There was a lot of news from Qualcomm at MWC, some of it expected, some of it not. The company released items from CPUs to radios to software to marketing alliances, all in a few short days.
The biggest news from Qualcomm was probably the most vague, they announced the new Snapdragon 820 family of SoCs. We say it was vague because of the scant few technical details given out at their press conference, but something is better than nothing. The 820 is said to be built on a leading edge but unspecified FinFET process and should sample in the second half of the year. SemiAccurate also expects the radio to be Cat 11 capable and be offered as a discrete chip for non-Qualcomm SoCs.
More importantly is the core that the 820 is based around, their first in-house 64-bit ARM v8 design. The marketing name for the core is now called Kyro which is not the internal code name, it is just, well, marketing. The core name is what SemiAccurate exclusively told you about in January, not Taipan. Anyone reporting it as Taipan is dead wrong, see here for more on that ‘Teh Intarwebs’ debacle. #echochamber
Next up is a bit more on the marketing side, Qualcomm is sponsoring the Mercedes F1 team. Officially they will be developing ‘connected car’ technology to help the team do a better job, not that they need much help right about now. That said if you have any idea about the sheer volume of data that a modern F1 car generates, not to mention the need to analyze it in realtime, you understand this partnership.
The spectrum around an F1 race is both scarce and overloaded and cars need to sends Gigabytes back to the pits every race so engineers can keep them from spewing stupidly expensive but impressively machined bits out the tailpipe (Note: It is singular as of 2014). If Qualcomm can help Mercedes do this, they are invaluable. Since this is F1 though, we think Mercedes F1 got at least a few hundred times this writer’s salary for the privilege, something that is not out of line for the sport.
Speaking of radios, another big Qualcomm announcement at MWC was the Cat 11 radio capable of 600Mbps or so. It was demoed at their booth and guess what, it performed as expected. The other side was an Ericsson unit, having a transmitter without a receiver doesn’t do much good, same for the inverse. If you question the real-world applicability of such speeds, take a look at the tests Signals did here.
Two more technologies that may seem easy on the surface but are a royal pain to implement were also shown off at MWC, Cat 6 dual SIM and LTE-U to Wi-Fi session handoff. If you think about it, doing Cat 6 is problematic because you need three discrete paths for your data. That means three discrete authorizations, billing, and related back-end stuff for a single session. If you have two SIMs, you could potentially double those paths. While it is quite doable, there is a lot of work that needs to be done to eliminate corner cases and other rare but potentially painful problems. It isn’t as simple as another SIM slot but it is now done. On the up side everything up to Cat 12 uses only three channels downstream so things should be good for a while.
LTE-U to Wi-Fi handoff is similarly problematic in minor ways. The short version is that the session can be handed off from LTE to Wi-Fi seamlessly or if you want to make an engineer’s life hard, you can use both at the same time. Handing off is a pain but it is not a terribly complex technical problem, implementations are where the problems are found. On the sharing side with the simultaneous use of both radios, things can get tricky.
Most of the problems here are solved, or at least avoided, by having the LTE radio scheduler schedule all of the packets on both sides. That means the Wi-Fi router must be aware of and coöperate with the LTE radio so both probably have to be controlled by the same entity. We say probably because the chances of two telcos cooperating, at least in the US, is about as likely as Comcast doing the right thing for a customer, IE far less than zero.
So for the handoff or sharing to occur, the Wi-Fi base station has to notify the LTE network that a device using said LTE network just logged on to it. The network then notifies the appropriate radio which then starts cooperating with the Wi-Fi router. This dance is the reason the Wi-Fi routers have to have firmware that is LTE-U aware, a minor change all told. From there the LTE radio shunts packets down whichever pipe is best, or both.
LTE-U is short for LTE-Unlicensed because it uses 5Ghz unlicensed spectrum which some say will be problematic for users. SemiAccurate is not up on the minutia of spectrum usage and interference headaches so we can’t say much there. Blasting LTE across a macro-cell area would basically trash all possibilities of Wi-Fi in the cell but luckily that isn’t going to happen.
Update March 12, 2015 @ 11:00am: LTE-U uses the 5GHz band, not the 2.4GHz as we previously stated. It has been corrected above.
The point of LTE-U is to use smaller micro- or pico-cells to do the work in an enclosed area. The vision for LTE-U to Wi-Fi is to have a small cell network in an office building or stadium which cooperates with the Wi-Fi network on premises. In that scenario a single vendor operating both networks is a highly likely prospect, and coöperation is probable even if it is two different vendors. When the customer says vendors should coöperate, they… they… nice weather here, how about where you are?
There were two other technologies shown off by Qualcomm at MWC mentioned during their keynote that deserve a much more detailed writeup. These two were the Zeroth AI/machine learning platform and the Sense ID ultrasonic fingerprint reader. Once you dig under the marketing terms on both, there is a lot of tech under the hood. Stay tuned.S|A
Latest posts by Charlie Demerjian (see all)
- Qualcomm ups Spectra ISP to 2.0 - Aug 15, 2017
- Coffee Lake is going to impact Intel’s margins - Aug 10, 2017
- SemiAccurate digs up Intel Coffee Lake specs - Aug 9, 2017
- Everspin hits the 1Gb milestone with new 28nm MRAMs - Aug 8, 2017
- Intel Xpoint/Optane still doesn’t work right - Aug 7, 2017