Four Stream Performance
TP-LINK supplied two C2600s so 4x4 thoughput could be tested. Two RTA2600's were set up in open air, six feet apart, one as a normal router and the other in wireless bridge mode (found in Advanced > System Tools > System Parameters). The in-house 5 GHz network was idle and only beaconing (no traffic). Channel was set to 153, bandwidth mode was in its default 80 MHz mode and everything else was set to defaults, including MU-MIMO enabled. Given the 4x4-to-4x4 nature of the link, MU-MIMO doesn't come into play.
I connected only one computer via Gigabit Ethernet at each end of the bridge. Baseline tests using Ethernet between the two computers (both coincidentally equipped with TP-LINK TG-3468 NICs) show the Ethernet link capable of a bit over 940 Mbps in both directions with each direction run separately. So the hardwired part of the link shouldn't be a limiting factor. I couldn't monitor link rates for these tests, since the router's admin GUI doesn't show them.
The simultaneous up and downlink test yielded 798 Mbps total throughput, with an almost 3-to-1 difference between down and uplink throughput as the IxChariot plot shows. The Linksys EA8500 yielded 830 Mbps for this test, with more battling between up and downlink.
TP-LINK Archer C2600 four stream throughput - simultaneous up/downlink
Separate up and downlink tests that yielded 739 and 652 Mbps vs. 727 Mbps downlink and 667 Mbps for the EA8500. The plot shows very steady throughput in each direction.
TP-LINK Archer C2600 four stream throughput - up and downlink
A second pair of clients was plugged into the router and bridge to see if a second connection pair would yield higher throughput and to ensure that the single Gigabit Ethernet connection was not limiting results. The plot shows a total 758 Mbps throughput, a bit less than the single-pair simultaneous uplink / downlink test above.
The significantly lower throughput of the second computer pair was suspicious and consistent on multiple runs. But a check with the two directly connected via Gigabit Ethernet yielded 786 Mbps downlink and 652 Mbps uplink. Both computers also reported Gigabit links when plugged into the each router. So I have no explanation for why the second pair of computers had significantly lower throughput.
TP-LINK Archer C2600 four stream throughput - simultaneous up/downlink, four computers
As an alternative, I used the original computer pair and added more connections for a total of eight. This raised total throughput to 933 Mbps downlink...
TP-LINK Archer C2600 four stream throughput - downlink 8 connections
...and 863 Mbps uplink.
TP-LINK Archer C2600 four stream throughput - uplink 8 connections
Our evolving MU-MIMO test process uses open-air testing with three vendor-supplied clients. TP-LINK sent three Xiaomi Mi Note Pro smartphones (available only in China), perhaps the same ones Amped Wireless provided to test its RTA2600.
Xiomi Mi Note Pro MU-MIMO enabled smartphone
MU-MIMO's improves 5 GHz downlink-only total bandwidth utilization by letting up to three devices share the same airtime (in its current implementation in 4x4 routers). So the test is simple:
- Disable MU-MIMO, connect three 1x1 clients, run a test to all three clients simultaneously, record each client throughput and calculate total throughput
- Repeat the test with MU-MIMO enabled
- Calculate the throughput differences
The same test conditions applied to the Linksys EA8500 were used. All tests used the IxChariot throughput.scr script with the test file size set to 5,000,000 Bytes, TCP/IP protocol, one connection per client, 90 second test length. Each test was run once. The in-house 5 GHz network was idle and beaconing (no traffic). Channel was set to 153, bandwidth mode was in its default 80 MHz mode and everything else was set to defaults.
In the sections to follow, all plots show downlink throughput only, since that is the only direction that MU-MIMO works in. Just to be sure, a few uplink tests were run and saw no significant throughput gain with MU-MIMO enabled was observed.