Router Speed tests

[caeci11ius]

I did a bit of this when I got my new VDSL2 connection, which was why I started looking into better performing routing and overclocking of my WNDR3700s (you can find the patch here if you're interested: viewtopic.php?f=7&t=7169).

You don't want to be testing with traffic sourced from any kind of disk - random traffic generated by the CPU into memory is fine. You can use a free tool like iperf (get it here: https://iperf.fr/ - source also available) to do it, and I'd suggest using a common method if you want to develop any kind of test database.

SmallNetBuilder did a nice article on this a few years back: http://www.smallnetbuilder.com/wireless ... ess-router ... you can use that as a starting point/setup guide. Basic summary though is you want to set up two hosts running your test tool, with one on the WAN and one on the LAN side. You need to set manual IPs and ensure routing between the two hosts is working, then connect your test tool through the routed connection and have it smash out a bunch of random I/O. If your ports are only 100mbit, you'd want to probably test up/down simultaneously to generate as much routed traffic as possible...

[allio]

Just to add a second result for the TL-WR1043ND 1.x on Gargoyle 1.9.0:

Upload: 116.98 Mbps, Download: 118.59 Mbps

With QoS turned on it couldn't break 20 Mbps.

I'm about to upgrade to 100/20 fibre, so it'll be interesting to see whether the old TP-link can handle something like a line-rate torrent without choking. I suspect that real world use will prove a lot more challenging than a single file read/write, and that an upgrade may be in my near future...

[pjablo]

[quote="allio"]Just to add a second result for the TL-WR1043ND 1.x on Gargoyle 1.9.0:

> Upload: 116.98 Mbps, Download: 118.59 Mbps

> With QoS turned on it couldn't break 20 Mbps.

First thing when you configure QoS is to declare a little less than your download speed in order to limit it on router and make ingress qos possible - I think your current download speed is about 20-25 Mbps so this limit works.

I have a WDR3600 and fast internet 200 Mbps - on LAN 1 Gps without firewall/qos rules max it is about 180 Mbps, with fw rules/limits - about 80-90 Mbps, on Wifi 5 Ghz router -<-> PC iperf test - about 95 Mbps, but real internet test on wifi - only 70-80 Mbps max.

I think it will be impossible to achieve max speed with qos/firewall rules on your TL-WR1043ND - which is slower than mine WDR3600

[Chilling_Silence]

In my testing:
TL-WR1043NDv2 gets ~140mbps happily, up to 160mbps in short bursts
TL-WR941ND gets ~55mbps happily, up to 60mbps in short bursts

The TL-WR1043NDv2 will do more when it's not NAT'ing but that's not much good.

Tests were done with a speedtest.net mini server through xmpplite on a Windows Laptop with gig ethernet to the WAN port, being run on a PC on the LAN (also gig ethernet).

I know the Archer C7 gets faster but I don't recall the values.

[Hercules]

I have a TL-WDR4300 currently running Gargoyle 1.5.10

In the next few months my house is planned to receive a faster connection and I will probably pay for a 50/20 service.

On another forum there was some scoffing (asus fanbois!) that my router won't cope with those speeds so just checking here whether anyone is using this router and has some experience as to whether it will cope ok?

Thanks all!

[ispyisail]

could be on the edge?

It also depends how much stuff your running e.g. QOS, Torrents etc etc

Only one way to find out

[Hercules]

Thanks. I have upgraded to 1.9.2 so dont know if that will improve things or not. As you say will see what happens in a few months.

[ispyisail]

1.9.x is more stable again (end of cycle)

[ektus]

iperf testing, setup:
iperf client: Windows 7 X64, Gigabit Ethernet connection
gigabit-switch
Archer C7 V2 LAN <-> WAN running Gargoyle 1.10.0
iperf server: Fritz!Box 7490

Code: Select all

iperf.exe -c 192.168.1.9 -p 4711
------------------------------------------------------------
Client connecting to 192.168.1.9, TCP port 4711
TCP window size: 8.00 KByte (default)
------------------------------------------------------------
[188] local 192.168.0.22 port 65248 connected with 192.168.1.9 port 4711
[ ID] Interval       Transfer     Bandwidth
[188]  0.0-10.0 sec   181 MBytes   152 Mbits/sec

Regards
Ektus.

[encro]

Have we got something like a CPU base line expectation when QoS is enabled to ensure no lapse in throughput?

For example: for every 1 Mbps throughput you would need 8.5 MHz of CPU
For 25 Mbps you'd need a 212 MHz CPU or greater (25 x 8.5 =212)
For 50 Mbps you'd need a 425 MHz CPU or greater etc
For 75 Mbps you'd need a 637 MHz CPU or greater etc
For 85 Mbps you'd need a 722 MHz CPU or greater etc
For 100 Mbps you'd need a 850 MHz CPU or greater etc
This is just a guess at approximating the clock speed needed, most likely doesn't represent a real world figure.