About the graphics card's video memory

video memory, also called frame cache, is used to store video cards Chipprocessed or about to be extracted. renders the data. Like computer memory, video memory is used to store The part of the graphic information to process.

Role

Video MemoryLike computer memory, video memory is the part used to store the graphics information to be processed. The picture we see on the display is made up of pixels, and each pixel points to 4 to 32 or even 64 bits of data to control its brightness and color, which must be saved through video memory, then by the display chip and CPU provisioning, and finally the results of the calculation into a graphics output to the display. Memory, like the motherboard, performs the function of storage, but it stores information such as the video card output to each pixel on the monitor. Video memory is a very important part of the graphics card, the display chip after processing data will be saved to the video memory, and then by the RAMDAC (digital-to-analog converter) from the memory read out the data and convert the digital signal to analog signal, and finally by the screen display. In the Advanced Graphics accelerator card, video memory is used not only to store graphics data, but also to be displayed on the chip for 3D function operations. In advanced display chips such as nvidia, "GPU" (graphics processing units) parallel to the CPU has been developed. High-density operations such as "T&l" (deformation and illumination) are performed by the GPU on the graphics card, thereby increasing the reliance on the memory. Obviously, the speed and bandwidth of video memory directly affect the overall speed of the graphics card because of the effect of the video card. Video memory as memory has undergone several stages of development, and even the development of video memory is more active than the motherboard, and has more varieties and types. The widely used memory type is SDRAM and Sgram, and the more superior DDR memory is first applied to the video card, which improves the overall performance of the graphics card. DDR to the success of the card on the guide, the full development of the motherboard system, a DDR "three years Alone" era is about to appear in front of the world.2 mainly produces The graphics card is a BGA package on the motherboard chipset, similar to the CPU (central processing unit), the industry is called the GPU (graphics processing unit), the market is mainly Nvidia and AMD two manufacturers. The graphics chip is equivalent to the CPU of the computer, but its main task is to process the display information, in the process of processing information, it will generate a lot of temporary data (unhandled, is processing, has been processed), which requires a special place to hold these temporary data, that is, video memory, it may also be a chip, It may also be part of the chip, which depends on the design of the hardware (discrete graphics and integrated graphics). As for the memory size. When the boot time is generally displayed, you can also click on the desktop properties--settings--advanced------Check the "Memory Size" (XP system), or right--screen resolution-Advanced settings--Adapter to view the adapter information (Win7 system) on the desktop. PCB: Is printed circuit board (printed circuit BOARD,PCB). It will almost appear in every electronic device. If you have electronic parts in a particular device, they are all mounted on PCBs of various sizes. In addition to fixing a variety of small parts, the main function of the PCB is to provide the above each part of the electrical connection. As electronic devices become more complex, more and more parts are needed, and the lines and parts of the PCB are becoming more and more dense. Bare boards (no parts on top) are often referred to as "printed circuit boards printed Wiring Board (PWB)". The substrate of the board itself is made of insulating and insulating materials, which are not easily bent. On the surface can be seen on the small line material is copper foil, the original copper foil is covered in the entire board, and in the middle part of the manufacturing process is etched off, leaving the part will become a mesh of small lines. These lines are called wire (conductor pattern) or wiring, and are used to provide circuit connections to parts on the PCB. Usually the color of the PCB is either green or brown, which is the color of the solder mask paint (solder masks). It is an insulating protective layer that protects the copper wire and prevents the part from being soldered to the wrong place. A layer of silk screen printing is also printed on the solder mask layer. It is usually printed with words and symbols (mostly white) to indicate the position of each part on the board. The screen