Server CPU, as the name suggests, is the CPU (Center Process Unit Central Processor) used on the server. We know that the server is an important device in the network. To accept less than dozens of people, there are more than 10,000 people, so there is a quick throughput of large data volume, super stability, long run, etc. strict requirements. Therefore, the CPU is the "brain" of the computer, which is the primary indicator for measuring server performance.
At present, the CPU of the server is still distinguishing between the CPU's instruction system, usually divided into two categories of Cisc type CPU and RISC type, and then there is a 64-bit VLIM (Very long instruction Word long instruction set architecture) instruction system CPU.
First, CISC type CPU
CISC is the abbreviation of English "Complex Instruction Set Computer", meaning "complex instruction set", which refers to the X86 (Intel CPU's naming) series CPU and its compatible CPU (other manufacturers such as AMD, VIA) The CPU produced is based on the architecture of the PC (Personal Computer). This CPU is generally 32-bit structures, so we also become the IA-32 CPU. (IA: Intel Architecture, Intel Architecture). CISC type CPU currently has two types of server CPUs with Intel's servers CPU and AMD. (1) Intel's server CPU (2) AMD server CPU
Second, RISC type CPU
RISC is an abbreviation of "Reduced Instruction Set Computing", meaning in Chinese means "Describe Collection". It developed on the basis of the CISC (Complex Instruction Set Computer) instruction system, some people test the CISC machine show that the frequency of use of various instructions is quite a very disappersion, and most commonly used is some relatively simple instructions, they only account for only The total number of instructions is 20%, but the frequency that appears in the program accounts for 80%. Complex instruction systems inevitably increase the complexity of the microprocessor, so that the processor has a long development time and cost. And complex instructions require complex operations, which will inevitably reduce the speed of your computer. Based on the above reasons, the RISC type CPU was born in the 1980s. Compared with the CISC type CPU, the RISC type CPU not only streamlines the instruction system, but also uses a "over-scale and overflow water structure", which greatly increases parallel processing power. (Parallel processing parallel processing means that one server has multiple CPUs. Parallel processing can greatly enhance the server's data processing power. Department level, enterprise-class server should support CPU parallel processing technology). That is, the architecture is in the same frequency, the CPU using the RISC architecture is much higher than the CPU performance of the CISC architecture, which is determined by the technical characteristics of the CPU. The CPU of this instruction system is currently employed in the medium and high-end servers, especially the high-end servers all use the CPU of the RISC instruction system. The RISC instruction system is more suitable for the operating system Unix of the high-end server, and now Linux is also a system similar to UNIX. The CPU of the RISC type CPU and Intel and AMD is not compatible on software and hardware. Currently, the CPU with RISC instructions in the medium and high-end servers have the following categories: (1) PowerPC processor (2) SPARC processor (3) PA-RISC processor (4) MIPS processor (5) alpha processor
From the current development of the current server, the PC server of the IA architecture (CISC architecture) characterized by "small, smart and stable" is a broader application with reliable performance, low price. In the Internet and LAN, for file services, print services, communication services, web services, email services, database services, application services, etc. Finally, it is worth noting that although the CPU is one of the most important factors that determine the performance of the server, if there is no support and cooperation of other parts, the CPU cannot play its performance. _________________________________________________________
Intel production CPU
Starting from the Pentium era, Intel (Intel) launched the CPU - Pentium Pro dedicated to the server, that is, "high energy pentium"; after entering the Pentium Three-Times, Intel has introduced the CPU - Xeon of the corresponding server (workstation), ie "Strong"; the server CPU corresponding to the Pentium Four is also called "Xeon". Such as Lenovo Wanquan 4200 server supports up to 4 Intel Pentium Xeon 700 MHz CPU, built-in 1MB or 2MB full speed cache. This server is a high-end enterprise-class server product of Lenovo. It is a large-scale enterprise, an important industry and other key departments to handle large data volume services, and the key task is good.
At present, Intel's CPU's product line is also very long, the following is technical parameters and specifications of various popular processors:
Note: A.atc is an Advanced Transfer Cache. The processor named "E" with ATC. B. The Front-Side Bus of the processor named "B" is 133MHz.
In the CPU produced by Intel, Pentium 4 (Pentium 4) and Celeron (Cace) are PC, Xeon (Strong), Xeonmp, and Itanium (Anteng) are for workstations and servers. Where ITANIUM is a 64-bit CPU that is completely different from other CPUs, and it is not considered for existing Windows applications. Although other processors have different working frequencies, FSB (front-end bus frequencies) and cache capacity, but internal design is basically the same, while ensuring software compatibility. Pentium 4 (Celeron) and Xeon's maximum difference is Xeon to build multiprocessor systems, while P4 is not. The P4 is built in the system that can only be used in a CPU, and Xeon can build a dual processor system with 2 CPUs, and Xeon MP can be used to build a system in 4 or more CPUs. "MP" is also "Multi Processing Platform" (multiprocessor platform).
Multi-processor systems can be used for high-process speed applications that can be used for single processors such as three-dimensional graphics and animated file coding, and can also be used in high load high speed applications such as database processing in the server (workstation). In addition, P4 (Pentium 4) uses 478 needle packages, Xeon (to strong) uses 604 pin packages, and supports their chipsets, so they cannot be used interchangeably.
