[Introduction] When the expert reaches out, he knows if there is any. UK's ImaginaTIon's latest neural network accelerator PowerVR 2NX NNA performance is closest to the competition 2 times, whether you are DSP + hardware, VPU + hardware or neural network DSP. It is expected to be the best choice for HiSili's next-generation Kirin chip.
Artificial intelligence has risen to the future strategic competitiveness of the country today because it will penetrate key equipment in all walks of life and become the core competitiveness of these devices. For example, today Huawei has already surpassed Apple's latest iPhone 8 by virtue of artificial intelligence technology. The neural network processing unit closest to the human brain computing model in the future will be the best choice for next-generation device machine learning and deep learning.
Chris Longstaff, senior director of product and technology marketing at Powermag, ImaginaTIon, said: "With the consideration of power, bandwidth, performance, reliability, security and latency, the next generation of artificial intelligence devices is unlikely to be implemented in software and cloud computing. Hardware for machine learning and deep learning. Neural networks are set to become the key processing unit for the next generation of artificial intelligence SoCs. In the future, it will be ubiquitous in SoCs like CPUs, GPUs and video codecs."
Neural networks, including convolutional neural networks (CNN), recurrent neural networks (RNN), and long- and short-term memory networks (LSTM), are driving explosive technological advances in various artificial intelligence industries. ImaginaTIon, a well-known GPU IP provider in the UK, recently announced a complete, stand-alone hardware IP neural network accelerator in Shenzhen that delivers industry-leading performance, power and area efficiency through the Neural Network-specific PowerVR architecture.
Chris Longstaff pointed out: "The performance, power and area efficiency of the PowerVR 2NX NNA Neural Network Accelerator is two times closer to the competition, regardless of whether the competitor is using DSP+ hardware, GPU+ hardware or neural network DSP."
Today, companies developing SoCs for the mobile, surveillance, automotive and consumer systems markets will be able to integrate high-performance computing with the new PowerVR 2NX NNA neural network accelerator with minimal power consumption and minimal chip area.
He exemplified that today's drones fly at speeds of more than 150 miles per hour. If there is no hardware processing unit and cloud computing is used for collision avoidance, it will require an anti-collision distance of 10-15 meters. If implemented with the PowerVR NNA SoC, you can avoid objects within a distance of less than 1 meter, and you can run multiple networks to identify and track objects while avoiding collisions.
Another obvious example is that if you use a high-end CPU to search and sort 1000 photos on a mobile device, it takes 60 seconds, and with PowerVR 2NX NNA, it takes only 2 seconds.
NNA is a basic type of processor that is expected to be as important as CPU and GPU, and both types of ImaginaTIon are available. NNA's potential applications are numerous, including feature detection and eye tracking in AR/VR headsets; pedestrian detection and driver alertness monitoring in automotive safety systems; facial recognition and crowd behavior analysis in intelligent surveillance systems Web scam detection, content advice, and predictive UX; speech recognition and response in virtual assistant applications; and collision avoidance and object tracking in drones.
"Today, large vendors such as Google, Apple, Facebook, Baidu and Huawei have applied AI/NN to mobile devices, while PowerVR 2NX is designed specifically for mobile and Android," said David Harold, vice president of marketing for Imagination. 2NX is not only the industry's first dedicated neural network hardware solution with support for high-flexibility bit widths from 16-bit to 4-bit, but also the lowest bandwidth neural network (NN) solution with an architecture that supports multiple operating systems, including Linux and Android."
According to the Embedded Vision Developer Survey conducted by the Embedded Vision Alliance (https://) in January 2017, 79% of respondents indicated that they have utilized or are planning to use neural networks to perform their products or services. Computer vision function. As technology continues to advance rapidly, more companies will be able to use neural networks to develop products and services. Imagination's customers have developed and deployed NN-based systems for multiple markets including security, mobile, automotive and set-top boxes.
Jeff Bier, founder of the Embedded Vision Alliance, said: "Many systems and application developers are using deep neural network algorithms to bring new sensing capabilities to their products. In many cases, the main challenge is to be able to address these requirements. Rugged algorithms provide sufficient processing performance while meeting stringent product cost and power constraints. Dedicated processors like PowerVR 2NX NNA are designed for neural network algorithms and will be able to drive these in many new applications. Deployment of powerful algorithms."
With the growing popularity of neural networks, dedicated hardware solutions such as 2NX NNA – which delivers 8x performance density improvements over pure DSP solutions – will deliver the highest performance at the lowest power and cost. . In addition, neural networks have historically been very bandwidth intensive, so the bandwidth requirements of memory will increase as the size of the neural network model grows. This creates significant challenges for SoC designers and OEMs when designing systems because of the bandwidth required to provide NNA. PowerVR 2NX minimizes the bandwidth requirements of external DDR memory and ensures that system performance is not limited by bandwidth. The popularization of PowerVR 2NX NNA-specific hardware will drive the development of various neural network-based applications.
PowerVR 2NX is a newly designed architecture from scratch that provides:
• Industry's highest unit milliwatt inference (mW) IP core to provide the lowest power consumption*
• Industry's highest in-plane/inference (mm2) IP core for the most cost-effective solution*
• Industry's lowest bandwidth solution* – support for weight and data flexibility, including low bandwidth modes down to 4 bits
• Industry-leading performance with a single core of 2048 MAC per cycle and increased performance with multi-core design
• Provides the most comprehensive system-level neural network solution based on hardware architecture
Chris Longstaff, senior director of product and technology marketing at Imagination, said: "The dedicated hardware for neural network acceleration will be the standard IP module for future SoCs, just like CPUs and GPUs. We are happy to be the first to introduce a complete hardware accelerator to the market. Supporting high-precision and flexible methods, the neural network can operate with the lowest power consumption and bandwidth, and provide absolute performance and unit area performance superior to the competitor's solution. The tools we provide will enable developers to enable and deploy very quickly. Executing its neural network can quickly increase revenue."
