ASUS routers are known for having lots of features. They also expose more controls for users to twiddle than other vendors, sometimes to not so good effect. With its implementation of Broadcom's Smart Connect, ASUS may finally have gone too far in exposing controls that other companies have not.
The screenshot below shows an extensive array of Smart Connect Rules at their default settings. ASUS is also the first AC3200 router to implement 2.4 / 5 GHz band steering, which means it can move clients between bands.
ASUS RT-AC3200 Smart Connect Rule page
The upshot of all this is that some users are reporting frequent disconnects and in many cases, disabling Smart Connect and manually assigning devices. I got to experience ASUS' Smart Connect disconnects while testing the feature. I used the same devices summarized in the table below used for NETGEAR R8000 and Linksys EA9200 Smart Connect testing.
|Moto X smartphone||1x1 AC||android-d4c097|
|NETGEAR R7000 in client bridge mode||3x3 AC||WLANTEST-STA|
|Laptop with NETGEAR A6200 USB adapter||2x2 AC||x220i|
|iPad 2nd gen||1x1 N||SNB - iPad 2|
|iPod Touch 5th gen||1x1 N||Tim-Touch-G5|
Table 3: Smart Connect Test devices
For Smart Connect testing, the RT-AC3200 was located in the wireless testbed upper test chamber with the door open. The bridge mode R7000 was in the lower chamber with the door closed. This allowed using the testbed programmable attenuators to control the signal and therefore the link rate of the R7000. I set 20 dB of attenuation so that the R7000 throughput wouldn't dominate the other AC devices. All other devices were located within 6 feet of the RT-AC3200, outside the test chamber and all received a nice, strong signal.
I found the RT-AC3200's band-steering was indeed dynamic, in fact, too much so. The Smart Connect test runs a minute and a half and it took multiple tries to complete a test run without a client disconnect. The table below shows four test runs and client connections for each. Note that there were no other networks active during testing (aside from beaconing) and devices were not physically moved between tests.
|Device||Type||Network Map ID||1||2||3||4|
|Moto X smartphone||1x1 AC||android-d4c097||5G-1||5G-1||2G||5G-2|
|NETGEAR R7000 in client bridge mode||3x3 AC||WLANTEST-STA||5G-1||5G-2||5G-2||5G-2|
|Laptop with NETGEAR A6200 USB adapter||2x2 AC||x220i||5G-2||5G-2||2G||5G-2|
|iPad 2nd gen||1x1 N||SNB - iPad 2||5G-1||5G-1||5G-2||5G-2|
|iPod Touch 5th gen||1x1 N||Tim-Touch-G5||5G-2||5G-1||5G-2||5G-2|
Table 4: Smart Connect Test devices
I noticed the two Apple devices were the most frequent disconnectors, switching to other remembered networks. Once I "forgot" all the in-range stored networks on the i-devices (and other clients for good measure), the disconnects stopped and I was able to run tests to completion.
I first disabled Smart Connect and connected all clients to the 5 GHz-2 radio set to Channel 153. The IxChariot plot below shows this test yielded 138 Mbps of total throughput from the five clients.
Total downlink throughput - All clients on 5GHz-2 radio
I then enabled Smart Connect with its default settings and ran another test. Client connections for this test are shown in column 3 in Table 4 above. This set of connections, with clients split between the 5GHz-2 and 2.4 GHz radios, produced 205 Mbps of total throughput, a 49% gain.
Total downlink throughput - Smart Connect run 1
I then power-cycled the router, let the clients reconnect and ran another test (connections shown in Table 4, column 5). This time all clients connected to the 5 GHz-2 radio and produced 184 Mbps total throughput, for a gain of only 33%.
Total downlink throughput - Smart Connect run 2
Finding out how clients connected was, frankly, a pain in the butt. The only way to get this information is to click on each device's link in the Client Status table, to bring up a pop-up with the information, as shown below.
Network Map w/ Client Status popup
It would be much easier to track Smart Connection decisions if ASUS implemented a tablular presentation of device connections like NETGEAR has in the R8000 (shown below).
NETGEAR R8000 device connection table
So how does the RT-AC3200's Smart Connect compare to NETGEAR and Linksys'? Table 5 shows the best Smart Connect gain in two connection scenarios was 49% for the AC3200 vs. almost 3X for both other routers. A big reason for this difference, however, is that the AC3200 started with higher throughput with all five clients on one 5 GHz radio. Still, ASUS' Smart Connect produced the lowest total throughput of the three AC3200 routers while NETGEAR's R8000 produced the highest.
|Product||Without (Mbps)||With (Mbps)||% gain|
Table 5: Smart Connect Total Throughput gain comparison
As has become its modus operandi, ASUS has released another work-in-progress with the RT-AC3200. Its Smart Connect implementation clearly needs work, since it did not consistently spread clients between its two 5 GHz radios to maximize throughput in our very simple test scenario and produced the lowest total throughput gain of the three AC3200 class routers tested so far.
The good news is that the RT-AC3200's wireless performance appears solid, if you disable Smart Connect and manually assign devices to radios. Still, if you have one of the top-performing AC1900 class routers in our Router Ranker, you're unlikely to gain significant range improvement by shelling out $300 for an RT-AC3200.
Our How To Buy A Wireless Router - 2015 Edition advises that AC1900 routers are the sweet spot right now. There is no benefit to be had with the current crop of first generation Quantenna-based 4x4 routers that are still waiting for their MU-MIMO enables. If you're in the mood to spend top dollar and have a busy WLAN with lots of dual-band clients, then AC3200 routers are the better bet. Just, again, don't expect them to improve range.
Right now, NETGEAR's R8000 Nighthawk X6, with its very conservative Smart Connect implementation, is still our AC3200 class recommendation. With its more comprehensive feature set and high storage throughput, the RT-AC3200 has the potential to take the lead. But you'll need to give it the usual three to six months for its bugs to be worked out by the all-too-willing ASUS Army of pay-for-the-privilege debuggers, uh, customers.