I was briefed by both Atheros and Ralink last week on their announcements at this week's Computex over in Taipei. While it wasn't the sole reason for the calls, I was able to get a feel for why we still haven't seen any three-stream N routers, even though the first were announced well over a year ago.
Ralink, Marvell were the first to announce 450 Mbps / three stream N chipsets around the end of 2008 / early 2009 and D-Link and TRENDnet jumped in to announce routers that used them. But neither router has ever shipped and it's unlikely they ever will.
Ralink said that high-end , i.e. expensive, routers need to be dual-band concurrent and that the initial three-stream N chipsets were both too expensive and CPU limited. As I've noted in earlier articles, going from two stream to three stream N isn't just a matter of upgrading firmware and drivers. The number crunching for handling the additional stream (plus optional N features such as beam forming) is a lot more intense than needed for dual-stream and far beyond the capability of dual-stream MAC/BB processors. So, moving to three-stream N requires a new BB/MAC.
But, at least in Ralink's experience, their initial stab at three stream processing (the RT2883) fell short of what was needed. So they had to go back and come up with a beefier chip, which they now have in the form of the RT3883, unveiled at this past January's Consumer Electronics show.
Atheros waited for the dust to settle a bit before announcing its AR9300 three-stream N chipset last November. That delay presumably gave them enough time to ensure that the devices had enough horsepower, because they aren't announcing any new three-stream N chipsets at Computex. Instead, their Computex announcement touts "router solutions" developed with AppliedMicro, Freescale, Ubicom and Cavium.
Atheros' explanation for the absence of any shipping three-stream routers to date was slightly different from Ralink's. While Atheros also cited high cost as a problem, they said that more work has been needed to improve stability and performance, especially for video streaming.
I asked both vendors whether they thought that HD video streaming was going to require dedicated adapter pair products that are video-optimized. Ralink said they didn't think so, but then cited their partnership with Celeno, which is focused on HD video distribution for service providers—certainly not a general-purpose wireless system that can handle both high speed data and HD video distribution.
Another vendor taking a dedicated video bridge approach is Quantenna, whose technology is supposed to be shipping this year in a NETGEAR product. (More about that here.)
I've rambled a bit, but the bottom line here is that I still don't get the sense that anyone is too excited about pushing forward with three-stream N products anytime soon. I think another part of the problem is the lack of client-side solutions, especially for folks who want to go from two-stream N to three.
Upgraders will most likely need USB adapters, since PCI slots are fast disappearing and ExpressCard slots never seemed to catch on. But with client side chipsets mainly designed for internal PCIe bus interface, it's more likely that we'll see three-stream N converters / bridges / APs like those that first appeared to add N capability to G routers.
So if you've been waiting for an N router to get your HD video streamed to that room to which no Ethernet can go, you'll need to keep waiting. Vendors are still trying to drive down cost and drive up performance (especially video) and it seems that it's taking much longer than they thought to get the job properly done.