printing surface is also known as the Icon Face (legend).3Working principleThe graphics card works by transmitting the required material and texture data into the video memory before the graphics card starts working (graphical rendering modeling). At the beginning of the work (modeling rendering), the data is transmitted through the AGP bus, and the chip will extract the data stored in the video memory via the AGP bus, and in addition to the modeling rendering data, there is a large number of vertex data and work instruction flows that need to be exchanged. The data is converted from RAMDAC to the analog signal output to the display end, which is the image we see. Display chip performance is increasing, its data processing ability is more and more strong, so that the memory data transmission and transmission rate is also more and more high, graphics card to the memory requirements are higher. Carrier, the size of the exchange rate of memory, the speed of the graphics card core performance is very important, and how to effectively improve the efficiency of video memory is the key to improve the performance of the entire display card.4difference KindAs an important part of the display card, video memory has been gradually changed with the development of the display chip. From the early Edoram, Mdram, SDRAM, SGRAM, VRAM, Wram until today's widely used DDR SDRAM memory has undergone many generations of progress. The type of memory used in the market mainly has sdram,ddr sdram,ddr sgram three kinds. SDRAM particles are mainly used on low-end graphics cards, the frequency is generally not more than 200MHz, in price and performance it has no advantage over DDR, so gradually replaced by DDR. DDR SDRAM is the mainstream in the market (including DDR2 and DDR3), on the one hand is the maturity of the process, the production of batches resulting in lower costs, making it cheaper, on the other hand, it can provide a high frequency of operation, resulting in excellent data processing performance. As for the DDR SGRAM, it is a product that the graphics card manufacturer has adapted specifically to the needs of the plotter, to enhance the graphics access processing and the efficiency of drawing control, from synchronous dynamic random access memory (SDRAM). Sgram allows to individually modify or access in-memory data in blocks (Blocks), which can work synchronously with the central processing unit (CPU), reduce the number of memory reads, increase the efficiency of the drawing controller, although its stability is good, and performance is good, but its overclocking performance is very poor, It is also very rarely used.FPM DRAM FPM DRAM(Fast page mode RAM): Fast paging pattern memory. is a commonly used memory in the 486 period (also used for video memory). 72 line, 5V voltage, bandwidth 32bit, the basic speed of 60ns or more. Its reading cycle begins with the triggering of a row in the DRAM array and then moves to the location indicated by the memory address, which contains the required data. The first piece of information must be validated and saved to the system in order to prepare for the next cycle. This introduces a "wait state", because the CPU has to wait for the memory to complete a cycle foolishly. One of the important reasons that FPM is widely used is that it is a standard and safe product and is cheap. But its performance flaws have led to its soon being replaced by Edo DRAM, the graphics card of this kind of memory no longer exists.EDO EDO(Extended Data out) DRAM, which is 5% faster than fpm Edo DRAM, because a logic circuit is set up inside the Edo that Edo can read the next data into memory before the last memory data read is finished. The Edo DRAM, designed for system memory, was originally very expensive, just because the PC market needed a replacement for FPM DRAM products, so it was widely used on fifth generation PCs. The Edo memory can work at 75MHz or higher, but its standard operating frequency is zero MHz, but its speed is not enough to meet the needs of the display chip, but also early become "antique-grade" products have the memory.SGRAM SGRAMis the synchronous graphics dram abbreviation, meaning that the synchronous graphics RAM is a graphics card designed for video memory, is a graphic reading and writing ability of the memory, from the SDRAM improved. It improves the past low performance memory transfer rate of the shortcomings of the performance of the display card to create conditions. Sgram reading and writing data is not one by one read, but in "block" units, thereby reducing the overall memory read and write times, improve the efficiency of the graphics controller. But its design and manufacturing cost is higher, more is used in the then more high-end graphics card. At present, this kind of video