Gathering micro-messages, almost all astrophysicists believe that the starting point of all things is BigBang - of course, what we are talking about is not that famous music combination, but a real expansion in the void. 13.7 billion years ago , The universe is formed by the expansion of a dense hot ball after the explosion, 'time' was born, the ball, is the first singularity.
Afterwards, as human beings develop in a similar way to acceleration, the more they multiply or even increase geometrically, and futurists now begin to predict whether the next singularity will arrive at 2045 or 2030? It is not important, but importantly, who will drive breakthroughs in basic research and development, especially basic research and development in information technology, which are closely linked to the next 'singularity'? 2018 MWC, we can see some clues.
5G Sprint: Qualcomm Completes Real Network Simulation Experiment
As a key infrastructure, 5G will produce up to $ 12 trillion worth of products and services by 2035. Qualcomm is an important contributor to 5G foundational technology and its chief executive, Steve Morinkov, said " The impact of the world economy will be similar to the emergence of electricity or cars.
During the MWC, Qualcomm released a number of significant results from its extensive 5G real-network simulation over the past few months, which led to significant interest in Qualcomm choosing between Frankfurt, Germany and San Francisco, USA, at 3.5GHz and 28GHz respectively The 5-M network simulation was conducted in the millimeter-wave band, and the final results showed that both 5-G simulations achieved a 5x network capacity gain versus the current state-of-the-art Cat. 20 gigabit LTE network The experiment also achieved 7 times and 23 times the response speed improvement, significantly improve the application experience.In the 5G millimeter wave simulation experiments in San Francisco, the average web browsing download speed average 1.4Gbps.In addition to the download speed and response speed improvement, 5G can also improve the quality of streaming video.Compared with 2K, 30 frames / second and 8-bit color video enjoyed by LTE users, 5G users can enjoy 8K, 120 frames / second and 10-bit color in simulation experiment Streaming video experience, not only for users will be a huge change, but also to provide developers with enhanced VR experience and video experience a vast space.
With a breakthrough in basic inventions, Qualcomm is still thinking about how to quickly share 5G technology with its partners. In the 5G era, because of the complexity of technology, handset manufacturers may have to build their end products with over 1,000 components. How to solve this problem? Qualcomm introduced the Qualcomm Snapdragon 5G Module solution during MWC and integrated more than 1,000 components in several modules covering digital, RF, connectivity and front-end functions including application processors, baseband modems, memory, PMICs, RF front-ends, antennas and passive components are all expected to ship in 2019. This means Qualcomm became the first company to deliver an epoch-making, turnkey, turnkey 5G modular solution for the business; it also means that terminal manufacturers can Faster production with less cost and less time.
The world's first release of the 5G chipset - Qualcomm Snapdragon X50 5G new air interface modem, has been the world's 20 terminal manufacturers selected to support them to create the first batch of 5G smart terminals, is expected to be released in 2019. These 20 Vendors include ASUS, Fujitsu, Fujitsu Connectivity Technologies, HMD Global, HTC, Inseego / Novatel Wireless, LG, NetComm Wireless, NETGEAR, One Plus, OPPO, Sharp, Sierra Wireless, Sony Mobile, Telit, Vivante Technology, Kai Technology, Xiaomi and ZTE.
In the meantime, several wireless network operators worldwide have also selected the Snapdragon X50 5G modem for 5G new air-port mobility trials below 6GHz and in the millimeter-wave band, including AT & T, BT, China Telecom, China Mobile, China Unicom , Deutsche Telekom, KDDI, Korea Telecom, LG Uplus, NTT DOCOMO, Orange, Singapore Telecom, SK Telecom, Sprint, Telstra, TIM, Verizon and Vodafone.
Interworking between different vendors is the most important part of industrial cooperation and an important step before 5G commercialization.In November 2017, Qualcomm, ZTE Corporation and China Mobile successfully achieved the world's first end-to-end 5G based on the 3GPP R15 standard During the MWC, Qualcomm also demonstrated 5G new air interface interoperability testing based on the 3GPP R15 standard with five equipment manufacturers including ZTE, Huawei, Ericsson, Nokia and Samsung, covering the following 6GHz and millimeter-wave frequency bands.
