Introduction
A raster display system is inseparable from the graphics processor. The graphics processor is an important component of the graphics system structure and connects the computer and the display terminal. The bond.
It should be said that there is a graphics processor (commonly known as a graphics card) if there is a display system, but the early graphics cards only contained simple memory and frame buffers, which actually only played a role in the storage and transfer of a graphic. All operations must be controlled by the CPU. This is sufficient for text and some simple graphics, but when dealing with complex scenes, especially some realistic three-dimensional scenes, this system alone cannot accomplish the task. Therefore, the later developed graphics cards have graphics processing functions. It not only stores graphics, but also performs most graphics functions, which greatly reduces the burden on the CPU and improves the display capacity and display speed. With the development of electronic technology, the technical content of graphics cards is getting higher and higher, and the functions are getting stronger and stronger. Many professional graphics cards have strong 3D processing capabilities, and these 3D graphics cards are gradually moving towards personal computers. Some professional graphics cards have even more transistors than contemporary CPUs. For example, in 2000, the RADEON graphics chip introduced by ATI in Canada contained 30 million transistors, reaching a filling rate of 1.5 billion pixels per second.
Composition
The graphics processor is composed of the following devices:
(1) Display the core of the main chip graphics card, commonly known as GPU, its main task is to control the system The input video information is constructed and rendered.
(2) The display buffer memory is used to store the graphics information to be displayed and the intermediate data of the graphics operations; the size and speed of the display buffer directly affect the performance of the main chip.
(3) The RAMD/A converter converts the binary number into an analog signal suitable for the display.
Development trends and problems
Problems in computing power and computing mode
The basis of the current GPU-the traditional Z-buffer algorithm can not meet the new Application requirements. In real-time graphics and video applications, more powerful general-purpose computing capabilities are required, such as support for collision detection and approximate physical simulation; in games, graphics processing algorithms must be combined with non-graphic algorithms such as artificial intelligence and scene management. The current GPU architecture cannot well solve the problems of transparency, high-quality anti-aliasing, motion blur, depth of field, and micro-polygon dyeing that need to be solved for film-level image quality. It cannot well support real-time ray tracing, Reyes (Renders everything You ever saw) and other more complex graphics algorithms are also difficult to deal with global illumination, dynamic and real-time display, shadows and reflections required for high-quality real-time 3D graphics. Need to study a new generation of GPU architecture to break these limitations. With the rapid development of VLSI technology, a new generation of GPU chips should have more powerful computing power, which can greatly improve graphics resolution, scene details (more triangles and texture details), and global approximation. The development trend of graphics processing systems is the fusion of graphics and non-graphic algorithms and the fusion of existing different coloring algorithms. The new generation of graphics system chips requires unified and flexible data structures, new programming models, and multiple parallel computing modes. We believe that the development trend is to solve the problems faced by current graphics processing system chips with a unified computing model of data-level parallelism, operation-level parallelism, and task-level parallelism based on a unified, regular parallel processing element array structure.
Problems in manufacturing process
Integrated circuits have developed to nanoscale technology and are constantly approaching physical limits. The so-called red wall problem has appeared: First, the delay of the line is more and more than the delay of the gate. The more important. Long-term transmission has not only the problem of transmission delay, but also the problem of energy consumption. Second, the feature size is so small that chip manufacturing defects are inevitable, and fault tolerance and fault avoidance technologies should be studied from three aspects: defect tolerance, fault tolerance, and error tolerance. Third, leakage current and power consumption have become very important, and independent power management technology should be adopted. Modern graphics processor chips have made significant progress in overcoming the red wall problem: using a large number of regular SIMD array structures; its distributed memory is close to the computing unit, reducing long-term effects; its hardware multithreading Covers part of the impact of storage delays. However, with the further development of the process, the current GPU architecture is difficult to adapt to future process development, and there is no system structure to deal with long-term problems, process deviations and process defects, especially how to adapt to the three-dimensional process. The gate thickness of the current state-of-the-art transistor is already about five atoms. In manufacturing, one atom missing will cause a 20% process deviation. Therefore, the deviation of the process has become an issue that cannot be ignored in SoC design. In particular, nanoelectronic integrated circuits after 2018 can produce regular nanodevices through random self-assembly. Therefore, the system structure of the new generation system chip must use a regular structure and tolerate process deviations, with fault tolerance, fault avoidance and reorganization capabilities. We believe that the use of adjacent technology between a large number of homogeneous processor elements to adapt to nano-scale technology and future three-dimensional technology, adopting new architecture and related low-power, fault-tolerant and fault-avoidance design strategies, is for future graphics processing systems The chip has important scientific significance.
Major suppliers
ATI
On August 20, 1985, ATI was established. In October of the same year, ATI developed the first graphics chip using ASIC technology. And graphics cards. In April 1992, ATI released the Mach32 graphics card with integrated graphics acceleration. In April 1998, ATI was selected as the market leader in the graphics chip industry by IDC, but at that time this chip did not have the title of GPU. For a long time, ATI referred to the graphics processor as the VPU. It was not until AMD acquired ATI that its graphics chip officially adopted the GPU name.
NVIDIA
NVIDIA first proposed the concept of GPU when it released the Geforce256 graphics processing chip in 1999. Since then, the core of NVIDIA graphics cards has been called GPU. GPU makes the graphics card reduce the dependence on the CPU, and perform part of the original CPU work, especially in 3D graphics processing. The core technologies used by the GPU include hardware T&L, cubic environment texture mapping and vertex blending, texture compression and bump mapping, dual texture four-pixel 256-bit rendering engine, etc., while the hardware T&L technology can be said to be GPU’s Sign.
Model examples
AMD
AMD notebook computer graphics products are mainly Mobility Radeon series, which have certain 3D performance, and their products mainly include R9 ( High-end), R7 (mid-end), R5 (low-end) three series:
Mobility Radeon R9 M200 series of Mobility Radeon R9 M295X/M280/M275X//M265X; /p>
Mobility Radeon R9 M395X/M385X/M375X//M365X of Mobility Radeon R9 M300 series;
Mobility Radeon R7 M200 series Mobility Radeon R7 M270/M265f M260X/M260;
Mobility Radeon R7 M300 series Mobility Radeon R7 M380/M370/M365/M360X/M340;
li>Mobility Radeon R5 M255/M230 for Mobility Radeon R5 M200 series;
Mobility Radeon R5 M300 series Mobility Radeon R5 M335/M330/M320/ M315 and so on.
nVIDIA
nVIDIA notebook computer graphics products mainly include GeForce 900M series mobile graphics cards, GeForce 800M series mobile graphics cards, and GeForce 700M series mobile graphics cards.
The GeForce9800M series mobile graphics cards mainly include GeForce GTX980M/ GTX970M/ GTX 960M/GTX950M/940M/930M/920M/910M, etc.
GeForce 800M series mobile graphics cards mainly include GeForce GTX880M/ GTX870M/ GTX860M/GTX850M/ 840M/830M/820M, etc.
GeForce 700M series mobile graphics cards mainly include GeForce GTX780M/ GTX770M/ GTX765M/GTX760M/GT755M/750M/GT745M/GT740M/GT730M/GT720M/ etc.