memory has been basically not used by manufacturers, the DDR memory is replaced. SDRAM, the synchronous DRAM (synchronous dynamic random memory), was once the most widely used type of memory on PC computers, even today SDRAM still occupies a niche in the market. Since it is "synchronous dynamic random memory", it means that its operating speed is synchronized with the system bus speed. SDRAM memory is divided into PC66, PC100, PC133 and other different specifications, and the number behind the specification represents the maximum function of the memory of the system bus speed, such as PC100, which means that the memory can be in the system bus 100MHz computer synchronization work. Synchronization with the system bus speed, which is synchronized with the system clock, avoids unnecessary wait cycles and reduces data storage time. Synchronization also enables the storage controller to know which clock pulse period is used by the data request, so that the data can begin to transmit at the time of the pulse rise. SDRAM uses a 3.3 volt operating voltage, a 168Pin DIMM interface with a bandwidth of 64 bits. SDRAM is not only used in memory, but also is more common in video memory. SDRAM can work synchronously with CPU without waiting period and reduce data transmission delay. Advantages: Low price, has been widely used in low-end graphics card. SDRAM after the DDR SDRAM became mainstream, the scenery is no longer, currently only in the lowest end of the product or thrift market to see this kind of video memory products. DDRVideo memory is divided into two, one is accustomed to the DDR memory, strictly speaking, DDR should be called DDR SDRAM. The other is DDR SGRAM, which is rarely used in this type of memory.DDR SDRAM DDR SDRAMIt is customary to call DDR SDRAM a DDR. DDR SDRAM is the abbreviation for double Data rate SDRAM and is the meaning of double-speed synchronous dynamic random memory. DDR SDRAM is developed on the basis of SDRAM, still use SDRAM production system, so for the memory manufacturers, only the production of ordinary SDRAM device slightly improved, you can achieve DDR memory, effectively reduce costs. RAM transmits data only once in a clock cycle, it transmits data during the rise of the clock, and DDR memory is transmitted two times in a clock cycle, it is capable of transmitting once in the rising and falling periods of the clock, so it is called double-rate synchronous dynamic random memory. DDR memory can achieve a higher data transfer rate at the same bus frequency as SDRAM. Compared to SDRAM: DDR uses a more advanced synchronization circuit, so that the specified address, data transmission and output of the main steps are independent execution, and maintain full synchronization with the CPU; DDR uses a DLL (delay Locked Loop, a delay-lock loop to provide a data filtering signal) technology, When the data is valid, the storage controller can use this data filter signal to accurately locate the data, output every 16 times, and resynchronize data from different memory modules. The DDL essentially does not need to increase the clock frequency to double the speed of the SDRAM, which allows the data to be read out on the rising and falling edges of the clock pulses, thus making it twice times faster than the standard Sdra. DDR SDRAM is the most widely used memory type, with more than 90% of graphics cards using this kind of video memory.DDR SGRAM DDR SGRAMis developed from the Sgram, but also in a clock cycle two times the data, it can be in the clock rise and fall period of each transmission of data. The data transfer rate can be increased by one-fold without increasing the frequency. The DDR Sgram is better than DDR SDRAM in performance, but it is still more cost-dependent than DDR SDRAM and is only used on fewer products. and its overclocking ability is weak, because of its structural problems overclocking easily damaged. DDR2 memory can be seen as an upgrade and extension of DDR memory, DDR2 memory to the "2bit Prefetch (2-bit prefetch)" Technology of DDR memory to "4 bit Prefetch (4-bit prefetch)" mechanism, At the same core frequency, its effective frequency is a full increase over the DDR memory, with the same memory bit width, the memory bandwidth has been increased by a whole, which is very useful for the performance improvement of the graphics card. Technically, the DRAM core of the DDR2 memory can be accessed in parallel, processing 4 data per access instead of 2 data for DDR memory, so DDR2 video memory enables processing of 4bit data per clock cycle, up to a factor of 2bit data processing compared to traditional DDR memory. Another improvement in DDR2 memory compared to DDR memory is that it replaces the traditional TSOP mode with a 144Pin spherical pin FBGA package, and the operating voltage