Posted: Thu Jun 02, 2016 18:13 Post subject: Accurate Wireless Speed Tests
I am a novice at networking and would like to know how are those with more networking experiences accurately test their wireless speed?
My setup is a Netgear R7000 running Kong's DD-WRT v3.0-r29620M in AP mode. The 2.4GHz radio is turned off and the 5 GHz is configured as shown in the attached file. My main router is an Intel Atom processor C2758 running pfSense. My WAN connection is symmetrical gigabit fiber.
The fastest wireless devices I have is a Samsung Galaxy S5 802.11ac 2x2 (866Mbps) and then an 802.11n 2x2 (300Mbps) laptop. On the S5, I use the SpeedTest app and my downloads max at 170Mbps with uploads at 300Mbps. On the laptop, using the SpeedTest browser my download max at 170Mbps with upload max at 200MBps. The status table always shows the RX and TX for the S5 to be 866M and for the laptop to be 300M.
It seems for some reason my wireless downloads are hitting a ceiling of 170Mbps. So I decided to run som LAN->WLAN testing with iperf3.
I launched iperf3 on the AP and from my laptop I performed a test with the defaults. Again 170Mbps downloads and 200Mbps uploads. I installed Magic Iperf app on the S5 and ran with the defaults and the download was around 170Mbps. Interesting.
So I then decided to up the TCP window size to 1024k and ran 8 parallel connections. This is where I started to see more promises. On the 802.11ac 2x2 client, iperf was reporting downloads and uploads of about 320Mbps.
So is using iperf more accurate? Doesn't the SpeedTest app saturate the bandwidth? Is something else going on?
_________________ Supermicro A1SRi-2758f pfSense 126.96.36.199 as primary router
Netgear R7000 Nighthawk AC1900 kongac V3.0-r29620M (OC to 1200MHz) in AP mode
I'm using the Tamos Throughput Test. It evaluates network perfomance from the application point of view. I.e. you can see actual speeds that can be achieved by software on your computer (with protocol overheads taken into account).
The tool consist of two parts client/server. The server should be installed on a PC or MAC, connected to the net by wire. The second part installs on a wireless client for testing and enjoing (or not) the results
I've added an example of WiFi stress test running on my macbook.
Most online flash based speedtests use a .jpg image, choosing which sized one to use based on
your initial connection speed. Images range from hundreds to many thousands of pixels and
most actually look like this ¬
Most folks don't have 30 ~ 100+MB jpg handy --- mp4 works well too