sfx2000
Part of the Furniture
VHT (802.11ac) in 2.4GHz is the basis for many vendor claims of AC1900 and higher performance.
It's been discussed in other threads that VHT/TurboQAM type connections are a challenge in the real world - I agree, I've been able to do this on a lab bench, but it's rare to see in real-world usage.
This post essentially supports those marketing claims, however, it's not the best setup for 2.4GHz - I stand by my position that VHT/TurbQAM/Up to 400,600Mbps modes are not appropriate for the 2.4GHz band - it does add a lot of interop concerns, and with many devices, actually has a negative impact on performance (also confirmed subjectively in this test).
Best settings for 2.4Ghz are still generally B/G/N mixed, 20MHz channels (one can try 20/40 auto) for best client performance...
Specifics... again, the objective was to see if one can actually get a non-standard connection in 2.4GHz with VHT mode - and yes, it does...
DUT - Dell Inspiron 15 5000 series (2015) - Windows 10 Build 1803 - internal Dell 1705 ath9k WiFi disabled in BIOS
Capture Device - Macbook Air 11" 2014 edition, OSX 10.13.6, Airtool 1.7, Wireshark 2.6.2
Observations - yes, while we were able to capture a VHT connection in 2.4GHz, there were a number of overlapping BSS's (other neighboring AP's out of my control) in the general area... the connection was established, but unstable at the network layer - max network throughput at the application level was less than 1 Mb/Sec (confirmed via iperf3/speedtest.net/fast.com/dslreports.com speedtests)
Resetting the Asus/Realtek drivers on the Laptop under test - 802.11n connection was established, to get 802.11ac 5GHz, one has to go into the driver advanced settings and set Preferred Band to 5GHz First, and there everything works as expected.
Attached is the Association Request and Response between the AP and Client under test. It's an interesting example, as this is a cross-vendor test - Marvell/Realtek - many others might be same vendor - e.g. Broadcom or QCA for both Client and AP.
Fun things in the PCAP's attached...
Look at the Marvell's HT capabilities - it really, really wants to beam form for 11n - it adverts pretty much everything in the 11n standard - you typically don't see this with Broadcom or QCA for Consumer AP's - I do see this from time to time in 11n for enterprise, but beamforming and spatial multiplexing in 11n are in conflict - do one or the other - and most chose correctly for SM vs. Beamforming (implicit or explicit) - one gets better returns on SM vs. Beamforming there...
The WRT's VHT Operation - MCS7 - which is appropriate perhaps, as the AP is set for 20/40 mode, and VHT7 is as high as you can get for 20MHz channels in 11ac - notice also in HT mode that it supports MCS32, HT-Duplicate, again, uncommon in most consumer AP's
The Realtek Client STA - it does not support by default SU beamforming (which is an optional feature actually in 11ac) - it does say that it's operating in 40MHz mode for CH6 in the association request...
Just kinda cool to catch this - most clients in 2.4GHz - while the chips might be capable of VHT, most drivers disable it - and for good reasons, IMHO...
FWIW - the caps show an VHT MCS7 operation - 2 streams, 40MHz = 300 Mbps, which is pretty much what one would expect with 11n in the same conditions - HT MCS 15 - this keeps things honest actually at the RF level...
Some sharp eyes might notice in the attach response that there's no RSN Info stanza in the PCAP - and that's because auth has already been completed, so it's implied, and not needed since no 11r support.
Also note that in the Attach Response the WMM - the AP defines the QoS relationship on the AP-STA link for 802.11, the client can request, but the AP grants what it's configured for.
It's been discussed in other threads that VHT/TurboQAM type connections are a challenge in the real world - I agree, I've been able to do this on a lab bench, but it's rare to see in real-world usage.
This post essentially supports those marketing claims, however, it's not the best setup for 2.4GHz - I stand by my position that VHT/TurbQAM/Up to 400,600Mbps modes are not appropriate for the 2.4GHz band - it does add a lot of interop concerns, and with many devices, actually has a negative impact on performance (also confirmed subjectively in this test).