is reduced from 2.5V to 1.8V. As the DDR2 memory provides higher frequency, the performance of the corresponding improved, but also brought high calorific disadvantage. Coupled with structural constraints that cannot be used in inexpensive TSOP packages, higher cost BGA packages have to be used (DDR2 's initial production capacity is insufficient and the cost is more problematic). High calorific value, expensive to use DDR2 memory graphics card, such as the first use of DDR2 memory GeForce FX 5800/5800ultra Series video card is a relatively unsuccessful product. For these reasons, DDR2 is not widely used on mainstream graphics cards.DDR3 DDR3Video memory can be regarded as an improved version of DDR2, the two have many similarities, mainly using 144Pin spherical pin FBGA package. But the DDR3 core has improved: DDR3 memory uses 0.11 micron production process, the power consumption is significantly lower than DDR2. In addition, DDR3 memory using the "Pseudo Open Drain" interface technology, as long as the voltage is appropriate, the display chip can directly support DDR3 memory. Of course, the long delay time of video memory particles (CAS latency) has been a major problem of high-frequency video memory, DDR3 is no exception, DDR3 CAS latency for 5/6/7/8, in contrast DDR2 for 3/4/5. Objectively speaking, DDR3 relative to DDR2 in technology and no rapid progress, but DDR3 's performance advantage is still relatively obvious: (1) power consumption and heat is small: absorb the lessons of DDR2, on the basis of cost control to reduce energy consumption and calorific value, making DDR3 easier to be accepted by users and manufacturers. (2) Higher operating frequency: Due to lower energy consumption, DDR3 can achieve a higher operating frequency, to a certain extent, to compensate for the long delay time of the shortcomings, but also as one of the selling point of the graphics card, which in the graphics card with DDR3 memory has been shown. (3) Reduce the overall cost of the graphics card: DDR2 memory particle size is 4M X 32bit, with the high-end graphics commonly used 128MB video memory will need 8. and DDR3 memory specifications for more than 8M X 32bit, single particle capacity is larger, 4 can constitute 128MB video memory. As a result, the graphics card PCB area can be reduced, the cost is effectively controlled, in addition, after the reduction of the number of particles, memory power consumption can be further reduced. (4) Good versatility: better compatibility with DDR2 than DDR changes to DDR2,DDR3. As the PIN, package and other key characteristics, with the DDR2 display core and the public version of the design of the graphics card can be modified to use DDR3 memory, which is a great benefit to manufacturers to reduce costs. At present, DDR3 has been widely used in most of the new high-end graphics cards.GDDR5GDDR5 (Graphics Double Data rate, version 5) SDRAM is a type of high-performance DRAM memory designed for high-frequency bandwidth requirements for computer applications. Like its previous version (GDDR4), GDDR5 is based on DDR3 SDRAM, which has a double rate compared to DDR2 SDRAM, but GDDR5 also has a 8-bit wide prefetch cache similar to GDDR4.Now Samsung and GDDR5 have started to develop their own plans. Samsung Electronics will start from the third quarter to take out 5000/3200mibps specifications of GDDR5 memory, this memory in 170FBGA package, gddr5x32 512Mbit G-die process, operating voltage 1.5v. At the end of this year to start production of 1Gb GDDR5 memory, specifications Gddr5x32,170fbga,d-die process, working voltage 1.5v. The GDDR5 is also expected to run the production of video memory at the end of this year, the first quarter of next year. Today's graphics card is generally equipped with the type of memory is GDDR3,GDDR4 only a small number of high-end models, and determined to leap the development of the singular dream has been pilot-run the world's first GDDR5.5Performance CapacityVideo MemoryMemory capacity is the number of local video memory on the video card, which is one of the key parameters for selecting a video card. The size of the memory capacity determines the ability of the memory to temporarily store data, and to some extent, it can also affect the performance of the graphics card. The memory capacity is gradually increasing with the development of the graphics card, and there is an increasing trend. Memory capacity from the early 512KB, 1MB, 2MB and other minimum capacity, the development to 8MB, 12MB, 16MB, 32MB, 64MB, has been to the current mainstream 512MB, 1GB and high-end graphics card 2GB, some professional graphics card even has a 4GB of video memory. In the maximum resolution of the graphics card, the maximum resolution is somewhat directly related to the memory, since