Significance of several indicators (CPU, FSB, Memory, GPU)
CPU frequency: frequency = outer frequency * multiplier. The clock frequency is also called the clock frequency, the unit is MHz, used to represent the CPU's arithmetic speed. Many people think that the CPU's main point is the speed of the CPU, in fact, this understanding is a one-sided. The clue of the CPU indicates the speed of the digital pulse signal in the CPU, and there is no direct relationship with the actual calculation capacity of the CPU. The frequency and actual calculation speed is related, but there is currently no certain formula to quantify the numerical relationship between the two, and the CPU's calculation speed depends on the performance indicators (cache size, instruction sets, "in all aspects of the CPU. The number of copies of the CPU, etc.). Since the frequency is not directly representing the speed of operation, it is likely to have a lower frequency higher CPU actual operation rate. Therefore, the frequency is only one aspect of the CPU performance, and does not represent the overall performance of the CPU.
The clock is the product of outer frequency and frequency. Why use foreign frequency and multiplier to describe CPUs? Because after 486, since the CPU operating frequency continues to increase, some other devices (such as board cards, hard drives, etc.) are limited by the process, and the higher frequency cannot be spaced, so that the CPU frequency is further improved. Therefore, multiplier technologies have occurred, which enables a multiplication of the internal operating frequency of the CPU to an external frequency, thereby reaching the improvement frequency by lifting the multiplier.
The FSB (front-end bus) is inside the PC, and one device communicates digital signals through the system bus (BUS). The CPU can communicate with memory, graphics cards and other devices via the front-end bus (FSB). The faster the FSB frequency, the processor gets more data in unit time, the higher the processor utilization. The front-end bus frequency directly affects the bus speed of the CPU and memory direct data exchange. Due to the adoption of special techniques, the bus existing in CPU and memory (CPU through the North Bridge and memory switching data) can be completed twice or even 4 times in a clock cycle, so it is equivalent to a frequency improvement. . (Ie the frequency of the CPU field.)
Intel and AMD are different from technology in FSB. Intel FSB frequency = CPU Overseas * 4, for example: 2.4C Overseofield 200MHz, FSB frequency 800MHz AMD FSB frequency = CPU Overseas * 2, for example: Athlon XP 2500 (Barton) Overseofa 166MHz, FSB frequency 333MHz
Division technology: Since the CPU is continuously improved, other devices cannot withstand such high frequencies, and therefore frequency division technology (implemented by the motherboard North Bridge chip). The frequency division technology is to reduce the Frequency of the CPU through the main board of the North Bridge chip, and then supplied to each board, hard disk and other devices. Early 66MHz field era is PCI device 2 divided, AGP equipment is not divided; later 100MHz Overseofa is a PCI device 3 divided, AGP device 2/3 divided (some 100MHz North Bridge chip also supports PCI equipment 4 Dividend); current North Bridge chip generally supports 133MHz extravader, ie PCI device 4 divided, AGP device 2 divided. Push it in this class. In summary, under the standard extra frequency (66MHz, 100MHz, 133MHz, 200MHz), the North Bridge chip makes the PCI device at 33MHz, and the AGP equipment is operated at 66MHz. Non-standard outer frequency: except 66MHz, 100 MHz, 133MHz, 200MH outfitting. When the outer frequency is in a non-standard frequency (often in the case of ultra-frequency), the frequency of the board or the hard disk may also be overclocking, which is not conducive to the stability of the board or hard disk. Fortunately, some motherboards now have locked the PCI and AGP bus frequencies (or support AGP bus frequencies) for convenience overclocking. The FSB bandwidth represents the data transfer speed of the FSB, units MB / S or GB / S. FSB bandwidth = FSB frequency * FSB bit wide / 8, now the FSB bit width is 64 bits. For example: P4 2.0A: FSB bandwidth = 400MHz * 64bit / 8 = 3.2gb / s.
Memory: DDR (DOUBLE DATA RATE) can transmit two data in a clock cycle, frequency constant, and double data transmission capacity. For example, the DDR333 actual operating frequency is 166MHz. The performance of the memory is determined by the bandwidth and delay time, and there is a channel-memory bus between the memory controller and the memory, and the speed of memory bus transmission data is called the memory bus bandwidth. Memory bandwidth (MB / S) = Running frequency (MHz) * Transmission magnification * Bus width (BIT) / 8. The DDR transmission magnification is 2. For example: DDR333 bandwidth = 166 * 2 * 64/8 = 2700MB / s.
Data The transmission between memory and CPU needs to pass through memory bus and FSB, and the speed is limited by the bus. If the memory bandwidth is low, it will become the bottleneck of data transmission, and the CPU cannot get the data required in time, affecting the performance of CPU performance. For example: Intel 865PE FSB bandwidth = 800 * 64/8 = 6.4GB / s; dual channel DDR400 memory bandwidth = 400 * 64 * 2/8 = 6.4Gb / s. Such a matching material does it. If you select the motherboard of the integrated graphics, because the display chip takes a certain amount of memory, it will consume some memory bandwidth to affect the overall performance of the system. At this time, the excess memory bandwidth is required to provide the display chip.
The most critical part of the graphics card: GPU (graphics processing chip) and memory. The relationship between GPUs and memory is similar to CPU and memory. AGP bandwidth, for example: AGP4X bandwidth = 66.6MHz * 4 * 32bit / 8 = 1GB / s; AGP8X bandwidth = 66.6MHz * 8 * 32bit / 8 = 2.1Gb / s. The AGP bus is just a channel for exchange data, providing a possibility of achieving higher bandwidth. True processing of data is a graphic chip, so it cannot be blindly conclusions that the AGP8X is 1x faster than AGP 4x. Several indicators affecting graphics card performance: 1. Pixel filling rate = core frequency * pixel pipeline, for example: G4MX440 pixel fill rate = 270 * 2 = 540Mpixels / s. The higher the pixel filling rate, the more pixels rendered per second, the stronger GPU processing capability. 2. Texture filling rate = core frequency * pixel pipeline * Each line texture unit, for example: G4TI4200 texture fill rate = 250 * 4 * 2 = 2000Mpixels / s. The three-dimensional object is applied to the texture is the basic rendering method. Now most games use multiple textures. Texture filling is completed by the pixel pipeline, and if the pixel line has a texture unit, the texture fill is equal to the pixel filling. The G4Ti4200 each pixel line has two texture units. 3. The memory bandwidth = memory frequency * Transmission magnification * The memory bit width / 8. G4 Ti4200 (128bit) Semism bandwidth = 250 * 2 * 128/8 = 8000Mb / s. Under the premise of rendering speed, the memory bandwidth will limit the performance of the GPU. 4. If you have a memory, you often see it.
