Multi-processing graphics technology delivers true scalability to ‘super-high’ performance points
Tokyo, Japan-- November 24, 2009 -- Imagination Technologies, the leading multimedia chip technologies company, will begin shipping fully verified production quality IP for its POWERVR SGX543MP multiprocessor graphics cores to partners before the end of December 2009. Multiple lead partners have already been working with beta versions of the core for several months and several SoCs for a range of markets incorporating SGX MP cores are already in advanced design.
Imagination reports that three of its partners now have licences for SGX MP technology, which continues to extend Imagination’s leadership of the embedded graphics acceleration market with multi-processor, shader-based solutions which address the rapidly growing demands for high performance graphics in a wide range of consumer electronics segments.
The technology can be delivered to customers in SGXMP2 (two-core) to SGXMP16 (16-core) variants.Graphics IP cores available from Imagination now range from SGX520, the world’s smallest OpenGL™ ES 2.0 mobile core, to SGX543MP solutions for high-performance console and computing devices. POWERVR SGX543MP delivers performance comparable to many desktops, laptops and games consoles.
Further details of the innovative technology behind POWERVR SGX543MP are being revealed to press and customers at the Embedded Technology Show in Yokohama, Japan from 18-20
Tony King-Smith, VP marketing Imagination Technologies says: “With the ability to combine up to sixteen SGX543MP GP-GPU cores on a single SoC, we are now able to deliver capabilities to our licensing partners previously only thought the domain of the discrete GPU chipset vendors, while maintaining our unrivalled power, area and bandwidth efficiency.”
The POWERVR SGX543MP family enables POWERVR SGX543 four-pipe programmable GP-GPU cores, to be integrated in a high performance, multi-processor graphics solution without performance or silicon area compromises. SGX543MP enables highly linear scaling of all aspects of GPU performance, specifically vertex shading, pixel shading, primitive setup and overall GP-GPU* functionality, whilst maintaining full software compatibility and with virtually no overhead in bandwidth usage.
At 200MHz core frequency an SGX543MP4 (four cores) will deliver 133 million polygons per second*** and fill rates in excess of 4Gpixels/sec**. Higher frequencies or a larger number of cores each deliver more performance. At 400MHz core frequency an SGX543MP8 (eight cores) will deliver 532 million polygons per second*** and fill rates in excess of 16Gpixels/sec**.
The highly efficient POWERVR SGX543MP family delivers near linear progression in vertex and pixel processing performance, unlike competitive solutions which scale only pixel performance. An SGX543MP2 delivers effectively twice the performance of a single SGX543 without compromise. And for a given workload the same bandwidth is required no matter how many cores are deployed – SGX543MP delivers faster performance by dividing the work on-demand, dynamically load balanced in parallel between cores.
The USSE2 (Universal Scalable Shader Engine v2), a key component of Series5XT architecture at the heart of SGX543MP is a scalable multi-threaded multimedia processing engine offering up to 2x the floating point throughput of earlier Series5 SGX IP cores. An extended instruction set with comprehensive vector operations and co-issue capabilities enables advanced geometry and pixel processing as well as GP-GPU tasks. These tasks are broken down into processing packets which are then scheduled across a number of multi-threaded execution units in the USSE2. This enables optimal hardware load balancing, maximum latency tolerance and efficient gate use, all accessed through a single software programming model and compiler.
POWERVR SGX543MP features:
- the highest performance per mW of any embedded graphics core
- highly linear scaling (over 95% efficiency) of performance in both geometry (vertex processing) and rasterisation (pixel/fragment processing)
- dynamic load balancing and on-demand task allocation at the pipeline level
- no fixed allocation of given pixels to specific cores, enabling maximum processing power to be allocated to the areas of highest on-screen action
- scalable GP-GPU compute power, which can be fully utilised through all Khronos APIs including OpenGL ES 2.0/1.1, OpenVG™ 1.x and OpenCL™
- use any number of cores, even or odd
- no additional work for software developers; using one driver stack for all SGX cores means applications see a common SGX architecture via the standard APIs regardless of number of cores used
- no additional CPU load when using multiple cores or loss of performance
- individual cores can be disabled based on workload for optimal power saving
* GP-GPU stands for General-Purpose computation on Graphics Processing Units.
** All fill rate figures stated assuming a scene depth complexity of x2.5
*** All polygon throughput figures are based on real and achievable sustained throughput in a real SoC; they are not theoretical figures that can never be achieved in any practical application
About Imagination Technologies
Imagination Technologies Group plc (LSE:IMG) – a global leader in multimedia and communication silicon technologies – creates and licenses market-leading processor cores for graphics, video, multi-threaded embedded processing/DSP and multi-standard communications applications. These silicon intellectual property (IP) solutions for systems-on-chip (SoC) are complemented by strong array of software tools and drivers as well as extensive developer and middleware ecosystems. Target markets includemobile phone, handheld multimedia, home consumer entertainment, mobile and low-power computing, and in-car electronics.Its licensees include many of the leading semiconductor and consumer electronics companies. Imagination has corporate headquarters in the United Kingdom, with sales and R&D offices worldwide. See: www.imgtec.com.