the data of these pixels is initially stored in the memory, so the memory capacity affects the maximum resolution. In the early video card memory capacity only with 512KB, 1MB, 2MB and other minimum capacity, memory capacity is really a bottleneck of the maximum resolution, but the current mainstream graphics card memory capacity, even 64MB has been eliminated, the mainstream entertainment-grade graphics card is already 512MB or 1GB, Some professional graphics cards even already have 4GB of video memory, in which case the memory capacity is no longer the factor that affects the maximum resolution. In the graphics card performance, as the display chip processing power more and more powerful, especially today's large 3D games and professional rendering needs to temporarily store more and more data, the required memory capacity is also increasing, memory capacity to some extent also affect the performance of the graphics card. For example, when the display core is strong enough and the memory capacity is small, there are a large number of large texture map data to be stored, if the memory capacity is not enough to store the data, then the display core at some time only idle to wait for these data processing is complete, This affects the performance of the display core performance, which also affects the graphics card. It is worth noting that the larger the memory capacity does not necessarily mean that the performance of the graphics card is higher, because the three factors that determine the performance of the graphics card first is the display chip, followed by the memory bandwidth (which depends on the memory bit width and memory frequency), and finally the memory capacity. A graphics card should be equipped with how much memory capacity is suitable for the display chip is determined by its use, that is, the memory capacity should match the performance of the display core is reasonable, the higher the performance of the chip due to its higher processing capacity of the memory capacity of the corresponding should also be greater, And the low performance of the display chip with high-capacity memory for its performance is not any help. For example, some commercially available Radeon 9550 graphics cards with 512MB of high-capacity memory are identical in terms of graphics performance and 128MB video memory Radeon 9550 graphics at the same time as the core frequency and memory frequency parameters, because Radeon 9550 the relatively low processing power of the display core determines that it does not make sense to be equipped with large-capacity memory, while the high-capacity video memory also brings the purchase cost improvement problem.DataThe number of data bits refers to the number of bits that can be transmitted within a clock cycle, which is an important factor in determining the bandwidth of the memory and is closely related to the performance of the graphics card. When the video memory type is the same and the working frequency is the same, the greater the number of data bits, the higher the performance. The memory bandwidth is calculated by running frequency × data bandwidth/8. With the current GEFORCE3 graphics card as an example, its memory system bandwidth =230mhzx2 (because of the use of DDR memory, so multiplied by 2) x128/8=7.36gb. The number of data bits is the video memory is also a very important parameter of the graphics card. The z buffers, frame buffers, and texture buffers can occupy a significant amount of memory bandwidth during the operation of the graphics card. Bandwidth is the 3D chip and the local memory transmission data Volume standard, this time the memory capacity is not important, also does not affect the bandwidth, the same memory bandwidth of the graphics card with 64MB and 32MB display performance difference is not small. Because of this time the bottleneck of the system in the memory bandwidth, when encountered a large number of pixel rendering work, memory bandwidth will cause data transmission congestion, causing the display chip waiting and affect the speed. The video memory is divided into 64-bit and 128-bit, and at the same operating frequency, the 64-bit memory bandwidth is only half of the 128-bit memory. This is why Geforce2 MX200 (64-bit SDR) is far less powerful than Geforce2 MX400 (128-bit SDR).Memory bit width memory Memory bit width is the number of bits that can transmit data in a clock period, the larger the number of bits, the greater the amount of data that can be transmitted instantaneously, which is one of the important parameters of video memory. There are 64-bit, 128-bit, and 256-bit video memory bits on the market, and the 64-bit video card, 128-bit graphics card, and 256-bit video card are used to refer to its corresponding memory bit width. The higher