China's three operators and many well-known Chinese terminals and equipment manufacturers are choosing the depth of cooperation with Qualcomm, which can also be seen Qualcomm for the development of China's wireless communications ecosystem attaches great importance to.In fact, just in January of this year, Qualcomm also A number of Chinese manufacturers have announced the '5G Pilot Program' to express global opportunities for 5G and strengthen cooperation to better support the upgrade of China's smart phone industry towards higher end and speed up the sea.
The good 5G experience of the future is also inseparable from the continued evolution of 4G LTE.MWC eve, Qualcomm also released a substantial increase in user experience 'artifact' - Snapdragon X24 LTE modem, 4G LTE brought to '2Gbps' era , Which is also the world's first chip to be released based on the 7nm FinFET process technology, which is also the third generation of Qualcomm's Gigabyte LTE chipset. On February 20, GSA released a report on the current status of LTE and 5G chips Research report, which states that the current fastest downlink is the Snapdragon X24 LTE modem.
AI: to the terminal side
AI (Artificial Intelligence) is another important human technology to achieve the next singularity, which gives the ability to learn different types of terminals.Traditionally, many of the neural network training and reasoning are done in the cloud or server-based Over the past few years, the entire model has undergone tremendous changes. Many artificial reasoning tasks, such as pattern matching, model checking, classification, recognition and detection, have gradually shifted from the cloud to the terminal side. This is very important for protecting data privacy and improving performance Overall reliability is of great benefit.
During the MWC, Qualcomm's AI Engine, an artificial intelligence engine announced, is a collection of hardware and software that accelerates terminal-side AI user experience on the Snapdragon mobile platform.Multiple programmable heterogeneous computing cores integrated in Snapdragon processors- Qualcomm Hexagon vector processors, Qualcomm Adreno GPUs and Qualcomm Kryo CPUs offer partners a broader choice of flexibility to support different AI functions, data and computational accuracy requirements, enabling the rapid and efficient execution of AI applications on the terminal side. The second component of AI Engine is software and tools, and software and tools are certainly more important when dealing with AI use cases.Hardware is an accelerating platform, and at the software level developers are able to do a great deal of innovation and build the latest AI use cases .
Currently, a number of smart phone manufacturers have used artificial intelligence engine AI Engine on Snapdragon mobile platforms to speed up artificial intelligence applications on their terminals, including millet, one plus, vivo, OPPO, Motorola, Asus, ZTE, Nubian , Hammers and black sharks, some of which are planning to use the AI Engine, an artificial intelligence engine, to optimize AI applications in its flagship Snapdragon smartphone of the future.
Based on its AI engine, AI Engine, Qualcomm and leading AI software development companies and leading cloud service providers have also established in-depth cooperation to bring exclusive use-case optimization to Snapdragon mobile platforms such as Shangtang Technology and Kuang Shi Face ++ A variety of pre-trained neural networks that support image and camera features include single-camera background blurring, facial unlocking and scene detection and recognition, etc. Tencent recently launched an interactive game named 'High Energy Dance Room' on its mobile QQ social platform Feature in the Android side of the phone QQ, the use of artificial intelligence engine AI Engine components to speed up the characteristics of the frame rate.In addition, Baidu also plans to fully support Qualcomm artificial intelligence engine AI Engine and its ecosystem.
Various types of terminals: to enhance the experience of the most critical
In this MWC, Qualcomm Snapdragon 845 mobile platform was awarded the annual GTI mobile technology innovation breakthrough award.Selecting party, Snapdragon 845 give full play to Qualcomm's industry-leading mobile heterogeneous computing expertise designed for immersive multimedia experience and design, It will also serve as a top-level artificial intelligence platform that extends the way people interact with the world.