AMD production CPU
AMD also produces an Athlon MP processor for workstations and servers. The internal design is basically the same as Athlon XP, but supports dual CPU. "MP" is also an abbreviation of the "Multi Processing Platform" (multiprocessor platform). PowerPC processor
In the 1990s, IBM (International Commercial Machinery), Apple (Apple) and Motorola (Motorola) developed the PowerPC chip success and manufactures PowerPC-based multiprocessor computers. The PowerPC architecture is characterized by good scalability, convenient and flexible. The first generation of PowerPC used a 0.6 micron production process, and the integration of the transistor reached 3 million single chips. In 1998, the copper chip came out and created a new historical era. In 2000, IBM started a large number of products using copper chips, such as the RS / 6000 X80 series products. Copper technology has replaced aluminum technology that has been used in 30 years, and the production process of silicon chips has reached a level of 0.20 microns, and single-chip integrates 200 million transistors, which greatly improves operational performance. The 1.8V low voltage operation (original 2.5V) greatly reduces the power consumption of the chip, easy to heat, thereby greatly improving the stability of the system. MIPS processor
MIPS Technology is a manufacturer designed and manufactured in high performance, high-end and embedded 32-bit and 64-bit processors, accounts for an important role in RISC processors. In 1984, the MIPS computer company was established. In 1992, SGI acquired MIPS computer. In 1998, MIPS was out of SGI and became MIPS Technology. MIPS company design RISC processor began in early 1980s, in 1986, the R2000 processor was launched, and the R3000 processor was pushed in 1988, and the first 64 commercial microproelectric R4000 was launched in 1991. The models of R8000 (in 1994) and R10000 (1996) and R12000 (in 1997) were launched later. Subsequently, MIPS's strategy changes, focusing on embedded systems. In 1999, MIPS issued MIPS32 and MIPS64 architectural standards to lay a foundation for future MIPS processors. The new architecture integrates all the original NIPS instruction sets and adds a lot of more powerful features. MIPS has developed high-performance, low-power 32-bit processor core (CORE) MIPS324KC and high-performance 64-bit processor core MIPS64 5KC. In 2000, MIPS issued a version of MIPS32 4KC and 64-bit MIPS 64 20KC processor kernel.
ALPHA processor
The Alpha processor was originally designed by DEC. After Compaq (Compaq) company acquired DEC, the Alpha processor continued to develop and applied to many high-end COMPAQ servers. Since 1995, 21164 chip has been developed, and the process is 0.5mm, and the frequency is 200MHz. In 1998, the new model 21264 was launched, and the clock at the time was 600MHz. At present, the new 21264 chip has reached 1 GHz, and the process is 0.18mm. The chip has a complete instruction predictive ability and a high storage system bandwidth (more than 1Gb / s), and wherein the multimedia processing capability has been enhanced. The 21264 chip maintains the ALPHA processor to run multiple operating systems, including TRU64Unix, OpenVMs, and Linux, etc., which have many mature applications, which is also an advantage of the Alpha processor.
PA-RISC processor
HP (HP) company's RISC chip PA-RISC came out in 1986. The first chip model is PA-8000, the frequency is 180MHz, and then the PA-8200, PA-8500 and PA-8600 models are launched. The 64-bit microprocessor PA-8700 developed by HP officially puts into servers and workstations in the first half of 2001. This new processor has a design frequency of 800 MHz. The process used by the PA-8700 is a 0.18 micron SOI copper CMOS process. The 7-layer copper conductor is used to interconnect, and the high-speed buffer on the chip reaches 2.25Mb, which is 50% higher than PA-8600. HP has launched the PA-8800 and PA-8900 processors, and its climbers reach 1GHz and 1.2GHz, respectively. RA-RISC is also the foundation of IA-64. In the future IA-64 chip, it will continue to maintain many of the important features of the PA-RISC chip, including the virtual storage architecture of PA-RISC, unified data format, floating point operation, multimedia, and graphic acceleration. SPARC processor
In 1987, Sun and TI cooperated to develop RISC microprocessor -sparc. The most prominent feature of the SPARC microprocessor is its scalability, which is the first microprocessing in the industry. The launch of SPARC has won Sun ahead of the high-end microprocessor market. In June 1999, Ultrasparc III debuted. It uses advanced 0.18 micron process manufacturing, all with 64-bit structure and VIS instruction sets, clock frequencies from 600MHz, which can be used to work with up to 1000 processors. The application of the Ultrasparc III and the Solaris operating system implements 100% binary compatible, fully supports customers' software investment, and supports many independent software vendors. In the 64-bit Ultrasparc III processor, Sun has mainly 3 series. The first is the scalable S series, mainly for high performance, easy to extend multiprocessor systems. At present, the frequency of UltrasParc IIS has reached 750 MHz. There are also models such as Ultrasparc IVS and Ultrasparc Vs. The frequency of Ultrasparc IVS is 1 GHz, and Ultrasparc Vs is 1.5GHz. Secondly, the integrated i series, which integrates a variety of system functions on one processor, providing a higher benefit to a single processor system. The frequency of Ultrasparc III i has reached 700 MHz, and the frequency of future UltrasParc IVI will reach 1GHz.