the memory bit width, the higher the performance price, so the 256-bit wide memory is more applied to the high-end graphics card, and the mainstream graphics card basically uses 128-bit video memory. You know the memory bandwidth = memory frequency x Memory bit width/8, then in the memory frequency equivalent, the memory width will determine the size of the memory bandwidth. For example, the same memory frequency of 500MHz 128-bit and 256-bit video memory, then both of the memory bandwidth will be: 128-bit =500mhz*128∕8=8gb/s, and 256-bit =500mhz*256∕8= 16gb/s, 128-bit, twice times, The importance of the video memory bit width in the memory data is visible. Video card memory is composed of a block of memory chips, the total memory width is also composed of memory particles of the bit width. Memory bit width = memory particle bit width × memory particle count. Memory particles with the relevant manufacturers of the number of memories, you can go online to find its number, you can understand its bit width, and then multiplied by the number of memory particles, you can get the video card bit width. This is the most accurate method, but the implementation of the more troublesome to teach you a more simple, but only to adapt to the general situation, there are some special circumstances, in most cases can be applied. At present, the format of memory is mainly TSOP and BGA two kinds, in general, the memory of BGA package is 32 bits/Star, and TSOP package is 16 bits? /capsules. If the video card uses four BGA package of video memory, then its bit width is 128 bits, and if it is eight tsop packaging particles, then the bit width is also 128 bits, but if the video card only uses four TSOP packaging particles, then the memory bit width is only 64 bits. This is only a general situation of the skills, not necessarily in line with all the situation, to achieve the most accurate judgment, or to look at the memory number it!BandwidthThe memory bandwidth is the bridge between the display chip and the memory, the larger the bandwidth, the faster the communication between the chip and the memory is displayed. In order to indicate this width, the memory bandwidth is in bytes per second. The bandwidth of the video memory is related to the bit width of the video memory and the speed of the video memory (that is, the operating frequency). The final conclusion: memory bandwidth = memory bit width x memory frequency/8. The speed of the memory is generally in NS units, the common display is 6ns, 5.5ns, 5ns, 4ns, 3.8ns, until 1.8ns. The corresponding operating frequency is 143MHz, 166MHz, 183MHz, 200MHz, 250MHz, until 550MHz. The calculation of the operating frequency is very simple-the inverse of the memory speed is the rated operating frequency of the video memory, such as the video memory clock period of 4ns, then the video memory running frequency is 1/4ns=250mhz (if the DDR memory is multiplied by 2). Memory Capacity = display Resolution x number of color digits/8bit. For example, the display resolution is basically 1024x768, the number of color bits is 32bit, then the required memory capacity =1024x768x32bit/8bit=3145728 byte, but this is for 2D graphics (normal plane), if it is 3D accelerator card, Then the required memory capacity is 1024X768X32BITX3/8BIT=9437184BYTE=9.216MB, which is the minimum requirement, but also must increase a certain capacity as the texture display memory, otherwise when the display resources are fully occupied, the computer only occupy the main memory as the texture memory , such two calls will cause the display performance to degrade, so the memory capacity of the real 3D accelerator card must be greater than 9.216MB. Workstation display memory are above 64MB. For example, 2D drawing applications, even in the case of 1600x1200, it is 1600X1200X32BIT/8BIT=7680000BYTE=7.5MB, if it is a three-dimensional drawing such as 3D Studio Max, then the capacity requirements are 7.5X3=22.5MB , but this is the minimum requirement, so the 32MB capacity of the video memory is to cope with this kind of 2D graphics or entertainment, the normal three-dimensional design. For workstations, the memory is more than 256M because it runs larger software and larger operations.Basic KnowledgeThe speed of the video memory is generally in NS units. Common display of 7ns, 6ns, 5.5ns, 5ns, 4ns or even 3.8ns of video memory. The corresponding rated operating frequencies are 143MHz, 166MHz, 183MHz, 200MHz and 250MHz respectively. Rated operating frequency =1/memory speed. Of course, for some of the better quality of memory, the actual maximum operating frequency of memory is a certain amount of margin. Video Memory overclocking is based on this principle, such as the rated frequency of 6ns of memory to the operating frequency of over 