During the show, Samsung Galaxy S9 and S9 +, Sony Xperia XZ2 and Asus ZenFone 5Z have announced the Snapdragon 845 mobile platform; and Snapdragon 845 will also support the subsequent release of millet MIX 2S and a series of domestic mobile phone.
In addition, Qualcomm announced the launch of the Snapdragon 700 Series mobile platform during the Barclays Festival as another highlight of Qualcomm's MWC in an effort to bring top-of-the-line technology and features to affordable endpoints. "Qualcomm said," From our top Qualcomm artificial intelligence Engine AI Engine With outstanding camera, terminal performance and power consumption, the optimized Snapdragon 700 Series will enable consumers to enjoy the experience they expect on the most advanced mobile devices at affordable prices. '
Thanks to the rapid development and accumulation of the smart phone platform, the speed and scale of innovation in the entire mobile technology are constantly improving and expanding.Mobile technology, as the largest technology innovation platform, is transforming and affecting many adjacent industries.
Qualcomm introduced a new virtual reality (VR) reference platform based on the Snapdragon 845 mobile platform to support the next wave of VR and stand-alone VR heads to help manufacturers and developers create immersive applications and the future Experience In the AR and VR space, Qualcomm has supported customers with more than 20 end-user devices including stand-alone head-mounted devices and XR-enabled smartphones, including the head of a VR ecosystem leader such as Google Daydream, Oculus and Vive Batch independent VR terminal.
In the PC space, Asus, Hewlett-Packard and Lenovo have introduced three new Windows 10 devices powered by the Qualcomm Snapdragon Mobile PC platform, which will leverage mobile LTE speed and mobile networks to support mobile computing anytime, anywhere. 'Always Connected PC' Connectivity and simplicity of smartphones combined with the power and innovative capabilities of Windows 10 PCs revolutionize the way consumers work and entertain. During MWC, Qualcomm and Microsoft jointly announced that they will work with the world's leading retailers to consistently connect consumers to sales PC, and with more operators worldwide to bring fast, easy and cost-effective wireless connectivity to always-connected PC products.
In the automotive sector, Qualcomm has been providing technology solutions to the automotive industry for 15 years and is currently the world's largest semiconductor information provider for automotive information processing and automotive Bluetooth connectivity solutions. 14 of the world's top 25 leading automotive manufacturer brands have adopted In addition, Qualcomm is working with the ecosystem of leading carmakers and automotive suppliers to accelerate the commercialization of C-V2X technology, such as the Peugeot Citroen The group and Qualcomm announced at MWC to jointly test the C-V2X communications technology, one of the objectives of which was to test the C-V2X technology and then to achieve direct communication between vehicles as the first step in deploying 5G for automotive applications .
In the field of IoT, Qualcomm has also plowed for many years, the daily number of IoT chips shipped more than 1 million, Qualcomm's IOT business revenues of more than 1 billion US dollars in 2017. At present, Differentiated chipset products, reference designs and more than 30 proprietary platforms.
2.Arm released a new generation of GPU Mali G52 / G31, adding machine learning capabilities
Set micro-grid March 6 reports
Today, Arm officially unveiled its newest generation of Mali multimedia suites, including the new video, display and image processors, in Beijing, which integrates seamlessly with existing DynamIQ-based CPUs and other Arm IPs to fully implement Arm A new generation of solutions for mainstream mobile devices and digital television (DTV).
Specific products include the Mali-G52 GPU for mainstream mobile phones, the biggest highlight of the machine learning capabilities; for entry-level phone Mali-G31 GPU is the smallest product; and Mali-D51 display processor and Mali-V52 video processor.
2017 ship 120 million Mali GPUs
According to IDC reports, more than 90% of intelligent terminals equipped with AI are based on the Arm architecture in today's market.According to official statistics, the current total delivery volume of ARM-based chips exceeds 125 billion, and is expected to reach 2021 Will reach 200 billion. Arm has more than 1,000 technology partners, the areas covered by the chip, network and cloud.
With its high performance and low power consumption, Arm has the absolute advantage in many areas, and in the field of GPUs, only Mali's GPU shipments in 2017 were as high as 120 million.
