Qualcomm’s IPQ8074 and QCA6290 look to be the first 802.11ax silicon to sample. This new standard is going to change WiFi in ways SemiAccurate thinks users probably won’t notice directly but will benefit from.
Lets start out with the two chips, IPQ8074 is meant for infrastructure and base-stations while the QCA6290 is the client device SoC. Both support 802.11ax and that is about where the similarities end. IPQ8074 is a 12-chain simultaneous dual-mode 2.4 and 5GHz radio. It can do 8×8 on 5GHz in ax mode plus four streams at 2.4GHz. Total throughput is said to be 4.8Gbps but you won’t see that on the client side as each 80MHz stream supports far less throughput.
You are probably wondering about the compute performance needed to push the signals, not just raw math from the DSPs but the control features as well. This need for number crunching means the IPQ8074 is a quad-A53 SoC with a lot of other accelerators on board too. Qualcomm built it on an unnamed 14nm process not for die size reasons but because it can be run of PoE. This is a killer feature for the corporate market, an 802.11ax 8×8 base station without external power, if it isn’t a holy grail, it sure is going to be warmly received by IT departments.
On the client side the QCA6290’s name should explain its entire feature set, but for those of you not up on current Qualcomm nomenclature it is a 2×2 MU-MIMO radio. In a change from the normal routine the client side can do simultaneous 2.4 and 5GHz operation as well for a claimed total throughput of 1.8Gbps. This simultaneous dual band on the client side seems like a waste of energy until you hear about the higher level features of the devices. The QCA6290 can do the full 8×8 sounding for all the signaling benefits it brings.
Speed is not the driving factor any more
802.11ax is not about raw speed on a single stream, it is about supporting more users in a given area of space. The speed goes up a bit mainly through the use of 80MHz channels and 1024QAM modulation but the real killer app is the density of devices supported. And the low energy use. And the ability to build higher (OSI) layer feature in a more sane manner than ever before. Ironically most of these things comes from a single new feature, a scheduler.
Instead of monitoring all packets and only passing the appropriate ones to the system, 802.11ax can schedule when to listen for a signal. If there are a hypothetical 10 time slots on a given channel, older 802.11 specs would listen to all 10, process all 10, and pass on the 1/10th of the data it was ‘supposed to hear’. 802.11ax effectively hears it’s packet then goes to sleep. When it is time for the next packet it wakes up and listens again, hopefully a bit before the needed packet arrives but you get the idea. This is said to save ~2/3rds the energy of prior specs and allows Qualcomm to add 2×2 MU-MIMO on mobile clients and support higher throughputs per channel at a net lower energy use. [Editor’s note: I wish I had the ability to take short naps so quickly.]
Since the IEEE only defines the MAC and PHY layers with 802.11ax, what are the higher level features? No not the naming schemes for the 802.11 standards, those are still determined by the old hag with one eye in the cave cackling over the boiling pot of wafer etch while reading sheep entrails for naming ideas. Sometimes the old ways still are the best and IEEE naming seems to be nothing if not tradition bound. (Note: Tell me I am wrong, 802.11a/b/g/n/ac/ad/ax/ay? Really?)
But on the tech side 802.11ax does do one very useful thing, it can pass a lot more data directly to the higher layers of the OSI stack. Instead of the layer 3+ software having to figure out channel conditions on their own, poll registers, and do lots of unnecessary work, ax can just pass the relevant data directly to processes that need it. This saves power, time, and coding costs, plus it gives a better answer than the old ways, sheep entrails or not.
This allows Qualcomm to do all sorts of useful things like SON (Self Organizing Networks) on top of 802.11ax hardware. They implement SON over 802.11ac but now it is easier and gets better data. What kind of things can SON and similar middleware do? Transparent handoff between base stations, handoff between 2.4 and 5GHz bands, flipping through channels on the fly to find clear ones, and much more. All the things you have been wanting for the last decade or so are now feasible or at least the bar to implement them has gone down significantly. Why? Scheduling and data passing. Now do you see how the simultaneous 2.5 and 5GHz operation on the QCA6290 is going to be useful?
Qualcomm is promising that the IPQ8074 and QCA6290 are going to sample in 1H/2017, the former is a 14nm product the latter is, well, not disclosed so you can guess what process that is on. In any case it looks like the first 802.11ax devices you see will be made with Qualcomm silicon. Energy use will go down, throughput will go up, but the real benefit is going to be in the higher level features. If things go well you probably won’t notice them, things will just work.S|A
Latest posts by Charlie Demerjian (see all)
- SemiAccurate has Skylake-SP die shots, sizes, and more - Jul 21, 2017
- Skylake-SP has a diverged core - Jul 19, 2017
- Intel’s Purley platform architecture is a step forward - Jul 13, 2017
- Intel launches Purley aka metal Xeons - Jul 11, 2017
- AMD’s Epyc has lots of connections - Jun 26, 2017