190MHz. Here is also a talk about the actual operating frequency of video memory and the equivalent operating frequency. DDR memory is capable of transmitting data on both the rising and falling edges of the clock, so the memory bandwidth is twice times the normal SDRAM at the same clock frequency and data bit width. In other words, the actual operating frequency of the DDR memory is twice times the normal SDRAM memory in the same memory speed. Similarly, the bandwidth of the DDR memory is twice times the normal SDRAM memory. For example, 5ns of SDRAM memory operating frequency of 200MHZ, and 5ns of the equivalent operating frequency of DDR memory is 400MHZ but to understand that the graphics card manufacturing, the manufacturer set the actual video memory operating frequency, and the actual operating frequency is not necessarily equal to the maximum frequency of memory. This kind of situation is now more common, such as video memory maximum can work at 650 MHz, while manufacturing video card operating frequency is set to 550 MHz, there is a certain amount of overclocking space memory. This is the current method used by manufacturers, video cards to overclocking as a selling point.ListFor video card memory, although and the motherboard used by the same name as DDR, DDR2 or even DDR3, but due to the specification parameter difference is large, can not be universal, so it may also be called video memory for Gddr, GDDR2, GDDR3. Memory speed corresponds to DDR frequency 6NS 166MHZ 333mhz5ns 200MHZ 400mhz4ns 250MHZ 500mhz3.6ns 278MHZ 556mhz3.3ns 300MHZ 600mhz2.8ns 360MHZ 720M Hz2.2ns 450MHZ 900mhz2ns 500MHZ 1000MHZ6Memory Frequencythe memory frequency is the default, in MHz (megahertz), the frequency at which the display is working on the video card. The video memory frequency reflects the speed of the video memory to a certain extent. Video memory frequency with the type of memory, the performance of different, SDRAM memory is generally working in the lower frequency, is generally 133MHz and 166MHz, memory frequency, mainly in the low-end video card use, DDR2 memory due to high cost and performance in general, so the use of a small amount. DDR3 memory is the most widely used memory type for high-end graphics. Different video memory can provide video memory frequency is also very different, mainly 400MHz, 500MHz, 600MHz, 650MHz, etc., high-end products also have 800MHz, 1200MHz, 1600MHz, or even higher. The memory frequency is correlated with the memory clock period, which is the reciprocal relationship, that is, the memory frequency =1/the memory clock period. If it is SDRAM memory, its clock period is 6ns, then its memory frequency is 1/6ns=166 MHz. For DDR SDRAM or DDR2, DDR3, whose clock period is 6ns, its memory frequency is 1/6ns=166 MHz, but to understand this is the actual frequency of DDR SDRAM, not what we normally call the DDR memory frequency. The DDR transmits two data in one cycle, which is twice times the SDRAM frequency, because it is transmitted in both the clock rise and fall periods. The traditionally called DDR frequency is its equivalent frequency, which is the equivalent frequency when the actual operating frequency is superior to 2. Therefore 6ns of DDR memory with a memory frequency of 1/6ns*2=333 MHz. The details can be seen below about the various video memory introduction. But to understand is the graphics card manufacturing, the manufacturer set the actual video memory operating frequency, and the actual operating frequency is not necessarily equal to the maximum frequency of memory. This kind of situation is more common, such as video memory maximum can work at 650 MHz, while manufacturing video card operating frequency is set to 550 MHz, there is a certain overclocking space memory. This is the manufacturer's usual method, the video card to overclocking as a selling point. GDDR3.7Memory PackageMemory packaging refers to the type of packaging technology used in memory particles, package is to wrap the memory chip, to avoid the chip and the outside contact, to prevent damage to the chip. Impurities and undesirable gases in the air, and even water vapor, will corrode the precision circuitry on the chip, resulting in a decline in electrical performance. The different packaging technology has great difference in manufacturing process and process, and it also plays a very important role in the performance of memory chip after encapsulation. Memory package form mainly QFP, tsop-ii, MBGA, etc., Tsop-ii, MBGA is more common. Many of the early SDRAM and DDR memory uses TSOP-II, and now with the increase in memory speed, more and more memory uses the MBGA package, especially DDR2 and DDR3 memory, all using MBGA encapsulation. In addition, many manufacturers also will DDR2 and DDR3 memory of the package