2NX includes hardware IP, software and tools to provide a complete neural network solution for SoC. It effectively performs all common layers of neural network operations. Depending on the computational needs of the inference task, it can be used independently - without additional hardware - or with other processors such as CPUs and GPUs.
PowerVR 2NX NNA is designed for inference engines that adapt to multiple markets, and its highly scalable architecture is suitable for more application markets in the future.
Mobile: With the upcoming release of Tensorflow Lite and the Android API, coupled with the rapid advancement of the Caffe2Go architecture, we will see a rapid explosion in the number of smartphone AI applications. Companies need efficient ways to perform a variety of reasoning tasks, including image recognition, speech recognition, and computational photography. PowerVR 2NX is the only IP solution available today that meets all the needs of mobile solution deployment with low power, small area, MMU and support for Android. In mobile devices where GPUs are indispensable, companies can use the new PowerVR Series9XE or 9XM GPUs in the same chip as 2NX NNAs as stand-alone GPUs. Its outstanding performance is enough to compete with the industry's stand-alone GPU.
Intelligent Surveillance: The number of surveillance cameras installed in homes and businesses is increasing dramatically, driving the need for neural network vision processing. Intelligent cameras based on these technologies can be used to make decisions based on security alerts, retail analytics, demographics, and interactive data. Considering bandwidth requirements, data confidentiality, and other factors, camera design must incorporate a degree of "edge" video information analysis and processing capabilities. Since these cameras typically do not have GPUs or only very small GPUs, as well as lower performance CPUs, they require an efficient, high performance, stand-alone neural network accelerator. The 2NX NNA is the ideal choice and highly scalable to meet both consumer and commercial needs.
Automotive: Automotive neural network applications include driver alertness monitoring, driver sight tracking, seat usage, road sign detection, driving route analysis, road user detection, driver identification, and more. These applications will continue to expand as the number of self-driving cars and intelligent transportation systems will increase in the coming years. In automotive systems, a complete hardware solution such as 2NX NNA is required to meet the relevant performance requirements.
Home entertainment: Set-top boxes and televisions will increasingly provide neural network-based solutions, such as the ability to adapt to specific user preferences, provide automatic child locks, and automatically pause and record programs based on user behavior. With these features, companies can increase their differentiation and increase revenue. The key to building a neural network on these devices is efficient bandwidth and low cost, as well as support for the NN API - these are important features of the 2NX NNA. In addition, there are many emerging NNA entertainment applications, including AR/VR.
Tony Chia, co-founder and vice president of mobile VR and AR, said: “The innovative technology from partner Imagination is to ensure that VR solutions provide consumers with a high degree of immersion. Key to game and visual experience. We are very excited about the potential application of neural networks in VR/AR, including tracking, control, gesture recognition, speech recognition, and more interactive experiences. Built-in PowerVR 2NX NNA chip The group provides the processing power and power efficiency required to push the VR/AR experience to a whole new level."
Simplify the work of developersImagination provides developers with all the tools they need to quickly and easily enable and execute their neural networks and ensure that computing bandwidth is perfectly balanced with accuracy. PowerVR 2NX development resources include mapping and fine-tuning tools, sample networks, evaluation tools and files. The complete PowerVR NX Mapping Tool easily converts from industry-standard machine learning frameworks, including Caffe and Tensorflow. Advanced network designers will be able to design and build neural networks on 2NX NNA to take full advantage of their hardware features.
Imagination also offers a universal Imagination DNN (Deep Neural Network) API that easily converts between CPU, GPU and NNA. This single API can be executed across multiple SoC configurations, making it easy to prototype on existing devices.
The PowerVR 2NX NNA is now available for licensing.
Imagination Releases PowerVR NNA Neural Network Accelerator
Provides twice the performance and only half the bandwidth compared to the closest competing solution
About Imagination TechnologiesImagination is a global technology leader whose products are used by billions of consumers around the world. Imagination has a complete portfolio of silicon IP (silicon intellectual property) products, including the key processors needed to create SoCs (system-on-a-chip) that can be used to develop a wide range of mobile, consumer and embedded electronics. The company's unique multimedia, processor and connectivity technologies enable customers to develop highly differentiated SoC platforms and bring products to market quickly. Imagination's authorized customers include leading semiconductor manufacturers, network vendors and OEM/ODM vendors that have developed the world's most iconic products.
ZGAR AZ MC Disposable
ZGAR electronic cigarette uses high-tech R&D, food grade disposable pod device and high-quality raw material. All package designs are Original IP. Our designer team is from Hong Kong. We have very high requirements for product quality, flavors taste and packaging design. The E-liquid is imported, materials are food grade, and assembly plant is medical-grade dust-free workshops.
Our products include disposable e-cigarettes, rechargeable e-cigarettes, rechargreable disposable vape pen, and various of flavors of cigarette cartridges. From 600puffs to 5000puffs, ZGAR bar Disposable offer high-tech R&D, E-cigarette improves battery capacity, We offer various of flavors and support customization. And printing designs can be customized. We have our own professional team and competitive quotations for any OEM or ODM works.
We supply OEM rechargeable disposable vape pen,OEM disposable electronic cigarette,ODM disposable vape pen,ODM disposable electronic cigarette,OEM/ODM vape pen e-cigarette,OEM/ODM atomizer device.
Disposable E-cigarette, ODM disposable electronic cigarette, vape pen atomizer , Device E-cig, OEM disposable electronic cigarette
ZGAR INTERNATIONAL TRADING CO., LTD. , https://www.szvape-pods.com