When you run more than one client on a single radio, the radio's total available bandwidth is shared among the clients. When those clients support different wireless standards with differing maximum link rates, that sharing gets interesting.
Figure 11 shows how bandwidth sharing went when both the Buffalo WLI-H4-D1300 bridge and an Intel Centrino Ultimate-N 6300 three-stream N client were both associated with the D1800H's 5 GHz radio. The D1800H was set to 11ac/n/a 1300 Mbps Mode (80 MHz) on Channel 36, with both clients using a WPA2/AES secured connection. Note that the Y axis doesn't go to 0 so that you can better see plot detail.
Figure 11: Mixed draft 11ac and 3 stream N clients
The draft 11ac pair starts first at about 138 Mbps. When the three-stream pair kicks in at 20 seconds, total bandwidth actually increases to (if my eyeballing is correct) around 150 Mbps (~ 83 Mbps for draft 11ac + 68 Mbps for 3 stream N). But the draft 11ac pair gives up 40% of its bandwidth to the cause. When the draft 11ac pair ends around 35 seconds in, the 3 stream N pair's throughput increases to an eyeballed average of around 85 Mbps, only about a 25% increase.
So in a mixed environment, a "1300 Mbps" draft 11ac client is clearly going to take the biggest throughput hit. Therefore, if you have a busy N wireless LAN and especially one with a lot of streaming and large file transfers, the "add, don't replace" rule for the G-to-N transition will certainly apply for the N to AC transition.
I like to run a simple "stress test" on new router architectures to see how wired routing and wireless throughput interact. I set up a test with four IxChariot simultaneous tests:
- LAN to WAN wired routing throughput
- WAN to LAN wired routing throughput
- Draft 802.11ac wireless downlink
- Three stream N, 2.4 GHz wireless downlink
This set of tests simultaneously ran the wired routing, 2.4 GHz radio and 5 GHz radio as fast as their bits would flow. In Figure 12, the routing tests start first, followed by both radios at the 10 second mark. It's hard to tell from the single test below, but after running many versions of the test, I can say that the 5 GHz radio's throughput when in 11ac mode is definitely affected by routing traffic. The effect is less pronounced than I've seen in other routers, however.
Figure 12: Buffalo WZR-D1800H stress test
This test usually shows some interaction between the routing and radio portions of a router and the D1800H is no exception. Figure 12 shows that routing throughput drops precipitously as soon as wireless activity starts. We've seen this in other routers, like the ASUS RT-N66U. Figure 13 shows a plot of the ASUS' stress test, where you can easily see wireless throughput drop when the router, even limited to 100 Mbps, kicks in at the 40 second mark.
Figure 13: ASUS RT-N66U wireless and wired stress test
I tried to duplicate the ASUS' test. But the Buffalo's wireless throughput just wasn't steady enough for me to try to increase the wireless loading by running simultaneous up/down tests on both radios. So I had to settle for using single downlink test for the router and both radios to get the clearer plot in Figure 14.
Figure 14: Buffalo WZR-D1800H stress test - take 2
In this test, I started the 11ac wireless pair first, then the 3 stream N pair on the 2.4 GHz radio 20 seconds in and finally the routing downlink at 40 seconds. Although the 11ac pair throughput bounces around a bit, you can see a slight throughput decrease when the 2.4 GHz radio kicks in. But there is a more easily detected change when the router starts. To keep the vertical axis compact, I limited the max routing throughput to 300 Mbps and also ran a test with it cranked down even lower to 100 Mbps. In either case, the 5 GHz radio throughput was more negatively affected as soon as routing activity started.
You can see the 2.4 GHz plot rise shortly after one minute and both routing and 5 GHz radio activity start. But in other test runs, the increase was lower than you see in Figure 14.
At any rate, there is interaction between the routing and wireless portions of the D1800H, particularly when operating at draft 11ac speeds. There is some radio-to-radio interaction, too. But the amount is much less than the router-to-radio effect.
There is both good and bad news for those itching to run out and buy a draft 11ac router. The good is that, when paired with its WLI-H4-D1300 partner, the WZR-D1800H can produce almost 450 Mbps of aggregate throughput when handling multiple clients. Even better, though, its that the pair can produce around 100 Mbps of usable throughput at my weakest signal test point for a single test client!
This is more bandwidth than I've ever seen available from any other 5 GHz wireless product and may even be capable of sustaining a trouble-free 1080p HD video stream. The catch, however, are the large and long throughput dropouts that I saw in many of my tests. So unless your HD streaming player has some decent buffering, you may still be out of luck! I will have to give HD streaming a shot in the coming weeks, after I clear out some of my review backlog.
The bad news is that you'll need to spend almost $400 to run the above experiment. And the more practical bad news is that the WZR-D1800H isn't a particularly good simultaneous dual-band three-stream N ("N900") router. If that is what you're looking for, you may be better off spending about the same price and picking up an ASUS RT-N66U.