth on the phone during a test and everywhere else but not any more, but the ATV does!! Strange

The relationship between WiFi and the Internet is just a weird thing, I think various latencies and packet transmission methods just creates all sorts of bottlenecks that aren't always obvious.

For example on my phone I might get 400Mbit over iperf3 to the LAN but struggle to hit 50Mbit over the Internet via WiFi, yet get higher than that over 4G. Its all rather unpredictable on a day to day basis.

This is why people defaulting to WiFi even where wired is a viable option is a really bad thing. WiFi is such a complicated beast its best to only use it when you really REALLY have to.