AMD Fusion architecture boosts ETX and XTX performance

AMD Fusion processors liberate ETX and XTX COM standards from looming form factor, power consumption, and performance constraints.

4ETX – one of the oldest and most widespread embedded COM standards – was threatened when Pentium M and Celeron M processors were discontinued, as was ETX’s sister standard XTX. This discontinuation left a gaping hole in the upper performance range, the demand of which continues to rise. AMD addresses this problem with its new, highly scalable embedded processor generation with Fusion architecture that enables the highest graphics and application processor performance with low power consumption for fanless designs.

COMs have established themselves on a broad basis over the last decade. They allow device developers to concentrate on their core competence. Developers can purchase know-how intensive and therefore costly solutions in the field of processor and RAM storage technology as pre-integrated solutions at a low price. In the long run, they can then change them quite simply and without much integration effort on the module level. Thanks to these options, systems become fully scalable and the long-term availability on the module level increases as a result of the independence from the currently implemented processor.

Since its introduction around eight years ago, ETX has established itself as one of the most widely used standards across all performance and application classes. Those that do not require ISA bus support can easily be switched to the identical XTX form factor standard. ETX and XTX enjoy immense popularity, also for current designs, due to the fact that SATA has been supported since the release of version 3.03 of the ETX standard.

Following last year’s discontinuation of the Pentium M and Celeron M processors and the 855/852 chipsets, doubts about the future of ETX and XTX for the upper performance range have arisen in many respects. This is aggravated by the fact that modern applications and the trend toward touch screens, sophisticated user interfaces, and increasingly larger displays with higher resolutions have placed much greater demands on graphics than previous processor generations could offer. In addition, maximum power consumption for fanless and battery-operated devices needs to be as low as possible and must not exceed the upper limit of 20 W. The classic approach with current processors lacks graphics performance when using low power consumption, especially in the 3D realm, or thermal power dissipation and space requirements increase disproportionately. AMD has addressed this problem with a fundamentally new approach: the G-series Fusion architecture.

Improved processing with Fusion

The essential features of the new AMD G-series processors are an ideal fit for the requirements of the small form factor embedded market, especially for users of existing ETX or XTX modules, which are mainly employed in industrial and building automation, kiosk applications, medical technology, gaming, and (interactive) digital signage. What all these applications have in common is the trend toward ever more realistic (3D) visualization with high resolutions up to HD. Even with the higher graphics performance required by high-resolution touch screen controls, the Fusion architecture lets designers easily upgrade their machines and devices by changing modules in order to enable modern and graphics-intensive user interfaces. This leads to completely new possibilities in areas such as medical imaging technology and analysis devices where Fusion architecture can fully exploit the advantages of parallel processing via the APU for the relevant evaluations.

Specifically, AMD’s Fusion architecture combines the Central Processing Unit (CPU) and Graphics Processing Unit (GPU) on one chip to an Accelerated Processing Unit (APU). The graphics processor is a General Purpose Graphics Processing Unit (GPGPU), which distinguishes itself from a standard GPU by its flexible parallel processing units. Thereby, the GPGPU can also be used for general, compute-intensive, parallelizable operations, and can also considerably increase the performance in the non-graphics sphere. Typical uses include numerical mathematical applications and all types of encoding/decoding tasks, encryption and network packet processing in particular.

Increasing efficiency with parallel and serial tasks

The Fusion architecture, highly suitable for XTX and ETX applications, takes advantage of the fact that processes run mostly serially on a standard CPU. Under these circumstances, parallelization can only actually occur in multiprocessor systems or virtually via time-splicing control of the individual, relatively large processes. The situation is different with a GPU: Here, tasks are distributed over many, very small and highly-specialized engines, which are linked with one another according to their respective tasks and which manage the various tasks in each timestep in parallel. What is special about a GPGPU is the fact that the individual processor tasks are not hard-wired as, for example, is the case with the vertex shader unit using a simple GPU. Instead, the particular tasks are freely configurable, similar to a network processor within a certain range. For the software developer this process remains transparent because the engines of the AMD G-series GPGPU are based on proven standards such as DirectCompute from Microsoft and OpenCL (not to be confused with OpenGL), and because AMD offers software development kits.

Since real task formulations are seldom purely serial, there are, in practice, enormous efficiency advantages when using a fused APU, which can process both serial and parallel tasks. Traditional CPU architecture and programming tools are usually less suitable for vectorial data models with parallel multi-threads. The traditional CPU performance benchmarks, which were used to compare the performance of platforms, are no longer meaningful. What will be decisive in the future, however, is the APU performance, which can differ in otherwise identical embedded systems according to the application, which can have more or less parallel tasks or threads. The standard 2D and 3D graphics benchmarks offer a good indication of the potential of the APU as they describe the maximum performance of the APU in a single graphics mode. Figures 1 and 2 illustrate the performance of preproduction Fusion APUs in comparison to other processors for a special application (overall performance) and for a 3D benchmark. Note that the maximum power consumption of the systems can differ quite substantially.

21
Figure 1: Graphics performance comparison: Internal tests demonstrate that the graphics performance of the AMD embedded G-series is higher than that of the Intel Core i5 520M.
(Click graphic to zoom by 1.9x)

22
Figure 2: Overall performance comparison: At normal clock rates the overall performance of the AMD G-T56N 1.6 GHz dual core APU is on par with Intel’s Core i5 520M.
(Click graphic to zoom by 1.9x)

21
Sidebar 1: The AMD G-series offers scalability for today’s graphics-intensive XTX and ETX applications.
(Click graphic to zoom)

Why choose an ETX or XTX module?

ETX and XTX (for applications without an ISA bus) are already implemented as proven standards in millions of baseboards. However, for new projects and redesigns it is also often wise to implement proven standards. This is confirmed by the wide range of ETX- and XTX-compatible baseboards and periphery offered (cooling and housing, for example). Designers can stick to the specific interfaces and I/Os around standardized COM, using familiar technology and their core competencies. In this way, they can quickly and easily implement a new processor architecture like Fusion by means of a simple module exchange without much effort in terms of hardware.

For this purpose, congatec offers the ETX COM conga-EAF and the XTX COM conga-XAF. Both modules are based on the AMD G-series Fusion processors and the embedded controller hub Hudson E1, and are available with single core processors with 1.2 GHz or 1.5 GHz, or with dual core processors with 1.0, 1.4 or 1.6 Hz and a thermal power dissipation of between 9 W and 18 W. They come equipped with a single DDR3 SODIMM module with up to 4 GB/1066 Hz but can also support the standard DDR3 with 1.5 V as well as DDR3L with 1.35 V. For more information on congatec System-On-Modules visit: http://congatec.us/products.html?&L=6.

A guaranteed future for ETX and XTX applications

The G-series Fusion architecture provides ETX and XTX small form factor applications with an excellent ratio of computing power to power consumption and enables battery-operated and/or fanless devices with high graphics requirements, often for the first time. What’s more, ETX and XTX modules with Fusion technology will remain future-proof for a long time to come, thereby facilitating easy entry into new dimensions of graphics performance and user-friendly interfaces.

Ron Mazza is General Manager at congatec. He has 33 years of management experience in technology companies, including five successful startups. Before helping found congatec, he contributed to the success of companies such as Zilog, Faraday Electronics, Phoenix Technologies, RFM, and VIA Technologies, mostly at the VP level or above. Before his transition to executive management, Mazza was a software engineer.

congatec Inc. 858-457-2600 ron.mazza@congatec.com www.congatec.us