Energy-Saving 2.5D Graphics Accelerator Gets Better Performance and Android OS Support
Updated Graphics Accelerator and Composition Engine IP Core available from CAST includes design improvements plus optimized Android support, increasing its ability to offload or replace power-hungry CPUs and GPUs in FPGA or ASIC SoCs
Woodcliff Lake, NJ, May 29, 2014 -- System designers struggling to deliver rich graphical experiences on mobile, wearable, and other power-challenged devices now have a better option in the form of the updated Think2.5D Graphics Accelerator and Composition Engine Core available from semiconductor IP provider CAST, Inc.
The Think2.5D Core is a rendering engine that accelerates two-dimensional graphics functions and pseudo-three-dimensional effects such as reflected and shadowed icons. Sourced from graphics experts Think Silicon, it has already been embraced for its power-saving ability to offload CPUs and/or GPUs in higher-end systems and even replace GPUs in lower-end embedded systems. Improvements in the new version now available include:
- a higher-performance architecture that can process up to four pixels/cycle, and can be scaled to eight pixels/cycle,
- internal compression to reduce demands on the frame buffer,
- improved power management functions, and
- an Android OS driver through which the core acts as a surface composer to accelerate graphics requests from the Skia graphics library, OpenGL®, or any graphics generating application running under the Android system.
Think2.5D is an extremely efficient composition engine that significantly offloads a system’s GPU. The core’s contributions are even more dramatic in GPU-less systems, where rendering rich, attractive Android graphics through the CPU alone is impractical due to processing inefficiency and excessive power consumption.
Compare user interface graphics composition in software versus faster Think2.5D hardware in this demo video clip, and see the difference live in the CAST DAC booth (#1024).
For example, in one such reference design with a 32-bit color SVGA display, a 100K-gate Think2.5D core clocked at just 125MHz delivers a peak of 40 frames per second, while the system’s ARM® Cortex™ A9 CPU clocked at a six times higher rate of 800MHz could only deliver a peak of 10 fps. Here Think2.5D enabled a four-times speedup and associated energy savings with a slower clock rate and relatively little additional silicon.
“Our engineers continue to respond to customer’s real-world challenges by developing innovative graphics processing architectures delivering better performance and energy savings with extremely small silicon footprints,” said George Sidiropoulos, managing director of Think Silicon. “With these Think2.5D extensions, we reached deep into the Android stack to provide vastly more efficient drawing and composition functions, without requiring the designer to know any Android magic and all in a compact and easy-to-use package.”
The company believes the Think2.5D Graphics Accelerator is one of the smallest such cores available, while also providing one of the best-available performance per watt measures in the industry. It uses a new modular, dual-pipeline architecture, with additional user-configurable pipelines available to handle higher operating frequencies. New intelligent power management functions including Dynamic Voltage and Frequency Scaling (DVFS) boost the power savings inherent in the efficient design. The core handles Android graphics functions like image movement, rotation, and scaling (and many others) by working with the Android Hardware Composer and Hardware Abstraction Layer to seamlessly intercept and more efficiently process these operations.
System integration of the Think2.5D Core is straightforward. It employs a standard AMBA AXI system interface, supports a variety of color formats and display resolutions for still images or video, and includes an Android OS driver library so that even design teams not familiar with Android can use the core effectively. The versatile Think2.5D can easily run in small FPGAs, such as Altera’s SoC FPGAs or the Xilinx Zynq line, as well as ASICs.
“This Think2.5D update continues our drive to make it easier and more power-and cost-effective for designers to bring their customers dazzling graphics and a great user experience in even the smallest, least expensive mobile devices,” said Nikos Zervas, chief operating officer for CAST. “The royalty-free Think2.5D opens the doors for some great new products like wearable electronics or smart watches, especially when coupled with one of the hyper-efficient 32-bit BA2X processors we provide.”
The Think2.5D IP Core is available now in RTL source code or FPGA netlists, with the new Android Composition Engine features and Android 4.2.2 drivers and support, as well as its ongoing DirectFB and Qt API support. Visitors to the Design Automation Conference (DAC) in San Francisco June 2–4, 2014 can see a demonstration of Think2.5D Core’s ten-times Android speed-up in action and learn more at CAST’s booth, #1024.
To learn more about CAST and its line of IP cores and subsystems, call +1 201.391.8300, visit www.cast-inc.com, or follow @castcores on Twitter.
|
CAST, Inc. Hot IP
Related News
- CAST Enters Graphics Acceleration IP Market with New Cores that Complement GPUs and System Processors
- Tuxera and Mentor Graphics Partner to Add exFAT and NTFS Support for Nucleus OS; Extend Android and Linux Leadership
- Fraunhofer's xHE-AAC Audio Codec Software Extends Native AAC Support In Android P For Better Quality At Low Bitrates
- Fujisoft supports the development of Android-powered embedded devices with the launch of their Graphics Accelerator for Android supporting Altera SoC
- Synopsys Announces Android Operating System Support for DesignWare ARC Processor Cores
Breaking News
- Jury is out in the Arm vs Qualcomm trial
- Ceva Seeks To Exploit Synergies in Portfolio with Nano NPU
- Synopsys Responds to U.K. Competition and Markets Authority's Phase 1 Announcement Regarding Ansys Acquisition
- Alphawave Semi Scales UCIe™ to 64 Gbps Enabling >20 Tbps/mm Bandwidth Density for Die-to-Die Chiplet Connectivity
- RaiderChip Hardware NPU adds Falcon-3 LLM to its supported AI models
Most Popular
E-mail This Article | Printer-Friendly Page |