This is the 2nd Wireless/DSP Newsletter produced by TIRIAS Research after Will Strauss from Forward Concepts passed the baton. In this issue we’ve gathered some of what we found at Mobile World Congress (MWC) and some of the more interesting headline news on mobile.
Based on reader feedback, we’ve added some formatting and a headline photo of TIRIAS Research’s Principal Analysts (From left to right: Kevin Krewell, Jim McGregor, and Paul Teich).
Intel Buys Mobileye
Mobileye is Intel’s second biggest acquisition of all time; only the Altera acquisition is larger. Intel is investing $10 billion to jumpstart its automotive program. When it comes to optical ADAS, Mobileye has a dominant share, but that is not to say its position is immune to competitors. Mobileye is focused on computer vision, but the general consensus is that autonomous vehicles will need radar and LIDAR as well, because the vehicle needs to see even those things that the human eye cannot. With the backing of Intel, Mobileye may now in be in a stronger position to integrate these other technologies.
However, Intel generally has a terrible reputation when it comes to acquisitions. In most cases, the acquired company dissolves into Intel culture and is never seen again. Often this destroys the technology and the acquisition fails. But, since Intel acquired Altera it is taking more hands-off approach. In the case of Altera there is no evidence that Altera’s business has been hurt by the acquisition. At TIRIAS Research, Kevin has taken a wait-and-see attitude and in an interview with Marketwatch, expressed his view on why this deal is different:
“That is extremely unusual for Intel that they would take their team and move it and put it under an umbrella of an acquisition,” said Kevin Krewell, a principal analyst with Tirias Research. “That is very non-Intel. It is a big commitment.”
Krewell of Tirias Research believes that Intel has performed well with its Altera acquisition so far, in part because it has taken a more hands-off approach than other deals in the past. He also hopes that Intel’s more open approach about computer architecture and design will have an impact on Mobileye, which develops vision systems for cars that are delivered in a black box and cannot be tampered with or changed. Auto makers, especially as they all gear up to add more automated functions to cars on the road to self-driving cars, want to add their own features or customize their offerings.
However, Principal Analyst Jim McGregor is less optimistic about Intel’s prospects:
Mobileye's technology, on the other hand, isn't as advanced as other systems provided by rival chipmakers Nvidia and Qualcomm, said Jim McGregor, principal analyst at Tirias Research.
McGregor points out that the automotive market used to be one of Intel's biggest business segments back in the 1980s, but abandoned it to focus on PCs and servers. Now Intel needs to look beyond those two markets.
"Intel abandoned automotive and now they're paying through the teeth to get back into it," McGregor said.
Intel’s Mobleye acquisition surprised some people and is certainly a controversial one. It’s difficult to say how well the combination of Intel and Mobileye will work out. It really depends heavily on how the automobile manufacturers view the synergy of the two companies working together. Competitors like Nvidia, NXP, Qualcomm, and Renesas will step up to compete with this new and potentially formidable foe. Also note that NXP will be part of Qualcomm later this year, making the leader in the automotive semiconductor segment (NXP) even more formidable.
You can read more about Kevin and Jim’s comments here:
AR and VR will need better audio
The big news for VR and AR at SXSW Interactive was mapping object-oriented audio onto ambisonic sound systems. The back half of that sentence needs a little definition:
Object-oriented audio treats each sound source as a separate object. Each audio object can move with respect to each other and observers.
Ambisonic sound systems use multiple audio channels to create a 3D soundstage around observers. The technique scales from headphones to large outdoor spaces.
A state-of-the-art VR/AR audio system maps sound emanating from multiple audio objects onto an ambisonic playback system in real-time. For bonus points, newer systems are starting to borrow advanced graphics techniques like ray tracing to generate audio reflections from the perspective of observers' ears to better simulate virtual spaces in VR and to match the audio characteristics of physical spaces in AR.
This technique can also be mapped to graphic artifacts for the hearing-impaired or simply to alert a person navigating a noisy environment that audio cues indicate they should look a certain direction.
All of this will drive demand for very low latency audio input and rendering systems, and that means audio DSP compute capabilities are going to see a lot of market pressure to evolve.
Road to 5G at Mobile World Congress
2017 is the year the mobile industry transitions to Gigabit LTE and begins 5G trials; the year that a true UHD 4k and AR/VR experience can be experienced in the palm of your hand; and the year that “IoT” transitions from the Internet of Things to the Intelligence of Things. All this is being enabled through new technologies, chipsets, and wireless technologies.
