Imagination launches a new graphics processor IMG A series that can be applied to any scene

Shanghai, China, December 3, 2019-Imagination Technologies announced the launch of its tenth generation PowerVR graphics processing architecture IMG A series (IMG A-Series). As the most powerful graphics processor (GPU) semiconductor intellectual property (IP) product it has ever released, the IMG A series develops and evolves the PowerVR GPU architecture to meet the graphics processing and computing requirements of various next-generation devices . The IMG A series is designed as a “GPU that can be applied to various scenarios”. It is the ultimate solution for a variety of vertical applications. It can support everything from automotive electronics (including assisted driving/autonomous driving), AIoT, computing to digital TV/ Applications such as set-top boxes/digital TV boxes (DTV/STB/OTT), mobile devices and servers.

The IMG A series achieves performance scalability from multiple dimensions, covering components that process 1 pixel (1 PPC) per clock cycle for the entry-level market, to high-performance devices with a speed of up to 2 trillion per second The core of sub-floating point arithmetic (2 TFLOP), and even multi-core solutions for cloud applications that exceed this value.

Dr. Ron Black, CEO of Imagination Technologies, said: “Since the launch of the first mobile PowerVR GPU 15 years ago, the IMG A series is our most important new GPU product and the best mobile GPU IP ever. In all application markets , It can provide the best performance with a low power budget for a longer running time. It is indeed a GPU that can be applied to all devices.”

The IMG A series provides significant improvements in every dimension. In the same clock and semiconductor process, compared with the PowerVR equipment currently being shipped, its performance has increased by 2.5 times, and the machine learning processing speed has increased by 8 times. , And the power consumption is reduced by 60%.

Compared with other GPU IP solutions currently available, the IMG A series can provide higher performance, lower power consumption (compared to competing products with the same clock and semiconductor process) and lower bandwidth (compared to competitors Use the same cache size), and all products occupy a smaller chip area. Its architecture can also provide powerful differentiation advantages, such as ensuring 50% image compression data (lossless in most cases, or visually lossless compression in special scenes).

Jon Peddie, chief researcher and founder of market research company Jon Peddie Research, said: “In fact, for mobile SoC, market leaders have their own GPU technology and are increasing market share at a rate of 5% per year. In order to prevent loss For its potential market, other mobile SoC companies need a convincing GPU to create some real competitiveness. Imagination’s A series products can do this.”

The IMG A series has been licensed to customers in multiple markets, and the first SoC devices equipped with this IP are expected to be available in 2020.

IMG A series IP core

The IMG A series is the highest performance mobile GPU IP ever released by Imagination Technologies, offering four high-performance configurations:

IMG AXT-64-2048 flagship version, with 2.0 TFLOPS, 64 Gpixels performance, and 8 TOPS AI performance.

IMG AXT-48-1536 is for high-end mobile phones; its performance is 1.5 TFLOPS, 48 Gpixels and 6 TOPS

IMG AXT-32-1024 targets high-performance mobile devices and automotive applications; its performance is 1 TFLOP, 32 Gpixels and 4TOPS

IMG AXT-16-512 is geared towards medium and high performance mobile devices and automotive applications; its performance is 0.5 TFLOPS, 16 Gpixels and 2 TOPS

IMG AXM-8-256 is geared towards mid-range mobile devices; its performance is 0.25 TFLOPS, 8 Gpixels and 1 TOPS

For lower-cost market segments, the IMG A series can provide the best area, cost and efficiency:

IMG AX-2-16 is for high-end IoT, entry-level DTV/STB, Display and other applications that require high fill rate; its performance is 2 PPC, 16 GFLOPS and 2 Gpixels

IMG AX-1-16 is geared towards entry-level mobile devices and the Internet of Things. It is the fastest GPU in its class that supports Vulkan; its performance is 1 PPC, 16 GFLOPS and 1 Gpixels

GPU for gamers

Due to the limitation of heat dissipation, most mobile GPUs are difficult to provide a consistent gaming experience, that is, they can run quickly in a short period of time, and then their performance drops to an unacceptable level as the chip heats up, and finally cannot be recovered. Lead to a disappointing user experience. The IMG A series provides consistent and reliable performance and longer game time at a consistent frame rate, avoiding overheating conditions such as clock downtime or even malfunction due to chip overheating.

Based on the Block Delayed Rendering (TBDR) technology, the PowerVR architecture GPU of the IMG A series only needs to draw the content visible on the screen, which gives it inherent advantages in bandwidth and power consumption.

Advanced power control technology using active dynamic voltage and frequency adjustment (DVFS) and deadline scheduling algorithms (Deadline Scheduling algorithms) can provide fast power control with ultra-low response delay.

If some parts of the GPU are not fully utilized or used for processing work, they will be immediately reduced in frequency or even go to sleep to ensure the best power efficiency.

The IMG A series aims to provide game users with the best experience by providing a continuous and stable frame rate.

This requires high GPU power efficiency. For this reason, the PowerVR series GPU is equipped with a firmware-based low-latency DVFS algorithm and has a GPIO control path directly connected to the system.

On this basis, IMG’s strong focus on visual perfect pixels is also reflected in many details, such as industry-leading visual lossless PVRIC compression technology, completely redesigned anisotropic sampling, and other completely redesigned algorithm implementations.

Jane, vice president of Chukong Technologies, said: “Today’s games require powerful graphics processing capabilities. Game operations have accounted for more than 40% of all smartphone usage, so having the best GPU is a key differentiating factor. And the combination of consistent GPU performance and long-lasting battery life, Imagination once again broke the limits of mobile graphics processing technology.”

The IMG A series is supported by an industry-leading tool set and software development kit (SDK), which helps developers optimize their products by providing new features such as hardware-based block heat distribution maps.

HyperLane

The IMG A series uses Imagination’s HyperLane technology; independent hardware control channels are each isolated in the memory, so that different tasks can be submitted to the GPU at the same time to lock the GPU to achieve multitasking. With the help of dynamic performance control, the GPU can distribute its performance among multiple tasks and perform all tasks while making the most of the GPU. Priority HyperLanes technology can provide dynamic splitting; for example, when performing an AI task, it can prioritize graphics requirements, provide it with sufficient computing resources, and use the remaining computing resources to process AI tasks. HyperLane technology can also isolate protected content for rights management. All IMG A series GPUs can support up to eight HyperLanes.

HyperLane also supports a new feature AI Synergy. This new option allows SoC designers to take advantage of the incredible levels of computing power in the IMG A series to accelerate their AI workloads. Through AI Synergy, GPU can provide graphics processing functions while using its spare resources to support programmable AI and fixed functions, that is, the highly optimized Imagination neural network accelerator. AI Synergy provides programmable artificial intelligence in the smallest chip area, while a unified software stack enables flexibility and outstanding performance.

Overview of IMG A series features

The best GPU IP ever

The performance density is 2.5 times higher than the best PowerVR device currently being shipped

·Extensible to be used in all GPU market areas

· Single core option from 1 PPC to 2.0 TFLOPS

·High-performance density can provide manufacturers with the best implementation cost and consumers with the best equipment price

·HyperLane technology includes support for multi-region hardware virtualization technology, which can provide stability, security and durability

·The automotive industry has long-term support for component supporting, and already has ISO 26262 certification conditions and documents

·PVRIC4.1 lossless or visually lossless image compression, which can always ensure that the bandwidth and storage space occupied are reduced by 50%

·Advanced and flexible power control technology

A unified AI API that can be used with the PowerVR neural network accelerator

·Support the latest API standards, including OpenGLES, Vulkan, OpenCL

The Links:   6DI50MA-050 6DI50A-060