called FBGA, which is more emphasis on the name of the pin arrangement, is actually the same package form. In addition, although MBGA and tsop-ii can achieve higher video memory frequency, but it is not easy to think that the MBGA package of memory must be better overclocking, because it is easy to overclock, more depends on the factory agreed to the default frequency and the actual video memory can achieve the gap between the frequency, including the design and manufacture of graphics card, Simply put, the MBGA package can reach higher frequencies, but its default frequency is also higher.QFP QFPis the package's abbreviation, is "The small square plane encapsulation" the meaning. The QFP package is used more frequently on earlier graphics cards, but less than 4ns in QFP package memory, because of the process and performance issues, is now gradually replaced by Tsop-ii and BGA. The QFP package has pins around the particles and is clearly recognizable. Tsop-ii (Thin Small out-line package, low profile small footprint). The TSOP package is a pin that is made around the chip and is directly attached to the surface of the PCB using SMT technology (Surface Mount Technology). TSOP package dimensions, parasitic parameters (the current greatly changes, causing output voltage disturbance) reduced, suitable for high-frequency applications, the operation is more convenient, high reliability. At the same time, TSOP package has the advantages of high yield and low price, so it has been widely used. The TSOP package is currently the most widely used memory package type. TSOP-II Package pins on both sides of the video memory.MBGA MBGARefers to miniature ball grid array packages, which are all referred to in English as the Micro Balls grid array package. Unlike the TSOP memory chip, the MBGA pin is not exposed, but is parasitic on the bottom of the chip in the form of tiny tin spheres, so this memory is not visible to the pin. MBGA has the advantages of less noise, good heat dissipation, excellent electrical performance, can be connected to more than a few feet, and can improve yield. The most prominent is due to the internal components of the smaller interval, the signal transmission delay is small, can make a greater frequency increase. The advantages of MBGA package are less noise, better heat dissipation, better electrical performance, more feet can be connected, and can improve the yield. The most prominent feature is that the internal components are spaced smaller, the signal transmission delay is short, can make the frequency of a larger increase. The MBGA memory performance is excellent compared to the TSOP package memory. But also on the circuit wiring requirements, the former as long as 66Pin, the lead is very long, and are lying on the PCB, design, welding, processing and testing is relatively easy, while the latter area of the former is only about 1/4, but there are 144Pin, each pin is a small size of tin ball, design and production is more difficult. Many of the early SDRAM and DDR memory uses TSOP-II, and now with the increase in memory speed, more and more memory uses the MBGA package, especially DDR2 and DDR3 memory, all using MBGA encapsulation. In addition, many manufacturers also refer to the DDR2 and DDR3 memory packages as FBGA, which is more emphasis on the name of the pin arrangement, is actually the same package form.8Clock CycleThe memory clock period is the repetition period of the memory clock pulse, which is an important index to measure the memory speed. The faster the memory speed, the greater the amount of data exchanged per unit of time, and the performance of the graphics card will be significantly improved in the same situation. The clock period for video memory is typically in NS (nanoseconds) and the operating frequency is measured in MHz. The memory clock cycle corresponds to the operating frequency one by one, and the relationship between them is: Operating frequency =1÷ clock cycle x1000. Then the memory frequency is 166MHz, then its clock cycle is 1÷166x1000=6ns. For DDR SDRAM or DDR2, DDR3 memory, the equivalent output frequency is used to describe the operating frequency. Since the data can be transmitted on both the rising and falling edges of the clock cycle, the memory bandwidth is twice times that of SDRAM in the same operating frequency and data bit width. In other words, the equivalent output frequency of the DDR SDRAM memory is twice times that of the SDRAM memory in the same memory clock cycle. For example, 5ns of SDRAM memory operating frequency of 200MHz, and 5ns of DDR SDRAM or DDR2, DDR3 memory equivalent operating frequency is 400MHz. Common Memory clock cycles are 5ns, 4ns, 3.8ns, 3.6ns, 3.3ns, 2.8ns, 2.0ns, 1.6ns, 1.1ns, or even lower.

About the graphics card's video memory