The SNB Forums are awash with waaay too much excitement about the new 4x4 Broadcom-based AC3100 and AC5300 class routers. Even folks who paid around $300 not too long ago for AC2350 / 2400 4x4 and AC3200 "tri-band" and suffered through bugs and compatibility problems are lining up to spend up to $500 (!) in hope that this time, they'll get "the one".
What has particularly caught my eye are claims of significantly improved Wi-Fi performance. This would seem to contradict my advice to stick with AC1900 routers as the best price / performance choice right now. The data in the Router Charts for AC2350, AC2600, AC3200 and AC5300 class routers back my advice. But my RF chamber based testing isn't the same as open-air testing, so maybe I'm missing something. So it was time to get out of the chamber and into the air for some real world results.
I used the same open-air test process used in Does An AC Router Improve N Device Performance? This time I used two test clients; an N600 class notebook and AC580 class smartphone.
The N600 test client was an Acer Aspire S7 Ultrabook running Windows 8 with an Intel Dual Band Wireless N 7260 Plus Bluetooth adapter with 220.127.116.11 Win 8 64 bit driver. This 2x2 N600 class adapter is found in many notebooks and has the advantage of not providing a performance edge to any router because Intel isn't used in consumer routers. 2x2 adapters in general, either single or dual-band, are also commonly included in many notebooks.
The 1x1 AC580 device was a Moto X XT1094 2nd generation smartphone. I don't know the maker of the radio device.
Since the question this time is whether it makes sense to trade up from an AC1900 router, I used the top-ranked NETGEAR R7000 Nighthawk as the baseline reference. To keep testing manageable, I ran up and downlink open air tests in only one location, with each router in my downstairs corner office and the client in test Location D.
The diagram below shows Location D on the upper level of an approximately 3300 square foot two-level home, approximately 35 feet away (direct path) from the router. There is one wood floor, one lower level sheetrock wall and a sheetrock ceiling between router and Client.
This location was chosen because it represents a mid-signal "transition" location for 2.4 GHz and the furthest operable 5 GHz location; precisely the kind of location where you would want to see a performance improvement from buying a new router.
Each IxChariot throughput script using a single TCP/IP stream was run once for 90 seconds. If the results were particularly low or unstable, the test was repeated. There were no other networks active during testing. As is our standard practice, WPA2/AES encryption was used for all testing with channel 6 and 20 MHz mode used in 2.4 GHz and channel 153 and Auto mode used in 5 GHz.
The routers tested are listed in Table 1. All were loaded with latest firmware before testing. All the processors are dual-core. I originally was going to test only the newer Broadcom-based 4x4 routers and the new MU-MIMO enabled NETGEAR R7500v2. But I decided to include the previous-generation 3x3 "tri-band" AC3200 class routers, because I have seen so many people itching to trade up from them.
|NETGEAR R7000 (reference)||18.104.22.168_1.1.79
||AC1900||Broadcom BCM4709 @ 1 GHz||$187|
||AC3200||Broadcom BCM4709 @ 1 GHz||$259|
||AC5300||Broadcom BCM47094 @ 1.4 GHz||$495|
||AC3100||Broadcom BCM47094 @ 1.4 GHz||$280|
||AC3200||Broadcom BCM4709 @ 1.4 GHz||$270|
||AC5300||Broadcom BCM47094 @ 1.4 GHz||$355|
|NETGEAR R7500v2||V22.214.171.124||AC2350||QCA IPQ8064 @ 1.4 GHz||$198|
Table 1: Routers and firmware
Neither link rate nor RSSI was monitored during testing, because both are only indirect indications of performance and vary greatly from device to device. Telling you an RSSI of -60 dBm was measured means little because your environment (router, device, physical environment, RF environment) is different than mine. In the end, it's delivered throughput that matters, so that's what was measured.
The Results - N600 Client
The percentage difference between the throughput measured using the NETGEAR R7000 reference router and the router under test was calculated. Results were then plotted for each band and traffic direction. If you want to do the math, the reference throughput measurements were: 2.4 GHz dn - 22.6 Mbps; 2.4 GHz up - 53.9 Mbps; 5 GHz dn - 73.1 Mbps; 5 GHz up - 43.4 Mbps.
2.4 GHz downlink results are the best of the four test runs, with all routers producing higher throughput. Note that trading up from both ASUS' RT-AC3200 and NETGEAR's R8000 AC3200 class routers resulted in higher throughput. (56% higher for the ASUS; 78% higher for the NETGEAR).
2.4 GHz Downlink Throughput Change - N600 client
2.4 GHz uplink results also showed many gains, but also the first losses. The only NETGEAR router to produce higher 2.4 GHz uplink throughput was the AC5300 class R8500 Nighthawk X8.
2.4 GHz Uplink Throughput Change - N600 client
As you might expect, 5 GHz downlink results were quite different. For AC3200 and AC5300 routers, I ran two sets of tests; one with the standard Channel 153, the other with Channel 44. The reason was that "tri-band" routers use internal antennas for one of the 5 GHz radios. Since the NETGEAR 8500's amplified antennas connect to the "low band" 5 GHz radio, I thought it would be good to see what difference they made compared to the non-amplified antennas on ASUS' RT-AC5300 and the two AC3200 class routers.
Turns out it's good I ran the Channel 44 tests, since they were the only ones showing throughput gains. The NETGEAR R8500 showed the most change, moving from a 23% throughput loss using the internal non-amplified antennas on Channel 153 to a 72% (!) gain with the external amplified antennas on Channel 44. Note that both low and high 5 GHz bands can use the same maximum transmit power levels since an FCC rule change in 2014. All routers in this test are designed to take advantage of that change.
5 GHz Downlink Throughput Change - N600 client
5 GHz uplink results are even more dismal; merely an 8% gain for only one product, ASUS' $500 RT-AC5300.