|Ⅱ Working principle of VRAM|
|Ⅲ VRAM parameters||1.Memory capacity|
|3.VRAM bit width|
The VRAM (video RAM) is a component used to store graphics information to be processed. The picture we see on the display is composed of individual pixels, and each pixel uses 4-32 or even 64-bit data to control its brightness and color. These data must be saved through the VRAM, and then deployed by the display chip and CPU, and finally, the calculation results are converted into graphics and output to the display. The VRAM, like the motherboard memory, performs the storage function, but the object it stores is the information of each pixel output by the graphics card to the display. After the display chip processes the data, it saves the data to the VRAM, and then the RAMDAC (digital-to-analog converter) reads the data from the VRAM and converts the digital signal into an analog signal. Finally, it is displayed on the screen.
In advanced graphics accelerator cards, the VRAM is not only used to store graphics data but also used by the display chip to perform 3D function operations. In advanced display chips such as nVIDIA, a "GPU" (graphics processing unit) parallel to the CPU has been developed. Due to the role played by the graphics card, obviously, the speed and bandwidth of the graphics memory directly affect the overall speed of the graphics card. As a memory, VRAM has gone through multiple stages of development just like motherboard memory. It can even be said that the development of VRAM is more active than motherboard memory and has more varieties and types. The widely used VRAM types are SDRAM and SGRAM. DDR memory with better performance was first applied to graphics cards.
The graphics card is a BGA-packaged chipset on the motherboard, similar to the CPU (Central Processing Unit), called GPU (Graphics Processing Unit) in the industry. There are mainly two vendors, nVIDIA and AMD, on the market. The graphics chip is equivalent to the CPU of the graphics card, but its main task is to process display information. In the process of processing information, it will generate a large amount of temporary data (unprocessed, processing, processed), which requires a special place to store these temporary data, that is, VRAM. It may also be a chip or just a part of the chip, depending on the hardware design (dedicated graphics and integrated graphics).
Before the graphics card starts to work (graphic rendering modeling), the required material and texture data are transferred to the VRAM. At the beginning of modeling and rendering, these data are transmitted through the AGP bus, and the display chip will extract the data stored in the VRAM through the AGP bus. In addition to the modeling and rendering data, there is a large amount of vertex data and work instruction streams that need to be exchanged. These data are converted into analog signals through RAMDAC and output to the display end, which is the image we see in the end. The performance of the display chip is increasing day by day, and its data processing ability is getting stronger and stronger, making the VRAM data transmission volume and transmission rate also higher and higher, and the requirements for the VRAM of the graphics card are also higher. The size of the VRAM exchange and the speed of the speed are crucial to the performance of the graphics card core, and how to effectively improve the performance of the VRAM has become the key to improving the performance of the entire graphics card.
8gb GDDR5 VRAM
VRAM capacity is the capacity of local VRAM on the graphics card, which is one of the key parameters for choosing a graphics card. The size of the VRAM capacity determines the ability of the VRAM to temporarily store data, and to a certain extent also affects the performance of the graphics card. The VRAM capacity is also gradually increasing with the development of graphics cards. The VRAM capacity has grown from the extremely small capacities of 512KB, 1MB, 2MB in the early days to 64MB, 128MB, 256MB, 512MB, 768MB, and up to the current mainstream 2GB, 4GB, 6GB and high-end graphics cards such as 8GB, 16GB, 32GB. Some professional graphics cards It even has 48GB of VRAM.
The data bits refer to the number of bits that can be transmitted in one clock cycle. It is an important factor in determining the VRAM bandwidth and is closely related to the performance of the graphics card. When the type of VRAM is the same and the operating frequency is the same, the larger the number of data bits, the higher its performance. The calculation method of the VRAM bandwidth is: operating frequency×data bandwidth/8. Taking the current GeForce3 graphics card as an example, its VRAM system bandwidth = 230MHz × 2 (because DDR VRAM is used, it is multiplied by 2) × 128/8=7.36GB.
The VRAM bit width is the number of bits of data that can be transmitted in one clock cycle of the VRAM. The larger the number of bits, the greater the amount of data that can be transmitted in an instant. This is one of the important parameters of VRAM. There are currently three types of VRAM widths on the market: 64-bit, 128-bit, and 256-bit. The 64-bit, 128-bit, and 256-bit graphics cards that people are accustomed to calling refer to their corresponding VRAM widths. The higher the VRAM bit width, the better the performance, the higher the price. Therefore, 256-bit wide VRAM is more used in high-end graphics cards, and mainstream graphics cards basically use 128-bit VRAM. Everyone knows that the VRAM bandwidth = the VRAM frequency X the VRAM bit width/8, so when the VRAM frequency is equivalent, the VRAM bit width will determine the size of the VRAM bandwidth. For example, 128-bit and 256-bit VRAM with the same VRAM frequency of 500MHz, then their VRAM bandwidth will be: 128-bit=500MHz*128∕8=8GB/s, and 256-bit=500MHz*256∕8= 16GB/ s, is 2 times of 128 bits, which shows the importance of VRAM bit width in VRAM data.
