The "All Front" test is meant to be a MU-MIMO torture test. It positions all three laptops on a rack in front of the router about 6 feet away as shown in the photo below. (The router looks closer than it is. And yes, the SNB test labs are quite posh...) Because the clients are so physically close, it's hard to optimize the formed beams for each one.
MU-MIMO All Front test setup
Nevertheless, the plot below shows essentially the same total throughput gain (32% vs. 31%) as the "optimal" triangle setup.
MU-MIMO Throughput change - All Front
Three Room A
I next wanted to see how MU-MIMO would work in a more real-world home situation with clients in different rooms. This test kept one laptop on the rack in the same room as the EA8500, the second laptop moved to a nearby bedroom about 25 feet and a few sheetrock walls away and the third was located in my office down the hall. This location (3 room A) was intentionally tough, about 50 feet away and a diagonal path through at least four sheetrock walls.
MU-MIMO 3 Room test floorplan
In the end, it almost proved too much for the office-located notebook, which slowed to a crawl with both MU-MIMO on and off. The take-away from this part of the test is that MU-MIMO won't help improve effective range, i.e. better throughput at a given location, if signal levels are low. But the test still yielded a 21% total throughput improvement for the other two clients.
MU-MIMO Throughput change - 3 room A
Three Room B
I tried a second multi-room test, this time moving the test notebook to a room adjacent to the bedroom the second laptop was relocated to (3 Room B on the floorplan above). The laptop sat about 35 feet and three walls away from the router. The results were interesting, with only the In Room laptop (with the strongest signal) showing significant throughput improvement (9%), but not as much as in previous tests. Total throughput improvement in this test was only 3%. Since this is within the variation I saw in uplink tests, which aren't affected by MU-MIMO, I call the result for this test as no improvement.
MU-MIMO Throughput change - 3 Room B
Three Room B - Run 2
I wanted to make sure I handn't made a setup mistake, so I reran the test, first putting the router through a MU-MIMO enable/disable cycle. This time, the In Room laptop throughput increased 72%, but the other two laptops' throughput decreased significantly more than before. The net result this time was an 11% throughput gain, significant, but not that impressive.
MU-MIMO Throughput change - 3 Room B - Run 2
The IxChariot plot below shows the test run throughput plot with MU-MIMO disabled.
MU-MIMO Throughput - 3 Room B - Run 2 configuration - MU-MIMO disabled
And with MU-MIMO enabled. With lower and unequal signal levels, the MU-MIMO algorithms have a tougher time reducing thoughput variation and equally distributing it.
MU-MIMO Throughput - 3 Room B - Run 2 configuration - MU-MIMO enabled
All Front w/ 2x2 STA
My last test threw a different curve ball, this time adding a 2x2 non MU-MIMO adapter into the mix. I moved all laptops back to the rack in the room with the EA8500, then disabled the QCA adapter in one of them. I then installed a NETGEAR A6200 2x2 USB client, connected everything and ran the tests. With two 1x1 clients still running MU-MIMO, we should expect to see improved throughput for the 2x2 client. Since MU-MIMO can move packets for both 1x1 clients in the same frame, more airtime is freed for the 2x2 client.
The results yielded a meh 9% total throughput gain, although the 2x2 adapter's individual throughput more than doubled!
MU-MIMO Throughput change - All In Front w/ 2x2 STA
I ran the test again the next day and got an 18% total throughput gain with the 2x2 adapter again getting the best individual improvement of 70%.
MU-MIMO Throughput change - All In Front w/ 2x2 STA - Run 2
The IxChariot plot below shows the test run throughput plot with MU-MIMO disabled. It's clear the 2x2 client gets the smallest throughput allocation, as you would expect with two 1x1 clients taking up more airtime.
MU-MIMO Throughput - All In Front w/ 2x2 STA - Run 2 configuration - MU-MIMO disabled
With MU-MIMO enabled, the 2x2 client's throughput increases dramatically, equalling both 1x1 clients' during the first part of the test run.