Arm's senior director of marketing, Ian Smythe, said it has become an inevitable trend for end-devices to support a rich, multi-layered user interface and a broad range of newest applications. More importantly, machine learning is no longer a feature of high-end smartphones There are configurations.Each level users want to easily use all kinds of APP applications equipped with machine learning capabilities, 3D games, mixed reality and 4k content is more and more popular, new games are also constantly emerging in the mainstream mobile phone.In the new generation of user needs Under the technology-driven innovation, Arm launched a new generation of solutions to provide users with a more hearty visual experience.
The new generation of GPU Mali G52 / G31 is released, emphasizing machine learning
First of all, from the GPU roadmap released by Arm, it has been fully deployed in high school and low end, including the G70 series for high-end, the G50 for mid-range, and the G30 for low-end.
Today's release of the Mali-G52 GPU targets mid-range products, but Arm's goal is to bring advanced features to mainstream devices.Compared with the previous generation G51, the G52 can cope with higher graphics complexity, allowing the power of mainstream mobile systems And bandwidth limitations to achieve more machine learning functions.
It is reported that Mali-G52 uses a wider execution engine, compared to the previous generation 4-thread, Mali-G52 provides up to 8 threads, in the same chip area, providing higher graphics performance.
Measured data show that the G52 delivers 30% higher performance density and 15% more energy efficiency than the previous generation, reducing power consumption and heat dissipation in the device, supporting longer gaming times, and even supporting high power-hungry technologies such as AR.
In addition, Mali-G52 GPUs also incorporate machine learning capabilities, and Arm officials say it is impractical to set up machine learning processors in mainstream devices that can be leveraged by each component of the system-on-chip (SoC) Optimized performance and combined with DynamIQ CPUs and Mali-G51 to improve machine learning performance for existing applications.
Another GPU product, the Mali-G31, is a new generation of Arm designed for low-end products that is positioned to deliver a superior user experience in the smallest die area.
The Mali-G31 is the first Ultra-Efficient GPU based on the innovative Bifrost architecture and is currently the smallest Arm processor with support for OpenGL ES 3.2 and the latest generation of Vulkan APIs, which developers can use for several Millions of devices.
In addition, the Mali-G31 has a 20% lower G51 size and 20% higher performance density than the previous generation using the Bifrost architecture, significantly reducing energy consumption while reducing chip area.
Display and video processor
The conference, the reality and video Arm also introduced two display and video processor.
Display Processor The Mali-D51 boasts the many advantages of the Mali-D71, a high-end display processor from 2017, and is integrated into the smallest DPU ever owned by Arm, the first mainstream display processor based on the Komeda architecture , Saving 30% energy over the entire system; doubling the scene complexity and supporting 8 levels of image processing like the Mali-D71; and 50% faster memory access latency for seamless, efficient content projection.
The Mali-D51 is fully optimized to work seamlessly with other IPs in the Mali Multimedia Suite and with Assertive Display 5 technology to bring even HDR (high dynamic range images) to mainstream devices; enhances with CoreLink MMU-600 System memory management efficiency.
The video processor Mali-V52 delivers twice the decoding performance of all major devices for 4K playback including 4K60 / 4K120 content, up to 20% more uploading quality and clearer, sharper video quality than the previous generation.
Arm said that the company is working hard to make more mainstream users feel the perfect visual experience. The release of the Mali Multimedia Suite today is one of Arm's many efforts. (Proofreading / Xiaoqiu)
3. Startup company borrows DNN to infer chips into AI market
The start-up company Gyrfalcon quickly emerged from an inferential chip for deep neural networks (DNN), but also faced an increasingly crowded market competition in the AI chip field...
Startup Gyrfalcon is rapidly emerging with a corollary chip for deep neural networks (DNNs), but it is also facing increasing crowding in the Artificial Intelligence (AI) chip market after the company first won One year after funding, the AI Altiplates chip, which has been available recently, is demonstrated, claiming multiple design orders in the areas of smartphones, security cameras and industrial automation equipment.