The VRAM bandwidth is the bridge between the display chip and the VRAM. The larger the bandwidth, the faster the communication between the display chip and the VRAM. In order to indicate this width, the unit of VRAM bandwidth is bytes/second. The bandwidth of the VRAM is related to the bit width of the VRAM and the speed of the VRAM (that is, the operating frequency). Finally draw a conclusion: VRAM bandwidth = VRAM bit width × VRAM frequency/8. The speed of the VRAM is generally in ns, and the common VRAM is 6ns, 5.5ns, 5ns, 4ns, 3.8ns, up to 1.8ns. The corresponding operating frequencies are 143MHz, 166MHz, 183MHz, 200MHz, 250MHz, up to 550MHz. The calculation method of operating frequency is very simple-the reciprocal of the VRAM speed is the rated operating frequency of the VRAM. For example, if the clock cycle of the VRAM is 4ns, the operating frequency of the VRAM is 1/4ns=250MHz.
The VRAM frequency refers to the frequency at which the VRAM is working on the graphics card by default, in MHz (megahertz). The VRAM frequency reflects the speed of the VRAM to a certain extent. The VRAM frequency varies with the type and performance of the VRAM. SDRAM VRAM generally works at a lower frequency, generally 133MHz and 166MHz. The VRAM frequency is mainly used on low-end graphics cards. The VRAM frequencies that different VRAM can provide are also very different, mainly 400MHz, 500MHz, 600MHz, 650MHz, etc., and there are 800MHz, 1200MHz, 1600MHz, and even higher in high-end products.
VRAM package refers to the type of packaging technology used by the VRAM particles. The package is to wrap the VRAM chip to prevent the chip from contacting the outside world. Impurities and undesirable gases in the air, and even water vapor, will corrode the precision circuits on the chip, thereby causing the electrical performance to decline. Different packaging technologies have great differences in manufacturing processes and processes, and the performance of the memory chip itself after packaging also plays a vital role. VRAM package forms mainly include QFP, TSOP-II, MBGA, etc., among which TSOP-II and MBGA are more common. Many early SDRAM and DDR VRAM used TSOP-II, but now with the improvement of VRAM speed, more and more VRAM uses MBGA package, especially DDR2 and DDR3 VRAM, all use MBGA package. In addition, many manufacturers also refer to the DDR2 and DDR3 memory packages as FBGA, which emphasizes the naming of the pin arrangement, which is actually the same package form. Although MBGA and TSOP-II can achieve higher VRAM frequency, it cannot be simply assumed that MBGA packaged VRAM must be better overclocked, because whether it is easy to overclock or not depends more on the default frequency set by the manufacturer and the actual VRAM. The gap between the frequencies that can be reached includes the design and manufacture of the graphics card. Simply put, the MBGA package can achieve higher frequencies, but the default frequency is also higher.
The VRAM clock cycle is the repetition period of the VRAM clock pulse, which is an important indicator to measure the speed of the VRAM. The faster the VRAM speed, the greater the amount of data exchanged per unit time, and the performance of the graphics card will be significantly improved under the same conditions. The clock cycle of the VRAM is generally in ns (nanoseconds), and the operating frequency is in MHz. The relationship between the VRAM clock cycle and the working frequency: working frequency=1÷clock cycle×1000. Then the memory frequency is 166MHz, then its clock cycle is 1÷166×1000=6ns. For DDR SDRAM or DDR2, DDR3 VRAM, the equivalent output frequency is used when describing its operating frequency. Because data can be transmitted on both the rising and falling edges of the clock cycle, the VRAM bandwidth is twice that of SDRAM when the operating frequency and data bit width are the same. In other words, under the same VRAM clock cycle, the equivalent output frequency of DDR SDRAM VRAM is twice that of SDRAM VRAM. For example, the operating frequency of 5ns SDRAM VRAM is 200MHz, while the equivalent operating frequency of 5ns DDR SDRAM or DDR2, DDR3 VRAM is 400MHz. Common VRAM clock cycles are 5ns, 4ns, 3.8ns, 3.6ns, 3.3ns, 2.8ns, 2.0ns, 1.6ns, 1.1ns, or even lower.