Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

MIPI DevCon 2016: Mobile Technology Expands to New Horizons

1,755 views

Published on

Smartphones are perhaps the most technically demanding semiconductor-based products, requiring high CPU and graphics performance within tight power, size and cost constraints. Processors forged in this crucible can also serve in emerging markets such as IoT, wearables and automotive. Linley Gwennap, founder and principal analyst of the
Linley Group, discusses the latest market and technology trends in the mobile market, including 5G, heterogeneous computing and sensor fusion. It also covers how these and other mobile technologies are impacting adjacent markets.

Published in: Mobile
  • Hurry up, Live Webinar starts in 6 minute! it's about the FREE Training Webinar: An insider system that made $23,481 in last 6 weeks! ♥♥♥ http://scamcb.com/zcodesys/pdf
       Reply 
    Are you sure you want to  Yes  No
    Your message goes here

MIPI DevCon 2016: Mobile Technology Expands to New Horizons

  1. 1. © 2016 The Linley Group September 14, 2016 Mobile Technology Expands to New Horizons Linley Gwennap, Principal Analyst, The Linley Group MIPI Developers Conference 2016
  2. 2. © 2016 The Linley Group September 14, 2016 About Linley Gwennap •  Founder, principal analyst, The Linley Group • Leading vendor of technical reports on semiconductor products •  Editor-in-chief of Microprocessor Report • Author of recent arGcles on ARM, Broadcom, Intel, Marvell, MediaTek, Nvidia, NXP, Qualcomm, Silicon Labs, et al •  Coauthor of A Guide to IoT Processors, “A Guide to Mobile Processors,” and “A Guide to Advanced AutomoGve Processors” •  Former CPU designer at HewleU-Packard 2
  3. 3. © 2016 The Linley Group September 14, 2016 Agenda •  Mobile trends •  Wearables trends •  IoT trends •  AutomoGve trends •  Conclusions 3
  4. 4. © 2016 The Linley Group September 14, 2016 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% 0 200 400 600 800 1000 1200 1400 1600 1800 2000 2013 2014 2015 2016 2017 2018 2019 2020 Millions of Units Smartphone Shipments ConAnue to Rise •  Smartphone penetraGon is 69% of all handsets and leveling off •  Most remaining growth is at low end in developing countries 1.95 billion smartphones in 2020 Samsung (Source: The Linley Group) Apple China Tier 1* Other *Huawei, Lenovo, Xiaomi, Yulong, ZTE 6.6% CAGR 2015–2020 Growth Rate Annual Growth 4
  5. 5. © 2016 The Linley Group September 14, 2016 Qualcomm Clings to Smartphone Processor Lead •  Qualcomm gains at high end • Galaxy S6 (2015) uses no Qualcomm • Galaxy S7 uses ~30% Qualcomm •  MediaTek gains share from Qualcomm in low end • LTE price war favors MediaTek •  Spreadtrum gains share as market grows at low end •  These vendors plus “internal” hold 98% of smartphone market 0% 10% 20% 30% 40% 50% 2013 2014 2015 2016f Qualcomm MediaTek Spreadtrum Internal* *Apple, Huawei, Samsung (Source: The Linley Group) 5
  6. 6. © 2016 The Linley Group September 14, 2016 Custom vs Standard IP Cores •  Many processor vendors license CPU, GPU, and other IP cores • E.g., CPUs from ARM, GPU from ImaginaGon •  Others design their own cores to achieve differenGaGon • Only the largest vendors, or those that serve mulGple markets, can afford this 6 All Standard IP Designs One Key IP Designs Most/All IP MediaTek Samsung (CPU) Qualcomm (CPU, GPU, DSP) Spreadtrum Apple (CPU) Intel (CPU, GPU) Rockchip Huawei (CPU) Arrow indicates recent development
  7. 7. © 2016 The Linley Group September 14, 2016 Heterogeneous Compute •  GPU, DSP, and other cores can offload specific tasks, improving power efficiency versus CPU somware • Special cores are custom designed for certain applicaGons, e.g. graphics, video, audio, image (photo) processing •  Shared virtual memory added in OpenCL 2.0 • Previous model (top) required manually copying data between GPU and CPU memory spaces • Shared memory allows GPU and CPU to operate on the same data structures • This approach simplifies somware and improves performance 7 GPU 
 Memory Main
 Memory GPU CPU On-chip Interconnect Shared
 Memory GPU CPU Coherent Interconnect
  8. 8. © 2016 The Linley Group September 14, 2016 5G Coming Soon? •  5G is a suite of different technologies • New bands for higher data rates • New protocols for improved service and reduced latency • Greater ability to aggregate mulGple types of connecGons simultaneously •  E.g. licensed/unlicensed, small cell/macro cell, LTE/5G •  3GPP Release 15 will standardize 5G with deployment likely in 2020 • First 5G smartphones likely to deploy at that Gme • Service and latency guarantees will improve support for IoT devices (M2M) •  Some operators (e.g. Verizon) plan to launch “5G” service in 2017 • Using a subset of prestandard 5G technologies for broadband service only 5G phase 2 2016 2017 2018 2019 2020 2021 2022 Rel 15 Rel 16 Rel 17 5G launch 5G trials Rel 14 Pre-5G launch Pre-5G trials 8
  9. 9. © 2016 The Linley Group September 14, 2016 Smartphones Add More Sensors •  9-axis + barometer for best posiGoning •  MulGple light sensors • E.g. proximity, ambient light, color •  Environmental • E.g. temperature, humidity •  New biometric sensors • E.g. fingerprint, heart rate •  Need to carefully manage baUery life • Use DSP or sensor hub to track sensors (Source: The Linley Group) 9
  10. 10. © 2016 The Linley Group September 14, 2016 Agenda •  Mobile trends •  Wearables trends •  IoT trends •  AutomoGve trends •  Conclusions 10
  11. 11. © 2016 The Linley Group September 14, 2016 Many Types of Wearables •  Smart bands lack displays but track steps, personal health • Fitbit ($79+) and Xiaomi Mi Band ($15) are most popular • Shipments of about 49 million in 2015 •  Smart watches use display to provide noGficaGons, apps • High-end watches support apps, typically cost $299 and up •  Apple Watch is most popular; others run Android Wear OS • Low-cost watches sell for as liUle as $50 • About 24 million smart watches shipped in 2015 •  Also smart clothing, medical wearables, ear buds, jewelry… 11 LG Urbane (Android) Apple Watch Fitbit Flex
  12. 12. © 2016 The Linley Group September 14, 2016 Leading Wearables Processors •  Most Fitbits use STMicro MCUs • Silicon Labs MCU in some versions •  Mi Band uses Cypress Bluetooth controller • Not designed as an MCU, but it can run simple code •  Apple Watch uses Apple processor • Originally designed for iPhone •  Most Android watches use Qualcomm Snapdragon • Snapdragon also used in smart glasses, VR headsets, and others •  Most low-cost watches, bands use standard microcontrollers 12 Apple S1 SiP
  13. 13. © 2016 The Linley Group September 14, 2016 Wearables: Cost Is Key •  Successful consumer wearables need: • Desirable use case (e.g., noGficaGons, call screening) • Easy to use (good UX, compaGble with PC/phone) • Long baUery life (mulGple days or weeks) • Price point below $100, preferably below $50 •  Phone makers will bundle with phone • $50 adder on $400 phone is almost a no-brainer •  Easy to take BOM cost well below $50 • Watches use small display, small baUery • Requires highly integrated mixed-signal SoC 13
  14. 14. © 2016 The Linley Group September 14, 2016 0% 2% 4% 6% 8% 10% 12% 14% 16% 0 50 100 150 200 250 300 350 400 2014 2015 2016 2017 2018 2019 2020 Percent of Smartphone Sales Millions of Units Smart Watch to Dominate Wearables Market •  Assumes Apple releases a popular smart watch in 2017 •  Followed by popular Android watches in 2018 116 million wearables in 2016 Watch Band Watch vs Smartphone % Other 38% CAGR 2015–2020 (Source: The Linley Group) 380 million wearables in 2020 14
  15. 15. © 2016 The Linley Group September 14, 2016 Agenda •  Mobile trends •  Wearables trends •  IoT trends •  AutomoGve trends •  Conclusions 15
  16. 16. © 2016 The Linley Group September 14, 2016 The IoT Connects Clients to Networks to the Cloud •  Most IoT clients (or edge devices) connect to the Internet through gateways • ConnecGons may be wired or wireless • A client senses and/or controls some aspect of its environment •  Gateways (or hubs) provide secure network access • Clients may first connect to “fog” (also a Cisco term) for storage and real-Gme analyGcs •  IoT services run on standard cloud servers • No hardware differences for IoT in the cloud 16
  17. 17. © 2016 The Linley Group September 14, 2016 Wireless SoCs Target IoT Clients, Low-End Wearables •  Sensor nodes and fitness bands employ MCUs running RTOS •  Several MCU vendors now offer complete wireless SoCs • IntegraGng CPU, memory, analog & digital I/O, radio • Minimizes cost and power •  On-chip flash and SRAM is sufficient for BLE and 802.15.4 • CPU frequencies typically <50MHz •  Manufacturing in 55nm and larger-geometry CMOS processes enables integraGon of RF components •  More complex Wi-Fi protocol requires faster CPUs, larger memories, typically off-chip flash 17
  18. 18. © 2016 The Linley Group September 14, 2016 IP Vendors Tuning Processor Cores for IoT •  New, low-power 32-bit CPUs (e.g. Cortex-A32) •  GPUs trimmed for smaller screens, longer baUery life • Examples: Mali-470, PowerVR G6020 •  CPU-IP vendors acquire/develop radio cores • BLE: ARM, Ceva, ImaginaGon, Synopsys • Wi-Fi: Cadence, Ceva, ImaginaGon 18 CPU Memory Radio Analog
  19. 19. © 2016 The Linley Group September 14, 2016 Vision of the IoT in Five Years •  In five years, smart home and smart appliances will comprise more than half of all IoT devices • Home-security and home-automaGon employ mulGple devices per household • Several smart appliances per home •  Industrial IoT will conGnue to grow at a steady pace, comprising 26% of IoT shipments in 2021 •  Wearables will mostly comprise smart watches, with 2-3 year replacement cycles driving shipments to 20% share of market 19
  20. 20. © 2016 The Linley Group September 14, 2016 Agenda •  Mobile trends •  Wearables trends •  IoT trends •  AutomoGve trends •  Conclusions 20
  21. 21. © 2016 The Linley Group September 14, 2016 In-Dash, Safety Drive AutomoAve IC Growth •  In-vehicle infotainment (IVI) in >50% of cars by 2022 • Car users want smartphone-like interface • Rear camera mandated in all US cars in 2018 •  Even faster growth in acGve safety (ADAS) • Autonomous braking (AEB) in all US cars in 2022 • AcGve cruise control, lane-keeping assist popular • Available for as liUle as $1,000 on Honda Civic LX •  Autonomous cars coming soon • First producGon vehicles on sale in 2018 • Requires mulGple cameras, other sensors 21 (Source: The Linley Group)
  22. 22. © 2016 The Linley Group September 14, 2016 IVI Systems Increase in Complexity •  In-dash systems omen use smartphone processors • E.g. Nvidia Tegra, NXP i.MX, TI OMAP (now DRA75x) • Intel Atom, Qualcomm Snapdragon also making inroads •  Surround view offers simulated overhead view • Processor must combine images from 4-6 cameras • Requires high processing perf, extra camera inputs •  Digital dashboard adds a second screen • IVI processor typically drives both screens •  IVI system may also control rear-seat displays • Up to four 1080p displays, moving to 4K in near future 22 Surround View (photo by BMW) Digital Dashboard (photo by Nvidia)
  23. 23. © 2016 The Linley Group September 14, 2016 ADAS Requires Complex Processing •  Some ADAS derived from smartphone processors • E.g. Nvidia Tegra, NXP i.MX, TI OMAP • Need to add vision-processing engine to boost perf •  Vision processing requires many pixel operaGons • Need parallel compute of 8-bit data types •  Neural networks omen used for object recogniGon • May use integer or floaGng-point coefficients •  May combine data from mulGple cameras, sensors • E.g., radar, lidar, ultrasonic, infrared sensors 23 Vision Processing in Urban Area (courtesy of Mobileye) Lane-Keeping Assist (LKA)
  24. 24. © 2016 The Linley Group September 14, 2016 The UlAmate Goal: Autonomous Vehicles •  Tesla Autopilot, Volvo Pilot Assist, etc available now but require conGnuous driver supervision •  IniGal autonomous vehicles likely in 2018 • Self-driving in certain condiGons only, such as highway driving and only in dayGme and good weather •  Ford promises fully autonomous vehicle in 2021 • Self-driving in a wide range of condiGons •  Incremental hardware BOM cost of <$5000 by 2022 • A huge win for taxi services (versus >$30,000/yr for drivers) • For personal cars, insurance savings could offset cost 24 Google Prototype Vehicle (photo by Linley Gwennap) Velodyne VLP-16 Lidar System
  25. 25. © 2016 The Linley Group September 14, 2016 Conclusions •  Smartphone growth is slowing, but the market is sGll huge •  Biggest opportunity in wearables is smart watches •  IoT growth is coming in smart home and smart appliances •  More than half of all cars will soon have in-dash display systems •  Many of these IoT and automoGve systems using mobile processors and mobile camera/display technology 25
  26. 26. © 2016 The Linley Group September 14, 2016 You have quesAons? We have answers.

×