New Computing Boon Beckons

Digital technology boosts many facets of research and development, with the computer serving as the single biggest invention drawing us into the future. Nowhere is that more apparent than with recent innovations in hardware architecture and growing use of multicore computer processing units.

The generalization of graphics processing units, and the imminent leap available between CPU and GPU cores, make a new era of interactive graphics possible. As a result, we will begin to see PCs and gaming consoles with unprecedented levels of color saturation, realism, and “immersiveness.”
But current graphics programming models known as Application Programming Interfaces were developed for the previous-generation GPU-only rendering pipelines. This severely hampers the type and quality of images possible on these systems.

Interactive Graphics Driving Innovation

Programmable graphics is the next step in interactive graphics development. This new era of imagery will cooperatively use the CPU, GPU, and complex data structures to efficiently synthesize future images according to next-generation software apps.

Next-generation interactive graphics programming models for these architectures are being developed and require new programming models, tools, and rendering systems designed to take full advantage of these new parallel heterogeneous systems. New graphics techniques, algorithms, and rendering engines that showcase unprecedented visual quality are making these images possible.

The single-processor computer is being replaced by parallel heterogeneous systems made up of processors supporting multiple styles of computation. CPU architects are no longer able to improve computational performance of the traditional heart of the computer system.

Instead of increasing clock speed, engineers are now providing a rapidly-increasing number of parallel coarse-grained cores, currently capable of delivering approximately 90 GFLOPS. (GFLOPS is 1 billion floating point operations per second).

Simultaneously, graphics processing units have evolved to be efficient fine-grained data-parallel co-processors that deliver much greater raw floating-point horsepower than today’s multi-core CPUs. New processors from AMD can produce 400 GFLOPS of peak performance via hundreds of computational units working in parallel.

In addition, although CPUs and GPUs have traditionally been separated by low-bandwidth and high-latency communication pathways, rapidly-improving interconnect technology (for example, AMD Torrenza and Intel Geneseo) and the promise of integrating CPUs and GPUs on a single chip (for example, AMD Fusion) allow CPUs and GPUs to share data much more efficiently. This enables graphics applications to intermix computational styles to optimally use the system’s computational resources.

Interactive 3D computer graphics is now the most computationally demanding consumer application. The economic force of the computer gaming industry and its appetite for computational power have driven the rapid development of current GPUs. In addition, the GPU programming model represents perhaps the only widely-adopted parallel programming model to date.

Unfortunately, this model assumes a GPU-only, unidirectional fixed-graphics pipeline. Creating a new programming model for interactive graphics which fully exposes the computational and communication abilities of these new architectures is necessary to enable a revolution in the quality and efficiency of interactive graphics and provide a killer app for these platforms.

The last five years have seen a significant amount of innovation in interactive graphics software and hardware; GPUs have progressed from being configurable fixed-function processors to highly-programmable data-parallel co-processors, while CPUs have evolved from single-core to task-parallel multicore processors.

The promise of programmable graphics illustrates the fact that GPU programmability has implications for computer graphics. User-defined data structures and algorithms are bringing tremendous flexibility, efficiency, and image quality improvements to interactive rendering.

What Our Future Could Bring

We are at the threshold of a new era of interactive computer graphics. We are no longer limited to today’s brute force, unidirectional rendering pipeline. Developers will soon be able to design adaptive, demand-driven renderers that efficiently and easily leverage all processors in new heterogeneous parallel systems.

New rendering algorithms that tightly couple the distinct capabilities of the CPU and the GPU will generate far richer and more realistic imagery, use processor resources more efficiently, and scale to hundreds both CPU and GPU cores.

We believe this new technology will be a boon for processor and chip manufacturers, including Intel and AMD. As applications such as 3D become more sophisticated, consumers will embrace these new technologies with vigor.

Apple, Google, Microsoft and other companies will seize new opportunities as the hardware and software technologies continue to evolve.

This post is from TechMan, WMB co-author who blogs about trends, issues and ideas affecting business, industry, technology and consumers.