Smart Connect - more
I next forced all clients off, clicked the SmartConnect box and let clients reconnect. The Attached Devices screen shows both AC clients (the R7000 and Moto X smartphone) connected to the low-band radio and the 1x1 N clients (iPod Touch and iPad) connected to the high-band radio. This represents a proper separation of "slower" and "faster" clients. But the assignment is opposite the description in product promotional material and data sheet, which identifies the low band radio for "dual-band WiFi devices" and the high band for "newer dual band AC WiFi devices".
Device connection - SmartConnect On
With the devices connected as above, I ran the same IxChariot script with the results shown below. Both AC devices do much better this time and the slower N devices do about the same. But total throughput this time is 258 Mbps, a 4X increase.
Throughput - SmartConnect On
Given the very unequal throughput distribution between the 3x3 and 1x1 AC devices, I decided to add a 2x2 AC device into the mix and repeat the experiment. I used the NETGEAR A6200 adapter in my trusty Lenovo x220i notebook with its internal wireless disabled. I also used only the iPad as the "slow" device.
The Attached Devices screen below shows all three devices connected to the high band radio with SmartConnect disabled. Note the 192.168.1.101 entry you see in some screens is the R7000 bridge IP, not the 192.168.1.102 Ethernet-connected client that the IxChariot test uses. In other words, consider the 192.168.1.101 and 102 devices as one wireless device.
Device connection second mix - SmartConnect Off
This produced total throughput of 83 Mbps shown below. Not much better than the first result.
Throughput second mix - SmartConnect Off
Enabling Smart Connect and reconnecting devices once again shows all AC devices on the low-band radio and the lone iPad on the high.
Device connection second mix - SmartConnect On
Total throughput this time is slightly lower (247 Mbps vs. 258), with the 3x3 device once again getting the bulk of the bandwidth. The 2x2 A6200 adapter (Pair 6 in the plot) gets only 29 Mbps, less then 10% of the total.
Throughput second mix - SmartConnect On
I also tried to see if I could get SmartConnect to assign the R7000 bridge to the "slow" device group by raising attenuation to the point where it had a very low link rate. But no matter what I did, SmartConnect figured out that it was a "fast" AC device and connected it with the other AC devices.
$300 is a lot to spend on a router, so they better be worth it. The R8000's value proposition is that its two 5 GHz radios will provide better total throughput utilization by separating "slow" and "fast" devices. It's not promising improved range. It's not promising higher (single device) speeds. But it is promising to fix a problem many of us have and don't realize it--that our slower wireless devices are making our faster ones look like they are broken.
My simple experiments show that the R8000 can deliver significantly higher wireless throughput with a mix of 5 GHz devices by assigning them to different radios. But the experiments also show that bandwidth sharing among devices connected to each radio may not be very equal. Perhaps it was the mix of devices I used. But between the 3x3 and 2x2 NETGEAR devices I used, the 3x3 got around 74% of total throughput and the 2x2 got only 11%.
Given the capabilities I saw in Broadcom demos, NETGEAR appears to have dialed XStream's capabilities down to a very simple level. Radio assignment appears to be based strictly on supported link rates, not the actual link rate in use. And NETGEAR has said it disabled dynamic radio assignment, citing problems it found with some devices not liking being steered. This bears another look, which I will do in Part 2, along with OpenVPN performance and full wireless testing.
In the meantime, don't be so quick to jump on this or the other "vacation special" that ASUS is trying to get you to buy. They are expensive, the firmware is not stable (especially so right now for the ASUS RT-AC87) and the technology has not been significantly shaken out to ensure trouble-free use. Let someone else take the pain for now.