Faster wifi with 1043nd V1

[d3fz]

Interesting topic, looking forward to the results from goyle100.

[Lantis]

It is a popular model (doesn’t mean it’s not getting long in the tooth).
I can only base it off the Polish statistics, where it is the 3rd most prevalent router.

Our users weren’t interested in anonymous statistics.
http://dl.eko.one.pl/stat.html

[kabi]

So I did some more investigation around strange HT40 usability.
Here are some more notes. I'm now running some latest 'beta' Gargoyle build from march.

1.) Whenever txantenna option is specified - it MUST be also given with rxantenna option. Valid setting for rxantenna is only '7'. For txantenna values 1..7 can be selected. When just only one option is given OR invalid value is used, these settings are simply ignored and any previous setting remains set in driver (that could have influenced my previous tests actually) - so this behavior is a bit unexpected.

2.) Since acceptance/usability of some WiFi channels looked strange to me, I've checked capacitors and one seemed a bit bloated so I've replaced it - but there are lot of them without any 'blowing part' in the middle so it would be probably hard to recognize if they are bad and I'm not going to replace them all

Anyway after replacement it seemed to be still same having 'weird' HT40 issues, but maybe channel usage seemed to improve a bit.

Further checking for possible trouble making options - it looks like dropping everywhere suggested lines from wireless config:

• list ht_capab 'SHORT-GI-40'
  list ht_capab 'DSSS_CCK-40'

actually made Windows boxes on the network happier.


But the most revealing part is probably that physically disconnecting 'the middle' antenna and using just 2 corner ones made it actually faster! With this setup and HT40 setting I've been able to reach over 100mbps in the single room mainly when the machine was 'alone' and has not shared networks with other devices. On the other hand 'older' devices then looked running slower (like just 8mbps). So I guess using HT20 with 3 antennas in place is reasonable compromise where 'fastest' boxes gets ~70mbps, middle level oldish phones ~30mbps and G devices are still over 18mbps.

ATM I'm not sure if 'the middle' antenna issue is a sign of approaching HW problem (it's aged device used 24x7) or it's 'by-design' how to serve '11n' only devices that fast - but in combined world it's way more limited. But still need to do more testings - now with better knowledge how those rx/txantenna options do behave and 2 or 3 antennas attached - lots of combination to benchmark so when one doesn't want to annoy home users it will take a while...

So yeah, probably time to think also about its replacement and 'Archer C7 ac1750 v2' actually does look quite interesting.

[kabi]

...there is no need for the GUI to expose it. The average user has no need to see the back end configuration.
For advanced user, it is there to use.

You should only use the country code that you are in. Using any other country code would be a violation of the relevant regulations for your country.
“The good neighbour policy”

When you actually mentioned this - where can the user in Gargoyle GUI actually set the proper country - to be a good neighbor ?

[Lantis]

They can't (my initial patch for this was rejected, i've been working on a version 2 with the requested changes for almost a year).

By default, Gargoyle is compliant globally (i.e. the minimum settings legal in ALL countries are adopted).


Please don't quote me out of context or mix and match my quotes. Those two quotes don't go together. The good neighbour policy refers to 20/40MHz channel coexistence in the 2.4GHz band.

[ispyisail]

So yeah, probably time to think also about its replacement and 'Archer C7 ac1750 v2' actually does look quite interesting.

This router is also starting to get dated

[kabi]

So yeah, probably time to think also about its replacement and 'Archer C7 ac1750 v2' actually does look quite interesting.

This router is also starting to get dated

So what is these days 'not dated' and doesn't cost a fortune and has 4 gigabits ethernets and 2.4+5GHz and is known to work WELL with openwrt ?

Next question - since you are building those Beta Gargoyle builds - I'm looking at why my router can't keep with traffic while purely mathematically it should have enough resources to handle my 150Mbps even with packet accounting - are there any 'measurement' of impact of DEBUG options used for compiled kernel - there is simply way too much debug info present in /sys/debug - it all has some cost to collect that much info - I'll probably have to try to create build environment myself to check this myself in details - but just asking ahead if there are some known stats about this ? (knowing myself heavy debug may easily cut away large portion of bandwidth in storage world...)

[Lantis]

Plenty of google results for why your third party router firmware can't route enough traffic.

It boils down to CPU speed, and the fact that third party firmware can't utilise hardware NAT.
Add bandwidth monitoring and QOS etc to the responsibilities for the CPU to handle, and it starts to struggle.

Removing kernel debugging options decreases image size in general. I do not know of anyone who has measured performance differences with these options.

[ispyisail]

So what is these days 'not dated' and doesn't cost a fortune and has 4 gigabits ethernets and 2.4+5GHz and is known to work WELL with openwrt ?

yeah........

Gargoyle is really only a GUI for OpenWRT.

OpenWRT and LEDE had a split and we got stuck with OpenWRT CC 15.05. That version is now 2 -3 years old? Only routers built in that time period work. OpenWRT and LEDE have re-merged and we are currently in the tested phase for a new base.

New routers will become available very soon

At the the current time the best routers for any speeds above 50mbps +/-

WRT1200AC v1 (Caiman)
WRT1200AC v2 (Caiman)
WRT1900AC v1 (Mamba)
WRT1900AC v2 (Cobra)
WRT1900ACS v1 (Shelby)
WRT1900ACS v2 (Shelby)
WRT3200ACM v1 (Rango)

With Openwrt/Gargoyle CPU power is king with fast internet connections

[kabi]

So I did some benchmarking - I've tested several firmwares and here are some comments. When testing on local network only where 1 box is connected on 1Gbps LAN and other is some fast Lenovo WiFi box and router was 'carefully' positioned in short distance to achieve best speed and using iperf.

• - TPLink official firmware 3.13.15 - give at best ~170Mbps.
  - Some recent beta DDWrt ~125Mbps
  - Gargoyle with disabled bwmon ~ 115Mbps
  - Gargoyle with enabled bwmon is limited to 100Mbps - not quite sure why as since CPU doesn't look all that busy...

And here are few more notes - my router really looks like some hw is not doing the right work - since removal of middle antenna makes it usable - with middle antenna WiFi connection is dropping and slow - I can just guess that running the router all the time with highest TX power 'slowly kills' radio components of this device.

Checking txantenna 1, 2, 4 each individually clearly shows all antennas can work and can stream traffic alone - so likely using 7 can give maximum 'power' of beam - but on my router this leads to major trouble as WiFi becomes slow - disconnecting middle antenna makes it way more stable (or 'txantenna 5')

Configuring WiFi with 'HT40' and 'noscan' likely gives the best speed with n-devices - but if the network is also composed of g-devices - they are working poorly - so dropping to HT20 and keeping defaults looks like g-devices are way more usable - while n-devices are still quite good.

Iperf speed testing surely shows those max speed can't be kept all the time and speed oscillates quite a lot with HT40 - and seems to be way more 'stable' with HT20.

Placement of antennas has major impact on resulting max speed - especially if the goal is to get big range coverage - using corner antennas set this way:

And I've also noticed DDWrt has BUG and for antenna settings '1+2' and '1+3' is using same number '3' for txantenna option.