Best settings for 2.4Ghz are still generally B/G/N mixed, 20MHz channels (one can try 20/40 auto) for best client performance...
Specifics... again, the objective was to see if one can actually get a non-standard connection in 2.4GHz with VHT mode - and yes, it does...
- AP - Linksys WRT-1900acV2 (Marvell based obviously)
- STA - Asus USB-AC56 (Realtek RTL8812AU) connected via USB3.0
- Upstream WAN is 150Mbps down/10 Mbps up - CoxHSI, Cable Modem Moto SB6183
DUT - Dell Inspiron 15 5000 series (2015) - Windows 10 Build 1803 - internal Dell 1705 ath9k WiFi disabled in BIOS
- Asus/Realtek Driver - default from Windows Update - v 1030.11.503.2016 - correctly ID's as the Asus USB-AC56 device
- Linksys WRT1900acV2 Firmware - v 2.0.8.187766
- VHT 2.4G - Not Support Broadcom Vendor (needed for Marvell based AP) - default is disabled - since this is a very early 802.11ac Wave 1 device, Broadcom had their way of supporting TurboQAM that Marvell/QCA, and other vendors did not
- Preferred Band - 2.4G First
Capture Device - Macbook Air 11" 2014 edition, OSX 10.13.6, Airtool 1.7, Wireshark 2.6.2
Observations - yes, while we were able to capture a VHT connection in 2.4GHz, there were a number of overlapping BSS's (other neighboring AP's out of my control) in the general area... the connection was established, but unstable at the network layer - max network throughput at the application level was less than 1 Mb/Sec (confirmed via iperf3/speedtest.net/fast.com/dslreports.com speedtests)
Resetting the Asus/Realtek drivers on the Laptop under test - 802.11n connection was established, to get 802.11ac 5GHz, one has to go into the driver advanced settings and set Preferred Band to 5GHz First, and there everything works as expected.
Attached is the Association Request and Response between the AP and Client under test. It's an interesting example, as this is a cross-vendor test - Marvell/Realtek - many others might be same vendor - e.g. Broadcom or QCA for both Client and AP.
Fun things in the PCAP's attached...
Look at the Marvell's HT capabilities - it really, really wants to beam form for 11n - it adverts pretty much everything in the 11n standard - you typically don't see this with Broadcom or QCA for Consumer AP's - I do see this from time to time in 11n for enterprise, but beamforming and spatial multiplexing in 11n are in conflict - do one or the other - and most chose correctly for SM vs. Beamforming (implicit or explicit) - one gets better returns on SM vs. Beamforming there...
The WRT's VHT Operation - MCS7 - which is appropriate perhaps, as the AP is set for 20/40 mode, and VHT7 is as high as you can get for 20MHz channels in 11ac - notice also in HT mode that it supports MCS32, HT-Duplicate, again, uncommon in most consumer AP's
The Realtek Client STA - it does not support by default SU beamforming (which is an optional feature actually in 11ac) - it does say that it's operating in 40MHz mode for CH6 in the association request...
Just kinda cool to catch this - most clients in 2.4GHz - while the chips might be capable of VHT, most drivers disable it - and for good reasons, IMHO...
FWIW - the caps show an VHT MCS7 operation - 2 streams, 40MHz = 300 Mbps, which is pretty much what one would expect with 11n in the same conditions - HT MCS 15 - this keeps things honest actually at the RF level...
Some sharp eyes might notice in the attach response that there's no RSN Info stanza in the PCAP - and that's because auth has already been completed, so it's implied, and not needed since no 11r support.
Also note that in the Attach Response the WMM - the AP defines the QoS relationship on the AP-STA link for 802.11, the client can request, but the AP grants what it's configured for.
Attachments
Last edited: