Set micro-network to launch integrated circuit micro-letter public number: ' Every day IC ', major news immediately released, every day IC, every day set micro-net, product micro-Cheng! Replication Laoyaoic Micro Letter public number search adds attention.
1. Qualcomm Snapdragon 845 big renovation hidden smartphone next year new trend;
Qualcomm (Qualcomm) 's High-profile Snapdragon 845 finally unveiled the mystery, as the expected new processor has greatly improved in many ways, from which it can glimpse Qualcomm's vision of the 2018 smartphone market, and the outside world can grasp some clues to the market's new trends over the next 1 years. Snapdragon 845 uses Samsung Electronics (Samsug Electronics) 10 nm process, and Qualcomm uses the new Kryo 385 central processing Unit (CPU) to make some significant improvements in the new processor. Kryo 385 uses arm (ARM) The latest A75 and A55 CPU core customer-made version, A75 is faster than the previous version of CPU processing speed 30%, the time pulse 2.8 GHz. Qualcomm also in Snapdragon 845 implanted 2MB L3 cache memory, To improve CPU efficiency and hysteresis. Kryo 385 also contains arm DYNAMIQ system controllers that can further improve efficiency and hysteresis, especially for performing tasks such as artificial intelligence (AI). The performance of the graphics processor (GPU), or Qualcomm's so-called ' visual processing subsystem ' (visual processing subsystem) is also upgraded. Qualcomm's new Adreno 630 is a new generation GPU with a completely redesigned architecture that improves performance by 30% compared to Snapdragon 835, and saves 30% power when the same performance is maintained. Overall, Snapdragon 845 still uses the Samsung 10 nm process, but both performance and power consumption have improved dramatically. Snapdragon 845 's Imaging signal processor (ISP) Spectra 280 also has a noticeable boost, with the new ISP focusing more on the quality and color gamut of the paint, giving photos and videos a more accurate, vivid and realistic picture. In addition, Qualcomm pioneered the Ultra HD Premium film recording, capable of filming 4K HDR films, once again leading the film recording technology. It is speculated that with the support of Spectra $number ISP, Snapdragon 840 will be able to record HDR movies with mobile phones with support for HDR displays. In addition, Snapdragon 845 also integrates the new hexagon 685 Digital signal Processor (DSP) to speed up the processing of different workloads, such as AI inference, with relatively low power consumption, which means that the use of Snapdragon 845 handsets can support more AI architectures. In the area of security, Qualcomm uses a completely new independent security module (secure Processing unit), which separates the security features of the Snapdragon 845 mission key from the rest of the SOC, including user passwords, Digital Key Management for application data and other user-sensitive information. Because more mobile subscribers use mobile payments and store user biometric data, Qualcomm does have to improve the processor's level of security protection. In addition, Qualcomm also promotes Snapdragon 845 's networking function. Network connectivity has always been a high throughput processor, Snapdragon 845 integrates its 2nd generation Gigabit LTE Solution Snapdragon X20 LTE data machine, compared to X16 Gigabit class LTE data machine, peak speed 20%, to 1.2 Gbps, Download a 3GB movie can be completed in less than 3 minutes. Snapdragon 845 also uses the new Bluetooth Tru wireless technology to solve a big problem with Bluetooth connectivity, with virtually unlimited connectivity to multiple devices, and for users around many Bluetooth devices, Snapdragon 845 processors will bring convenience. From the Snapdragon 845 renovation of the situation, the mobile phone will be more than the past 1 years to launch the product superior, not only faster processing speed, energy efficiency and overall performance also enhanced, regardless of the implementation of AI functions or visual quality is expected to leap forward, The 2018 market's flagship phone seems to be worth looking forward to. Digitimes
2.2017 years of global semiconductor products, manufacturing equipment and materials sales will be a record high;
With the expansion of the types and quantities of various terminals in semiconductor products, it not only promotes the growth of the market demand of all kinds of semiconductor products, but also makes the demand of semiconductor manufacturing equipment and materials ascend. Coupled with the increase in average price (ASP) of products such as memory, and the role of ASP, such as silicon raw materials, the international semiconductor Equipment Materials Industry Association (SEMI) estimates that 2017 global semiconductor products, manufacturing equipment and materials sales will hit a record high. As for semiconductor products, global semiconductor sales are expected to hit a record $400 billion trillion for the first year in 2017, as the average monthly sales of semiconductors in the world have increased for 15 consecutive months and 8 months in a row. Data showed that global semiconductor sales topped $300 billion in 2013 for the first time, to 3,055. After $8.4 billion, sales grew by another $100 billion in 4 years. By contrast, the global semiconductor market has grown from $200 billion to $300 billion and has been in the 13-year period (2000-2013 years). The $100 billion, which grew to $200 billion, went through 6 years (1994-2000 years). In addition to semiconductor products, the estimated 2017 global semiconductor manufacturing equipment and materials sales will also break the 2000 and 2011 record, and then innovation High. Data show that 1987-1994 years of global semiconductor manufacturing equipment and materials market size roughly equal. However, with the rise of the 8-inch (200mm) wafer process in the mid 90, the semiconductor manufacturing equipment market began to rise rapidly and fell sharply after a record of $480 billion in 2000. 2001 was smaller than $300 billion, and 2002 was about $200 billion. Since then, the market size has rebounded, but it has been cyclical and has never surpassed the 2000 record. However, the estimated 2017 global semiconductor manufacturing equipment sales will be expected to reach $550 billion trillion, a new record. The scale of the global semiconductor material market, which fell in 2001 after a surge in 2000, began to resume growth in 2002, compared with the large fluctuations in the size of the semiconductor manufacturing equipment market. After 2000 years of 2004, it set a new record in 2011 (only 2009 had slipped). Subsequent declines in sales, as well as temporary fluctuations, have not been significant, and have started to grow again in 2016. Estimates of 2017 global semiconductor material sales will surpass the 2011 record. Semi said that for the material market, the key factor that led to sales decline is the decline of the material itself ASP, which is the decline of silicon raw material ASP. As a result, shipments of silicon alone have reached a record high, but the raw material sales are still far below the 2007 peak. In addition, although the 2017 silicon raw materials ASP has rebounded, but still only more than half of the 2008 ASP. In the global semiconductor manufacturing industry, with the rise of TSMC, Samsung Electronics and other semiconductor manufacturers in Asia Pacific, the focus of the global semiconductor manufacturing industry has shifted from Japan to the Asia-Pacific region over the past 14 years. Japan remained the world's largest semiconductor manufacturing equipment and materials market in 2003, accounting for 26% of the world. However, 2017 years of Taiwan, South Korea and mainland China will share the global market 61%, significantly higher than the 2003 33%. The proportion of Japanese territory will fall to just 13%. Data show that in 2009 Taiwan squeezed Japan into the world's largest semiconductor manufacturing equipment and materials market. In 2010, South Korea again squeezed Japan, the world's second largest market, Japan retreated to the third. 2016 the mainland again squeezed Japan, the world's third largest market. South Korea is expected to overtake Taiwan in 2017 as the world's largest market for semiconductor manufacturing equipment and materials. The mainland will remain in the third largest market position. Digitimes
3. Intel next year AI, IoT and 5G new battlefield firepower fully open;
Intel's more than a year of layoffs have finally come to an end, with the consolidation of manpower and strategic objectives, the gradual increase in operations, Better-than-expected performance in the 3rd quarter, and a marked pick-up in earnings over the same period in 2016, with Intel optimistic about the 4th quarter and Full-year performance. Not only is Intel still dominant in the PC and data center platforms, but it is also fully open to new battlefields such as artificial intelligence (AI), Internet of Things and 5G, with the combination of technology, marketing and eco-chains, which is expected to offset Intel's decline in the traditional PC sector in 2018. Intel announced in April 2016 that it would lay off about 11% of its workforce, when about 130,000 of Intel's global staff estimated that some 12,000 people had been abolished, leading to a disruption in supply chain management, and Intel had been operating gradually since the second half of the transition period from 2017 onwards. Supply chain operators believe that despite the PC market growth, Qualcomm (QUALCOMM), Ultra Micro (AMD), Nvidia and other aggressively grabbed into the data center platform battlefield, but Intel offerings of a number of innovative technology and platform products, and with the global supply chain close cooperation, the rival group has so far difficult to shake the Intel PC, Data center platform two big profit prop. In the PC battle, Intel decided to join hands to suppress Nvidia sitting big, check and balance its drawing chip the city dominates the electricity market situation, and compared to the previous and Nvidia signed GPU licensing mode, Intel and the micro-cooperation does not explain too much, only announced the use of embedded multi-crystal Interconnect bridge ( EMIB) Packaging Technology, integrated H-series processor Kaby Lake, micro-independent graphics card and HBM2 memory. Intel more angle-digging ultra-micro drawing chip The Koduri, Raja, in addition to being appointed as the chief architect and also the new core and vision computing business and Edge Computing solution, Intel will restart its stand-alone graphics chip business to deter Nvidia's attempts to dominate the high margin electricity market. Intel by the NB platform processor city accounted for 90% advantage, as well as the 80% DT map, coupled with the overall strengthening of the graphics chip technology research and development strength, 2018 from the PC market revenue, profit should be able to keep a flat score. And in the development of the AI and the Internet Key Data center platform, Intel stability in more than 90% territory, in recent years, including NVIDIA, Qualcomm, Micro and IBM in succession to the Intel post, in which Nvidia stressed that its GPU parallel computing technology to win the performance of Intel, And the Ultra micro also starts with the brand-new EPYC series processor, IBM is the lock AI application launches the next generation processor Power9, as well as the first support OPENCAPI Next generation Data Center Server architecture's power BAE AC922, It can greatly shorten the training time of deep learning artificial neural network. Google in 2016, the first release of the machine learning chip TPU, the recent Diane Bryant, Intel Data Center Business general, Google development AI own chip and control the big cloud of data war attempt is quite obvious. While the various centaurs are fully in the data center platform battlefield, as Intel's 3rd quarter earnings show, the data center platform business revenue continues to grow, Intel has the technology and close eco-chain integration advantage, the opponent group is still difficult to break down in the short term, Intel 2018 Data Center performance will continue to grow. Looking forward to 2018, Intel, in addition to continuing to master the PC and data center platform market advantage, in the automatic driving, robotics, intelligent city, such as AI and IoT new battlefield will be full fire, have the opportunity to offset the decline of the traditional PC sector, the annual performance is flat upward. As for the 5G technology war that will erupt in 2020, Intel has already deployed in advance, from terminal to cloud all with 5G focus, a number of important technologies are being developed, such as NFV technology has been built, for 5G gradually lay the foundation. In addition, Intel is also involved in a number of 5G standards, and with a number of telecommunications and equipment manufacturers to set up the Intel Receptacle Builders Network Eco-Cooperative group, including equipment, software, application services, such as more than 200 companies to build 5G network, development and testing, such as dynamic network slices, Multi-access edge operation, millimeter wave technology, and jointly promote the 5G and various services. Digitimes
4.NVIDIA open GPU cloud service for AI researchers;
NGC further imports Nvidia TENSORRT inference Accelerator, ONNX compatibility mode and supports Mxnet 1.0
The AI researchers using the desktop GPU will now be able to capture the powerful computing power of the NVIDIA GPU Cloud (NGC) through NVIDIA Titan, and announce the expansion of the NGC feature to import new software and other important features into the container, providing researchers with a broader range of , a more powerful tool portfolio to help promote the research and development of AI and efficient computing.
NVIDIA Pascal architecture Titan GPU users can register the NGC account for free and get a complete range of software and tools for GPU optimization for depth learning and HPC. Other supported computing platforms include Nvidia DGX-1 and DGX redevelop and solutions driven by the Nvidia Volta architecture on Amazon EC2.
With the rapid expansion of the NGC container login service, its supporting software includes the TensorFlow and Pytorch Advanced Learning Framework for NVIDIA, Third-party-managed HPC applications, nvidia HPC professional visualization tools and the programmable inference accelerator nvidia TENSORRT 3.0.
The first wave of NGC companies, including GE Healthcare, is also the first medical device manufacturer to use NGC, which will use the deep learning software in the NGC container to accelerate the delivery of the most sophisticated AI systems to 500,000 of imaging devices around the world to improve medical effectiveness.
New NGC container, update and feature Nvidia in addition to supporting TENSORRT on the NGC container login service, the following related updates are announced: • Import open neural Network Exchange (ONNX) to TENSORRT Support and Supply Mxnet 1 First edition • Support Baidu Paddlepaddle AI framework
Onnx stems from a set of open software designed by Facebook and Microsoft for developers to model exchange in different frameworks, and in the TENSORRT development container, Nvidia constructs a converter to import the ONNX model into the TENSORRT inference engine, Allow developers to easily deploy models with low latency and high yield.
These additional resources provide a one-stop service for developers to support a variety of AI computing requirements from research and application development to training and deployment.
The October-line NGC is open free to users who perform operations on Amazon Web services through the NVIDIA Volta GPU, as well as all Nvidia DGX-1 and DGX station users. Nvidia will also continue to expand NGC. Eettaiwan
5. Samsung, LG expanded software talent for the development of AI, robot career preparation
In order to meet the advent of Industry 4.0 new era, Korea Samsung Electronics (Samsung Electronics), LG Electronics (Electronics) recently actively reinforcing artificial intelligence (AI), robotics and other related fields of experts and software personnel, Expect to be able to pre-empt the future development potential of the technology emerging areas. Korean media newspim quotes the South Korean industry news, pointed out that a little earlier November Samsung Life Home Appliances ministry, is expected to hire a large number of database design management, service environment and platform development, diversification Platform Module development and related fields of research and development personnel, is expected to be published in January 2018 hire list. Samsung related Personage said, because Samsung Medium and long term target for home appliance product Thing Network (IoT), increase carry AI function, therefore is strengthening the related link technology, and will manage to the user's data, therefore plans to expand the related domain talented person to use. Samsung's strategy is intended to give all home appliances ' smart ' functions by 2020, expand product connectivity, and expand the application of ' Samsung Connect '. The so-called ' Samsung Connect ' is a device-class and operating system (OS) that controls all products with an integrated single application (APP) through cloud services. In addition, Samsung also intends to use a general conversational voice service, allowing users to manipulate the will can be directly understood by the machine. Through such an AI platform, the surrounding images can be analyzed, and the understanding of the correct meaning, and even can be predicted, the biography of Samsung is committed to the development of such technology. In addition to the home appliances with internet function, the relevant technology can also be used for self-driving intelligent vehicle and service robot. Samsung has also organized fine-tuning, finished product research and development in charge of the Unit DMC Institute and Software Center for the ' Samsung research ', but for the industry rumors of service-oriented robots, Samsung has said no relevant research and development plans are underway. LG recently also published the year-end personnel and organization, including H&A Business Headquarters Intelligent Solution to the Liu Hui promotion of special services, LG, this is also a large-scale promote AI and other software field experts, the relevant Shuo, doctoral talent will be completed in mid-December. LG plans all future products have AI and cloud function, is also active from the outside to extend the software talent, the original investment is expected to expand twice times before 2020, research and development manpower will be expanded more than 50%. It is understood that Lekine employs a range of robotics-related technologies, including mobile and communication technologies. At present, LG's robot career has some achievements, such as cleaning robot, navigating robot and weeding robot, navigation robot products have started in Korea Incheon Airport and shopping Mall services, February 2018, South Korea Pingchang Winter Olympics, is expected to see the gold Tour robot figure. At the beginning of 2017, Samsung also began to strengthen the robotics core technology research and development, and began to extend the relevant field of the top experts. Samsung related people said Ai, large data and IoT, Industrial 4.0 industry Revolution Core technology and the current technology and product integration, related business operation model is floating on the table. Digitimes