While many companies made technology and product announcements leading up to MWC, there are many more to come. There were a few smartphone announcements at CES in January from companies like Asus and Huawei, but there will more at MWC and many on the latest versions of chipsets like the Snapdragon 835 by Qualcomm, the Helio X30 from MediaTek, and the Exynos 8895 from Samsung, all leveraging the bleeding-edge 10nm manufacturing technology from Samsung and TSMC. BlackBerry, Huawei, LG, Sony, and Nokia all announced new phones (more to come on this as we get our hands on them).
And just as the latest generation of smartphones are announced, there were announcements on the technology enabling the next generation of chipsets. Qualcomm has already announced a new integrated Gigabit modem and RF front-end, while ARM announced two acquisitions and its first cellular modem.
Huawei at Mobile World Congress
Following the announcements of the Mate 9 for business users and the Honor 6X for on-line purchases, Huawei is using MWC to launch its consumer-centric products. The company launched the P10 and P10 Plus smartphones and its second-generation smartwatch Watch 2. The P10 and P10 Plus leverage the same Kirin 960 processor used in the Mate 9 with enhanced hardware and software application capabilities. The P10 series includes dual cameras with Leica lenses front and back. The most interesting features of the new devices, however, is the expansion of the artificial intelligence (AI) capabilities of Hiborg and the addition of Highlights.
Hiborg is an AI framework that analyzes how a device is used and information about the users. For example, the device recognizes the location when the user turns on (or takes out of airplane mode) in a new country to provide data assistant services, such as exchange rates and contact information for the user’s country’s embassy.
Huawei Highlights was developed in conjunction with GoPro and is based on the GoPro Quick software platform for easy image organization and viewing.
Artificial Intelligence behind mobile apps evolves
Across MWC, OCP Summit, and SXSW Interactive, it was evident that pattern analytics are increasingly leaning on GPU acceleration. This demand spans the market, from classic big data analytics to cutting-edge deep learning techniques. One key aspect of deep learning is that in many ways it automates the job of a data scientist. Programmers with little to no background in statistics or parallel programming can create sophisticated pattern matching capabilities by selecting seed models from libraries, and these systems scale very well by adding GPU acceleration. Think of deep learning as a key capability that will democratize pattern analytics.
Because deep learning based systems (machine learning and artificial intelligence) scale so well using GPUs, all of the cloud giants are designing specialized server architectures to integrate GPUs more tightly with their traditional processor based compute nodes, and all of the enterprise OEMs are doing the same for their traditional IT customers who want to up-level their analytics game.
The OCP Olympus HGX-1 system with its NVIDIA P100 modules and the OpenPOWER (really Google/Rackspace) Zaius system with its IBM POWER9 CPUs and OpenCAPI accelerators both require high performance design, layout, and debugging capabilities.
5G is coming closer to reality through NB-IoT and Gigabit LTE
5G is more than just a bump in speeds, it is the culmination of several advances in wireless interfaces for both high-speed high-bandwidth data and low-power low data rate applications, both of which are key applications for the next generation of DSPs. Whether it’s an in-house solution from companies like Qualcomm or third-party IP providers like CEVA and Synopsys, all the DSP vendors are offering new solutions for 5G. On the low-end, there are a growing number of narrow-band (NB-IoT) modem solutions for low-power sensor and machine-to-machine applications. As indicated above, even ARM introduced a modem running on a Cortex-M33 MCU.
Not to be done on the high-side is Gigabit LTE. Technically, Gigabit LTE is a step before 5G, which promises speeds of 10Gb/s or higher, but just achieving gigabit-class solutions is a significant step. Just prior to MWC, Telstra announced the deployment of Gigabit LTE in Brisbane, Melbourne, and Sydney in conjunction with Ericsson, Netgear, and Qualcomm. The week after MWC, Sprint demoed a Gigabit LTE solution at a Pelicans professional basketball game in conjunction with Motorola and Qualcomm.
Additional 5G modems and Gigabit LTE networks will be rolled out throughout 2017. Stay tuned for more details.
As always, we encourage your feedback
Kevin Krewell, Jim McGregor, Paul Teich
TIRIAS Research is a high-tech research and advisory firm, an independent third-party resource to high-tech companies. We provide custom research and advisory services on technologies, markets and ecosystems to a select group of technology industry leaders. Our Principal Analysts have decades of in-depth expertise in silicon, software, and systems specification, design and deployment.
If this is a pass-along copy and you want to be added to our free newsletter mail list email us at newsletter (at) tiriasresearch.com.
We respect your privacy. To remove your name from future mailings please reply to this message with the word BYE as the subject.