Data centers usually perform in-depth neural network training and use a series of servers to perform inference tasks. At present, more and more clients and embedded systems use accelerators to accelerate inference work from fields such as automobiles to mobile phones.
Companies such as Apple, Google and Huawei have introduced in-built SoC smartphones with inference modules, and Google and Microsoft have also built inferences accelerators for their data centers.
Among several suppliers, Movidius, which is currently affiliated with Intel, has begun shipping inference chips, while Cadence, Ceva and Synopsys supply IP blocks. Imagination and ARM also have plans to release. In addition, there are at least a dozen AI start-up companies still in stealth mode, including Groq, a group of former Google chip developers co-founded to infer chips.
Linley Gwennap, head of market watcher The Linley Group, said: "Much of the movement about AI inference chips comes primarily from IP providers; for the most part it will be an integrated chip market."
Gyrfalcon has been promoting its Lightspeeur SPR2801 processor-available chip for Convolutional Neural Networks (CNN) since autumn last Fall. It uses TSMC's 28nm process and 7x7-mm package with tens of thousands of dedicated cores and embedded static Random Access Memory (SRAM) delivers 9.3 TOPS / W efficiency without the need for an external dynamic random access memory (DRAM) These are mainly arithmetic logic units (ALUs) with associated memory and logic for controlling data flow element.
The first generation of the Gyrfalcon Lightspeer 2801, with APiM technology, provides clouds and edges with parallelism of several orders of magnitude more power than the rest of the market for deep training and reasoning tasks. The accelerator card works with M.2 and PCIe Interface, support TensorFlow, Caffe and other open source environment.
The new chip concept came from the idea put forward by its co-founder and chief scientist Lin Yang in his 1988 doctoral dissertation when he conducted a CNN study at UC Berkeley in California The chip is designed to accelerate CNN, which is one of the most widely used neural network applications.
Frank Lin, co-founder and president of Gyrfalcon, said the top five smartphone makers will import the chip in a future flagship handset, a major insurer in China that will be designed into its surveillance cameras and a Shanghai-based Steelmakers will also use the chip in their inspection system.
Yang Lin said: 'We have achieved rapid growth after securing the first tranche in February last year. We verified the FPGA in April and completed the chip manufacturing at TSMC in mid-September.'
Developed in partnership with Socionext and QuickLogic
Gyrfalcon announced in December last year that its chip will be used in the video analytics platform of SoCionext, with its parallel processor. This product developed in cooperation with Socionext, QuickLogic will be available in March, Turkey security monitoring System manufacturer Bilkon Ltd. will be its first customer.
QuickLogic is evaluating the Gyrfalcon chip with its next-generation sensor hub for detecting the possibility of awakening words in various 'hearable' products. Timothy Saxe, chief technology officer of QuickLogic, said: 'In a dozen companies to build inference chips. Of the companies evaluated by QuickLogic, they are closer to our market than other companies. '
Saxe said that artificial assistants such as Amazon's 'Alexa will make people want a wider range of vocabulary for more natural language processing, so we must speed up this process through alliances.'
Gyrfalcon plans to launch the second product to support higher performance in June of this year, and will release the new serial chip before the end of this year. It will also explore the possibility of custom-made low-power 50mW custom chips for a large number of customers.
The start-up company designed a PCI Express card for use with server chips. In addition, it will follow Movidius's footsteps and launch a USB dongle with the new chip in April to attract broader CNN software development. Business
'We hope that not only PhDs can perform this task - this USB dongle supports several pre-loaded neural network models and is very easy to use. ' Yang Lin also assisted China Mobile before joining Gyrfalcon. Promote its international business.
However, he did not reveal how much money Gyrfalcon has received so far from the two rounds of investments made by China, South Korea, Japan and the United States. He said only: 'By 2018, we have raised enough funds to support the overall operation, Design orders and potential orders, expected to be profitable this year, and even hope to breakeven.
So far, the company has 41 employees, most of whom are in Silicon Valley headquarters.
