What about GDDR3 in the video card?

Compared with the DDR2, the GDDR3 architecture is closer to the DDR2 and is developed by DDR2 as the video card's dedicated memory. In terms of speed, GDDR3 is closer to DDR2 and has a certain gap with the newer DDR3. and GDDR3 and DDR3 are completely different two-generation products. Compared with GDDR, the main innovations of GDDR3 are as follows: The working voltage is reduced from 2.5V to 1.8V, the signal terminals in the chip replace the terminal signal line in GDDR, the output driver with dynamic control impedance, 4-bit prefetching and one-way single port data selection. The combined effects of all these features are higher data rates, better signal integrity, and lower power consumption. As a result of these changes, GDDR3 memory can obtain a much higher data rate than the GDDR and DDR2 standards.

Basic functions and task assignment in three-dimensional graphics rendering system

In interactive three-dimensional games, the computation of graphs can be divided into several basic steps of sequential execution. In the first step, three-dimensional scenes are generated based on the instruction results given by the players. The three-dimensional scene represents the arrangement and position information of all objects in the virtual three-dimensional world. This part of the calculation is done by the CPU in the PC. The CPU then passes the three-dimensional scene to the GPU. The task of the GPU is to convert the three-dimensional scene into a two-dimensional image that the display can display. This task performed by the GPU is called a three-dimensional rendering.

Many effects must be considered when producing the actual image, such as color, texture, multiple extended light sources, shading, reflection, transparency, light absorption, opaque materials, and so on. Achieving all of these results requires strong computing power and a particularly fast and wide memory interface to enable random access to the memory within the shortest latency. The main drivers for memory bandwidth and capacity are the number of parameters that must be stored and quickly available, as well as the storage of intermediate results for highly repeatable computations. Do not forget that all of these calculations must be completed in real time, with a capacity of more than 40 frames per second.

Integrated and stand-alone graphics devices

In general, the graphics system can be divided into two major categories, namely, integrated graphics system and independent graphics system. In an integrated graphics system, a graphics processing unit is embedded within a PC chipset located on a notebook and desktop board. For texture storage and caching, these integration systems use the PC's main memory. This approach limits the rendering performance of three-dimensional graphics in two ways. First, the maximum memory bandwidth is limited by the standard main memory bandwidth, and secondly, the graphics system must share this bandwidth with the CPU and other client programs that access the memory simultaneously on the PC. Benchmark tests performed from "Tom's Hardware Guide" make it clear that the integration system does not provide enough rendering capability for advanced 3D games.

The independent graphics system consists of a graphic processing unit which is physically separate and works independently, and a dedicated graphics memory directly connected to the independent GPU. The standalone graphics system is connected to the PC chipset via the standard PCI-E (formerly AGP) bus. The three-dimensional processing performance of these independent GPU is now far beyond the performance of integrated graphics processors, as is the requirement for memory bandwidth. Benchmark tests show that the performance of independent graphics systems is 3 to 20 times times larger than integrated systems. Memory I/O technology and memory bandwidth play a key role in achieving this performance leap.

The Fever level and high-end graphics cards use specialized x32 structure graphics memory. These systems now use 500~800mhz clock-frequency GDDR3 memory to form a frame cache of up to 512MB. These configurations provide up to 410Gbps of memory bandwidth to the GPU, equivalent to 12 times times the current most advanced PC main memory bandwidth. The newest high-end notebook computer graphics system uses 500MHz GDDR3 memory to compose 256MB frame cache, the memory bandwidth that provides to GPU still can reach 5 times times of the main memory bandwidth of these laptops.

Mainstream graphics systems usually use x16-structured memory components. Most of the new mainstream graphics systems use a clock frequency of about 400MHz DDR2 I/O technology. The performance of these systems is still significantly higher than the integrated graphics system.