AMD – APU, OpenCl and Eyefinity

 September 14, 2011 – San Francisco, CA – We had a chance to meet with AMD recently to overview their new APU which is a combination CPU & GPU single chip with an embedded data and memory manager. They have shipped over 12M of these devices to date.

AMD APP SDK for OpenCL

One of the challenges of modern systems, is that software speeds are not keeping pace with CPU and process migration. To address this challenge, the OpenCL API was developed. OpenCL focuses on handling the issues of distributed core, multi-core CPU and multi-thread GPU systems and the ability to jointly utilize both CPU and GPU for a single program. OpenCL is currently the #1 API for developers and is supported by AMD, Intel, Nvidia, ARM, and Imagination Technologies.

The language is the primary choice of developers for game and digital content creation as it is multi-platform and hardware agnostic. The same code will work in a CPU only system, CPU/GPU system, and embedded controller/GPU system. The OpenGL language takes care of identifying the available resources, and utilizing them. This arrangement allows for creation of code that is compatible with both standalone devices and cloud/data center environments. The language is not just for general purpose applications, but embedded devices such as the HP Printers, use the OpenCL language as their base programming interface.

With the capabilities of the OpenCL language, and the optimized results that are delivered with the high performance, low power integrated solution of the AMD APUs, AMD has taken an active effort to bring an awareness and education program for students, engineering and software developers to learn how to use OpenCL to provide optimized hardware performance. 

AMD has developed course-work, core program outlines, labs and auto-grading systems for software “homework assignments”. The programs are targeted towards Computer Science programs which are software targeted, and machine hardware level transactions are nor necessarily part of the standard curriculum. The course-work is aided by an SDK that is available from AMD [AMD Accelerated Parallel Processing (APP) SDK ] . The fact that the OpenCL community is an open source environment and it is supported by multiple open source development tools in addition to commercial tools. The OpenCL language has also been incorporated into Java under the Java bindings for OpenCL  program, so application development can now receive the ability to transparently scale based on resource availability.

They additionally had an update on their Eyefinity multi-display products . They were showing a new 5×1 horizontal gaming configuration (similar to the 3×1 shown below) running from and 8 core FX CPU with Dual Radeon graphics connected via display port to the monitors.

AMD Eyefinity 3×1 display