Maximum Wireless Throughput
Since throughput vs. attenuation plots are now done with a 2x2 STA, we use the Ixia Veriwave that can emulate up to 4x4 ac devices to test maximum throughput. The results posted are an average of 10 one minute test runs. This benchmark is fully described in the Revision 9 test process description. The charts have the same filters applied for the average throughput tests shown previously. Keep in mind it is difficult to achieve these numbers in real life. They require both a 4x4 STA and a test client with performance far above what most real clients can do.
2.4 GHz results show all products within ~ 25 Mbps of each other on downlink and ~30 Mbps on uplink. The R9000 does better than the Talon AD7200 in both cases, but not significantly so on uplink.
Maximum Wireless Throughput comparison - 2.4 GHz
The 5 GHz maximum comparison has the R9000 at the bottom of the chart, with significantly lower throughput than its R7800 sibling (743 vs. 928 Mbps). I found the R9000 had to be tested with a lower maximum link rate (MCS8), while the R7800 was able to run at the top MCS9 rate. The R9000's uplink throughput was also in the mid 700 Mbps range, but was still better than the ASUS RT-AC88U and Archer C3150 for this benchmark.
Maximum Wireless Throughput comparison - 5 GHz
Since NETGEAR sent two R9000s for testing, I tested maximum throughput that way, too. One router was set to bridge mode and located 6' away in open air. The router was set to Channel 6, Up to 800 Mbps (40 MHz B/W mode) and 20/40 MHz coexistence disabled for 2.4 GHz tests. Channel 40 and Up to 1733 Mbps mode was set for the 5 GHz tests. The IxChariot throughput.scr script was used with test file size set to 5,000,000 bytes and four connections to generate traffic for the test.
The 2.4 GHz downlink result of 234 Mbps is about half of the Veriwave's. This appears to be due to lower link rate; maximum observed rates reported by the bridge during testing were 364 Mbps downlink and 324 Mbps uplink.
4x4 bridge mode throughput - 2.4 GHz downlink
2.4 GHz uplink throughput was also pretty low at 172 Mbps. In contrast, the Veriwave was able to run its 2.4 GHz maximum throughput tests at a full MCS31 (600 Mbps) rate in both directions. The Veriwave does not support 256 QAM rates in 2.4 GHz because they are not part of the 802.11ac standard.
4x4 bridge mode throughput - 2.4 GHz uplink
The 5 GHz downlink results were better than the Veriwave's at 899 Mbps vs. 743 Mbps.
4x4 bridge mode throughput -5 GHz downlink
5 GHz uplink throughput was also higher, coming in at 931 Mbps vs. 759 Mbps for the Veriwave. Maximum link rates reported by the NETGEAR bridge were 1560 Mbps downlink and 1404 Mbps uplink. This once again illustrates that high link rate doesn't necessarily mean highest throughput.
4x4 bridge mode throughput - 2.4 GHz uplink
60 GHz (802.11ad) Performance
NETGEAR sent an Acer TravelMate P648-M-59KW notebook so I could test 802.11ad performance. It's part of Acer's TravelMate P6 line, with 2.3 GHz Intel Core i5 6200U processor, 8 GB of DDR3 RAM, 256 GB SSD and a Qualcomm Atheros "Sparrow" 802.11ad adapter.
One aspect of 11ad I failed to mention in the TP-Link review is that Wi-Fi and AD connections can be simultaneous. This may be obvious due to the separate radios. But I've learned to not assume anything when it comes to Wi-Fi. The screenshot below shows the Windows connection properties with 5 GHz and 60 GHz connections to the router. Both connections were idle at the time, hence the low 60 GHz link rate. 11ad connections set up just like Wi-Fi connections by design. The unification of 60 GHz and Wi-Fi connection management is one of WiGig's key advantages over previous proprietary 60 GHz protocols.
Wi-Fi and AD simultaneous connections
I did not have a machine with an 10 GbE SFP+ adapter available during testing. So testing was limited by the maximum throughput of a single Gigabit Ethernet connection. NETGEAR's reviewer guide indicated the other way to get > 1 Gbps throughput to / from the 11ad radio would be to run traffic through the 2.4 and 5 GHz radios in addition to a single Ethernet connection. But I did not try this because I can't think of a real-world use application that would do that.
I set the notebook and router up about 4 feet apart with the rear of the notebook facing the front of the R9000. I first tried a drag-and-drop transfer of a 1.15 GB Windows backup file. Throughput varied during the transfer, but the 127 MB/s shown in the screenshot is about the highest value I saw during the transfer. I didn't see the maximum 4.6 Gbps link rate while testing; the usual link rate reported was 3.8 Gbps.
802.11ad file transfer throughput
For a different look, I set up an IxChariot test using the High Performance Throughput script with 8 connections. The screenshots below show a solid 939 Mbps downlink...
60 GHz throughput test - Total Throughput downlink
... and 940 Mbps uplink. Note that once things settled down, throughput was very evenly distributed among the eight connections.
60 GHz throughput test - Total Throughput uplink
I didn't try any distance or next room tests; the TP-Link review already demonstrated that 60 GHz is for line-of-sight use only.
I didn't test the R9000's MU-MIMO performance because the Veriwave system requires direct connection and the R9000's active antennas don't make that practical. Who cares about MU-MIMO anymore anyway? It's like the 3D TV of Wi-Fi; it's included as a sales come-on, but hardly anyone uses it.
I noted in the TP-Link Talon AD7200 review wrap-up that most buyers will experience AD7200 class routers at best as AC2600 class 4x4 routers because they won't have a device with an AD radio. And my advice still remains to buy an AC2600 class router if you want a 4x4 router, the best of which will cost you less than half the R9000's $500 price tag (although as I write this Amazon's price is $450).
That said, NETGEAR has made the R9000 a more compelling buy than the TP-Link Talon if you're after big router bragging rights. The 1.7 GHz quad-core Annapurna Labs processor provides enough compute brawn to provide more powerful storage based features, like Plex and Amazon Cloud backup. And it was a smart more to bump the Gigabit port count up to six, make two of the ports support 802.3ad link aggregation for uplink to a smart switch or connection to a multi-port NAS and include the first 10 GbE port in a consumer router, even if it is SFP+ vs. copper. The R9000 also outranks the TP-Link on Performance on our Router Ranker.
But 10GbE, aggregated ports, 160 MHz bandwidth, 802.11ad and MU-MIMO make the R9000 pretty darn expensive. And how many of use can even use these features now or in the foreseeable future?
Unless you have 11AD devices right now, there is no reason to buy an AD7200 router. You're better off to wait for those enterprising folks over in Taiwan to cobble together 11ad "AP" / bridges, just as they did in 11ac's early days. Sure, the Gigabit Ethernet connection they will probably have (USB 3.0 is also a possibility for a bit more bandwidth) will be a bottleneck. But you'll get more consistent throughput than with 5 GHz Wi-Fi and the benefit of an additional radio that adds bandwidth vs. stealing it from the 5 GHz radio. You'll also be able to put the 11ad radio where you want it, which probably isn't where you want your main router.
Not so long ago, the only thing you got by spending more money on a router was a bigger number on the box that did little or nothing to address the problems you were trying to solve: coverage and capacity. But now you have the choice of spending those same big bucks on multi-AP Wi-Fi systems (aka "mesh") that spend that money more wisely because they're likely to produce better Wi-Fi today with the devices you already have.
Fortunately, NETGEAR has you covered on both approaches. You can buy the R9000, a big-number router with a feature set you won't find on any other router today, or you can buy Orbi, the best Wi-Fi system I've seen so far. Choose wisely, grasshopper.