Imagination Unveils ‘GPU of Everything’ Family

Article By : Nitin Dahad

Imagination Technologies launches its most significant new GPU family for 15 years, the IMG A-series.

Making its most significant new GPU family announcement for 15 years, Imagination Technologies this week unveiled its IMG A-Series, the tenth generation of its PowerVR architecture. Calling it the ‘GPU of everything’, the new cores introduce a new GPU architecture with single 128-thread wide ALU to boost performance and power efficiency.

The company said the A-Series is 2.5x faster (or 150%) for the same area and same power compared to its currently shipping PowerVR GPUs, making it inherently more power efficient. For the same performance it is now 60% lower power.

The new GPU family enables performance scalability ranging from an entry-level 1 pixel per clock (PPC) part right up to 2 TFLOP cores for performance devices, and beyond that to multi-core solutions for cloud applications. This scalability is why Imagination is calling the IMG A-Series the ‘GPU of everything’, because of its ability to address multiple markets, from automotive, AIoT (many firms are now using AIoT to denote intelligence embedded into Internet of things (IoT) devices), and computing through to DTV/STB/OTT, mobile and server.

The main source of speed improvement is from changes to the ALU architecture. In the A-Series Imagination has moved from a dual MAD x32-wide ALU to a single MAD 128-thread wide ALU. By moving to a 128-thread wide unit it said it is now able to perform many more instructions per clock, which is fundamental to improving power efficiency. The A-Series is also scalable from one to four scalable processing units (SPUs) to create cores that match market requirements while still keeping area size to a minimum.

A-Series 128-wide ALU

The main source of speed improvement in the A-Series comes from the 128-thread wide ALU architecture (Source: Imagination Technologies)

Imagination said that compared to other GPU IP solutions available today IMG A-Series delivers higher performance, lower power (compared to competitors at the same clock and process), and lower bandwidth (at the same cache size as competitors), all in smaller silicon size. Its architecture enables differentiators such as guaranteed 50% image compression data (lossless in most cases, or visually lossless in exceptional cases).

Jon Peddie, founder of Jon Peddie Research, said, “The simple fact is that for mobile SoCs (systems on chip) the market leader owns its own GPU technology and is increasing market share at rate of 5% year-on-year. In order to stop the losses to their own potential share the other mobile SoC companies need a compelling GPU that will deliver some real competition. Imagination’s A-Series can do that.”

The new IMG A-Series is already licensed for multiple markets and the first SoC devices are expected in 2020, according to Imagination. It is available in various configurations:

  • IMG AXT-64-2048 for flagship performance; 2.0 TFLOPS, 64 Gpixels and 8 TOPS of AI performance
  • IMG AXT-48-1536 for premium mobile; 1.5 TFLOPS, 48 Gpixels and 6 TOPS
  • IMG AXT-32-1024 for high-performance mobile and automotive; 1 TFLOPs, 32 Gpixels and 4 TOPS
  • IMG AXT-16-512 for high-mid-performance mobile and automotive; 0.5 TFLOPS, 16 Gpixels and 2 TOPS
  • IMG AXM-8-256 for mid-range mobile; 0.25 TFLOPS, 8 Gpixels and 1 TOPS

For lower cost segments, the IMG A-Series offers the following area, cost and efficiency optimized variants:

  • IMG AXE-2-16 for premium IoT, entry DTV/STB, display and other fillrate driven applications; 2 PPC, 16 GFLOPS and 2 Gpixels
  • IMG AXE-1-16 for entry-level mobile and IoT and the fastest Vulkan-capable GPU in its class; 1 PPC, 16 GFLOPS and 1 Gpixels

HyperLanes for GPU multitasking
One of technologies Imagination introduced with the new family is what it calls ‘HyperLane technology’: individual hardware control lanes, each isolated in memory, enabling different tasks to be submitted to the GPU simultaneously for secure GPU multitasking. With dynamic performance control, the GPU can spread its performance across these multiple tasks, executing them all, while maximizing GPU utilization. Priority HyperLanes deliver a dynamic split; for example, prioritizing graphics and delivering all the required performance for that application while an AI task runs alongside using the remaining performance. HyperLane technology can also isolate protected content for rights management. The IMG A-Series GPUs support up to eight HyperLanes.

HyperLane technology

The new HyperLane Technology in the IMG A-Series enables different tasks to be submitted to the GPU simultaneously for secure GPU multitasking (Source: Imagination Technologies)

HyperLane also enables a new option to enable SoC designers to utilize the compute capability in the IMG A-Series to accelerate their AI workloads. Called AI Synergy, it is an option that allows the GPU to deliver graphics performance while using its spare resource to enable programmable AI alongside a fixed-function, highly optimized Imagination neural network accelerator. This delivers programmable AI in the lowest silicon area, while a unified software stack enables flexibility and performance.

Another new feature of the new GPU family is that cache sizes can now be configured to their preference, allowing developers to significantly reduce GPU bandwidth requirements. More system-level cache can now be put down to reduce the need to go to system memory or to allow the use of less expensive external RAM to reduce power, thermals and system cost.

Not forgetting its gaming heritage
According to Imagination, most mobile GPUs struggle to deliver consistent gaming experiences due to thermal constraints – they are fast for a short while, then slow to unacceptable levels and then never recover, resulting in a disappointing user experience. The company claims the IMG A-Series delivers sustained performance for extended game play at consistent frame rates avoiding thermal panic clock throttling, glitches or overheating.

The IMG A-Series’ PowerVR architecture has an inherent bandwidth and power advantage due to its tile-based deferred rendering technology, which means it only draws what is visible on screen. This is combined with advanced direct flexible power signalling using Pro-Active DVFS (dynamic voltage and frequency scaling) and Deadline Scheduling algorithms, which deliver fast power control with ultra-low latency response. If parts of the GPU aren’t fully utilized or needed for processing, they are immediately slowed down or even put to sleep to ensure optimal power efficiency.

For optimum gaming user experience, the IMG A-Series delivers sustained stable framerates. This requires high GPU power efficiency, which is provided by PowerVR’s smart flexible firmware based low-latency DVFS algorithms with a direct GPIO control path into the system. This is combined with a focus on delivering visually impeccable pixels via features such as the industry-leading visually lossless PVRIC compression and completely reworked anisotropic sampling and level of detail algorithms.

An executive at Chukong Technologies, said that gaming requires considerable graphics processing capability, and with gaming accounting for over 40% of all smartphone use, having the best GPU is a key differentiator. This is where it said Imagination is pushing the boundaries of what is possible with mobile graphics, combining fast and consistent GPU performance with sustained battery life. 

Virtual Event - PowerUP Asia 2024 is coming (May 21-23, 2024)

Power Semiconductor Innovations Toward Green Goals, Decarbonization and Sustainability

Day 1: GaN and SiC Semiconductors

Day 2: Power Semiconductors in Low- and High-Power Applications

Day 3: Power Semiconductor Packaging Technologies and Renewable Energy

Register to watch 30+ conference speeches and visit booths, download technical whitepapers.

Leave a comment