@RTouris, usually UDP is used to measure WiFi throughput, but the built-in speed test of CPEs seems to be TCP in my opinion.

 

UDP gives you the maximum WiFi throughput, which should nearly match the speed provided by the WiFi layer (e.g. N300). TCP gives you the throughput for typical data over, which in an ideal environment is around 70% of the WiFi speed in 802.11n mode on a CPE according to my experience (e.g. 214 Mbps for CPE/WBS). The former is the WiFi throughput, while the latter is the goodput for the device being tested.

 

What results do you achieve with the built-in speed test on a link?

@R1D2 On a newly established link serving a 100mbps DSL line over 1.5km with two CPE510s I get ~30dB of SNR over a channel width of 20Mhz with MCS maxing at 144mbps, PharOS speedtest gives me ~70mbps both up/down, direct downloads max out @ 30mbps down tops :/ 

 

Alt. vendor on same settings maxes out ~130mbps over UDP / ~100mbps TCP and indeed reaches ~90mbps on speedtest sessions...I just don't get it w/ TP-Link nowadays?

Hi @RTouris,

 

did you use iperf or a web-based speed test?

 

In my experience only iperf gives reproducable results. If using speed tests on the web, make sure to always use the same test server (which otherwise is selected automatically) and repeat tests at least a few times.

 

Web-based speed tests vary heavily depending on the server selected automatically and the actual network load. dslreports.com is mostly very accurate, but an iperf test server running in the LAN is almost always accurate. DSL lines also don't guarantee a constant throughput.

 

Also consider the fact that modern APs use different antenna designs those days and Pharos CPE designs are a bit older.

@R1D2 I'm using a synthesis of speed tests, both iperf based off a server that operates at AP's location and an average of speedtest.net results from servers in the local area where there's no in-between hops, connections abroad etc. What puzzles me is the shear difference between the TP-Link's built-in UDP results and the actual throughput compared to the alt. vendor....Granted these are sector antennas compared to parabolic PtP antennas used in the alternative deployment, but given the comparable achieved SNRs and maxing of the MCSs I fail to substantiate the difference, it just makes little to no-sense...

 

In an attempt to remedy some such ill-effects in the past I tried deploying some CPE610s, which combined with CPE510 didn't give me the results I expected. More specifically I would get the extra bonus of the added dBi on the CPE610 which lead to a stable downstream (CPE610 set as the CL), however for some reason the signal from the CPE610 towards the CPE510 wouldn't get through properly in three different cases I tried and would subsequently lead to deteriorating upstream, so I just gave up using that combo. :/

Hi @RTouris,

 

when referring to iperf, I mean the standalone utility, not the CPE's built-in iperf.

 

But as I said, antenna designs have greatly improved in last 5 years, so even with standalone iperf you could probably get similar results (but at least accurate results).

 

BTW: speedtest.net is the most unreliable bandwidth test I ever saw:

 

with regards to iperf, yes it's the standalone terminal executable on a server / client setup which tests both TCP and UDP according to set flags. My concerns lay with the difference between the two setups given that the link appears to be of equal magnitudes with both vendors. I can't blame speedtest.net for the variance, it's definitely 100% the link and the antenna tech that's inherently different here which makes the difference. Granted the CPE510 can act as a PtmP AP and doesn't require a L4 "license" to do so, but given the appearances and given facts (vendor's gear is equally if not even cheaper to acquire) I'd be eager to start weighing my options with regards to aging TP-Link gear... :/