With our updated VeriWave test set in place and a new standardized MU-MIMO test process, I went back and retested all the MU-MIMO enabled routers that have crossed the SmallNetBuilder test bench to date. The only exception is Amped Wireless' Athena. I didn't have one on hand and decided to not request one. Amped announced its redesigned replacement, the ATHENA-R2 at CES 2016, which is scheduled to start shipping any day now. So we'll just test the new one when Amped sends it for review.
I checked for new firmware before each router was tested. The products and firmware versions in this test are compiled in Table 1.
|TP-LINK Archer C2600||AC2600||QCA9980||No change||1.1.2 Build 20150924 Rel. 66045|
|TRENDnet TEW-827DRU||AC2600||QCA9980||No change||1.00b011|
Table 1: Routers and firmware versions
Keep in mind all products tested are based on Qualcomm Atheros MU-MIMO radios. Table 1 shows three different QCA MU-MIMO radios used: the original 4x4 QCA9980; latest generation 4x4 QCA9984 and 3x3 QCA9982.
Products using Broadcom MU-MIMO capable radios have only "beta" level firmware released at this time. I'm waiting for released firmware before adding Broadcom-based MU-MIMO routers to the Router Charts. But I did test an ASUS RT-AC88U with beta MU-MIMO firmware that I'll show the results for later on.
The primary benefit of MU-MIMO is higher total throughput when multiple MU-MIMO devices are in use. The first chart shows a 2-to-1 difference between best and worst performing products for maximum MU-MIMO throughput gain.
MU-MIMO throughput - maximum
The 802.11ac standard supports up to four clients in a single AP-to-STA transmission. In practice, today's 4x4 AP and router designs support only three STAs per transmission. The diagram taken from Is MU-MIMO Ready For Prime Time? shows what happens with four MU-MIMO STAs. The second transmit frame swaps out MU STA #1 and swaps in MU-STA #4. STA #1 then gets serviced in another transmission as a normal (SU) STA.
Four STAs - MU
As more MU STAs are added, each must be swapped in and out. Ideally, the maximum throughput reached with three MU STAs would continue indefinitely. But in reality, the overhead of efficiently scheduling MU and SU airtime takes its toll and throughput decreases.
The plot below compares the NETGEAR R7800, which had the highest total throughput gain with the TP-LINK Archer C2600, which had the lowest. Also included is the Linksys EA7500, the first 3x3 router supporting MU-MIMO. The NETGEAR yielded peak throughput of 807 Mbps with three MU-MIMO STAs, as expected. But the TP-LINK didn't produce its lower peak throughput (390 Mbps) until six MU STAs were in use.
MU-MIMO Throughput vs. STA
The 3x3 Linksys EA7500 is interesting. Total throughput gain peaks with two devices at 462 Mbps, slightly better than half the NETGEAR R7800's total MU-MIMO gain. Four devices drop throughput back down below the single device starting point, which then jumps back up with five devices. Past that, results were obtained only with 6, 8 and 14 devices. In the missing tests, all MU-MIMO STAs were able to connect, but would not pass the test TCP/IP traffic.
Without the benefit of MU-MIMO, routers should share available throughput evenly among all connected clients, assuming ideal conditions. Our test makes conditions as ideal as possible, using a cabled environment and identical simulated STAs. So the SU (single user) MIMO tests are a good indication of how efficent routers are at serving multiple STAs.
The plot below shows average SU-MIMO throughput for the six tested routers. Hmmm, what's up with the TP-LINK?
SU-MIMO throughput - average
The next plot again shows the TP-LINK Archer C2600, NETGEAR R7800 and Linksys EA7500. Of these three, the NETGEAR is the only one showing proper SU behavior. The Linksys would not pass traffic beyond 6 STAs and the TP-LINK not only starts out with less than half the proper throughput, but also starts significantly declining after 5 STAs. Throughput then falls to single digits beyond 12 STAs.
SU-MIMO Throughput vs. STA
Pulling the TP-LINK plot from Is MU-MIMO Ready For Prime Time? shows SU performance was also problematic then.