said by owlyn:
WiFi results are not reliable indicators of your actual speed. There are troo many variables. I recently had to change my WiFi channel due to interference. A neighbor apparently got a new router and was using the same channel as I was. After changing the channel, speed and reliability went back to normal.
Testing here in a clean RF environment within a faraday cage lab environment using 2:2:2 configuration with various enterprise Cisco APs. Associated rate is 300Mbps with RSSI of -29dB and lower. The WiFi connection is as ideal as you can get for 5Ghz and is more than capable of testing against a 50/10 cable Internet connection. The downstream comes out at 50mbps+/- and 2Mbps+/- upstream when running ShaperProbe on Windows host machines configured with an Intel WiFi chipset radio. Mind you it does not do this to internal or external iperf servers.
In addition using any number of Macs connected via WiFi with ShaperProbe results in ideal 50/10 results, nor does the referred slow upload results occur when using other WiFi chipset radios such as those from Atheros, Broadcom or RALink.
There is definitely something happening with the use of ShaperProbe in combination with Intel WiFi chipset radios on Windows boxes. The end result is poor estimated capacity and shaping rate for the upload.