Compile: Susan Hong
(Reference text: Startup Claims AI Design Wins, by Rick Merritt)
4. Samsung Electronics R & D spending over last year 15.5 billion US dollars, net profit of 39 billion US dollars!
According to foreign media reports, according to industry data, Samsung Electronics invested more than 16.8 trillion won (about 15.53 billion U.S. dollars) in research and development last year. Meanwhile, the company's liquid assets still remain at 30 trillion won (about 27.74 billion U.S. dollar), but a decrease of about 1.5 trillion won from the previous year, mainly due to the significant investment Samsung made in terms of facilities and shareholder returns.
Compared with 2016, Samsung Electronics last year's research and development costs increased by 2.01 trillion won (about 1.86 billion US dollars), or 13.59%.
In addition, Samsung's realizable assets are still at 30 trillion won level.Specific, Samsung's current cash and liquid assets totaled 30.55 trillion won (about 28.224 billion US dollars), a decrease of 1.57 compared with 2016 Trillion won (about 1.45 billion US dollars), or 4.88%.
Samsung held short-term financial products 5.69% to 49.45 trillion won (about 45.72 billion US dollars), short-term salable financial assets fell 12.29% to 3.19 trillion won (about 2.95 billion US dollars).
Samsung Electronics net profit hit a record 42.19 trillion won (about 39.01 billion won) in 2017. However, last year's investment in Samsung Electronics was also huge, with only 43 trillion won (KRW 397.6 million) for facilities alone One hundred million U.S. dollars) .
In 2017, Samsung Electronics also repurchased 9.2 trillion won (approximately 8.51 billion U.S. dollars) of stocks and paid 5.8 trillion won (approximately 5.36 billion U.S. dollars) in dividends to shareholders. Under a management policy favorable to shareholders , The liquidity of Samsung Electronics stocks has slipped.
A senior business executive stated: 'Samsung Electronics has decided to return at least 50% of its free cash flow to shareholders within the next three years from this year. Therefore, Samsung Electronics may face the problem of insufficient cash in the future.'
Meanwhile, Samsung Electronics's total assets (including capital and debt) are 301.75 trillion won (about 279.0 billion U.S. dollars), breaking the 300 trillion won mark for the first time in more than 50 years since its founding.
5.Qorvo Demonstrates the Industry's First 5G RF Front-End Module at the 20th GTI Live Demo
Gathering micro-network news, Qorvo and global test, measurement and control system leader National Instruments (NI) tested the first commercially available 5G RF front-end module (FEM). Test demonstration at the 20th GTI seminar in London, UK For the first time.
Qorvo's QMI19000 5G FEM combines a power amplifier and a low-noise amplifier in a single package dedicated to mobile devices operating in the 3.4 GHz spectrum. FEM testing is performed using an advanced NI PXI system that can be implemented to some degree Helps customers design and test 5G technology to deploy 5G early on in mobile devices.
With the flexibility of the NI PXI system, Qorvo's 5G FEM designs can be quickly tested using a variety of waveforms, and the same hardware can be reused. The PXI test system is based on NI Vector Signal Transceiver (VST) technology for smarter implementation Automated Testing. NI's software-centric approach helps semiconductor customers meet the toughest test coverage, measurement quality, and product launch requirements.
Paul Cooper, Qorvo's mobile product network operator and director of standard development strategies, said: 'The NI PXI test system has a wide bandwidth, excellent RF performance and flexibility, and plays a vital role in helping us introduce the industry's first commercial 5G FEM. Qorvo has always been committed to innovation, as demonstrated at the 20th GTI Symposium in London.
Qorvo's high-performance RF solutions simplify design, reduce product footprint, save power, improve system performance, and accelerate the deployment of carrier aggregation technology. Qorvo combines system-level expertise, extensive manufacturing scale, and the industry's broadest portfolio of products and technologies. Help leading manufacturers accelerate the release of next-generation LTE, LTE-A and IoT products. Qorvo's core RF solutions set the standard for next-generation connectivity and provide unmatched integration and performance at the heart of the connected world.