What is a Server CPU?

Server CPU is the CPU (central processing unit) used on the server. The server is an important device in the network. It has to receive access from a few dozens to thousands of people, so it has strict requirements for fast throughput of large data volume, super stability, and long time operation. Therefore, the CPU is the "brain" of the computer and is the primary indicator of server performance.

Ⅰ Classification of server CPU 

The CPU of the server is still distinguished by the instruction system of CPU, which is usually divided into two categories:  CISC  CPU and  RISC CPU, and later there is a 64-bit  VLIW  (Very Long Instruction Word) instruction system CPU.

CISC server CPU

CISC is the abbreviation of "Complex Instruction Set Computing". It refers to the x86 (an Intel CPU naming specification) series CPUs produced by Intel and its compatible CPUs (CISC), which are based on the PC (personal computer) architecture. CISC type CPUs are mainly Intel server CPUs and AMD server CPUs.

Intel server CPU Xeon

Advantages

From the current server development status, the IA architecture (CISC architecture) PC server, which is characterized by "small, compact and stable", has been more widely used due to its reliable performance and low price. In the field of Internet and LAN, they are used for file service, printing service, communication service, Web service, e-mail service, database service, application service, etc.

Disadvantages

The biggest drawback of  IA-64  microprocessors  is their lack of compatibility with x86, and Intel, in order for  IA-64 processors to better run software from both dynasties, has on IA-64 processors (Itanium, Itanium2 ...) An x86-to-IA-64 decoder was introduced so that it could translate x86 instructions to IA-64 instructions. This decoder was not the most efficient decoder, nor was it the best way to run x86 code (the best way was to run x86 code directly on an x86 processor), so Itanium and Itanium2 had very poor performance when running x86 applications. This became the root cause of the creation of X86-64.

One final point worth noting is that although the CPU is one of the most important factors in determining server performance, it cannot perform as well as it should without the support and cooperation of other accessories.

The biggest drawback of IA-64 microprocessors is their lack of compatibility with x86. Intel, on the other hand, introduced x86-to-IA-64 decoders on IA-64 processors such as Itanium and Itanium2 in order for IA-64 processors to better run software from both dynasties. This enables the translation of x86 instructions into IA-64 instructions. This decoder was not the most efficient decoder, nor was it the best way to run x86 code (the best way is to run x86 code directly on an x86 processor), so Itanium and Itanium2 had very poor performance when running x86 applications. This became the root cause of the creation of X86-64.

Although the CPU is one of the most important factors in determining the performance of a server, it cannot perform as well as it should without the support and cooperation of other accessories.

RISC server CPU

RISC is the abbreviation of "Reduced Instruction Set Computing". It is developed on the basis of the CISC (Complex Instruction Set Computer) instruction system, and some people have tested CISC machines to show that the frequency of various instructions is quite disparate. The most frequently used are some relatively simple instructions, which account for only 20% of the total number of instructions, but 80% of the frequency in the program. A complex instruction system is bound to increase the complexity of the microprocessor, making the processor development time long and costly. And complex instructions require complex operations, which will inevitably reduce the speed of the computer.

For these reasons, the  RISC CPU was born in the 1980s, which not only streamlined the instruction system compared with the CISC CPU but also adopted a structure called "superscalar and super pipeline", which greatly increased the parallel processing capability. (Parallel processing means that a server has multiple CPUs processing simultaneously. Parallel processing can greatly improve the data processing capability of the server. Department-level and enterprise-level servers should support CPU parallel processing technology.) In other words, the architecture is much higher performance than CPUs with CISC architecture at the same frequency, which is determined by the technical characteristics of CPUs. RISC instruction system is more suitable for UNIX, the operating system of high-grade servers, and Linux is also a UNIX-like operating system.

The main types of CPUs that use RISC instructions in mid- to high- grade servers are as follows.

(1) PowerPC processors

(2) SPARC processors

(3) PA-RISC processors

(4) MIPS processor

(5) Alpha processor

Currently, the two major CPU processor instruction systems, CISC and RISC architectures are both taking advantage of each other's strengths and weaknesses and moving toward convergence. CISC borrows ideas from RISC to optimize the efficiency of the instruction system, while RISC introduces enhanced instructions to improve the efficiency of complex task processing. Therefore, there is no need to be overly concerned about the difference between CISC and RISC, both architectures are very advanced and will evolve over time.

CISC complex instruction set is characterized by a large number of instructions, one instruction performs multiple functions. The advantage is the high efficiency of executing specific functions, such as multimedia processing. The disadvantage is the complex system design and low execution efficiency. Typical architectures include x86.

The RISC reduced instruction set is characterized by fewer instructions, and complex tasks are completed by combining multiple lean instructions. The advantages are high execution efficiency for common tasks and low power consumption. The disadvantage is low processing efficiency for some complex tasks, such as multimedia processing. Typical architectures are ARM, Power, MIPS, Alpha, and SPARC.

VLIW server CPU

VLIW is the abbreviation of "Very Long Instruction Word". VLIW architecture adopts advanced EPIC (Explicit Parallel Instruction) design, we also call this architecture "IA-64 architecture". EPIC processors are mainly Intel's IA-64 (including Intel's Anthem processor) and AMD's x86-64.

Ⅱ Server CPU vs. desktop CPU

1. Different instruction sets

The common CPUs equipped with home or work computers are usually CISC complex instruction sets, which pursue a large and comprehensive instruction set and try to integrate various common functions into one piece, but the calling speed and hit rate are lower compared to server CPUs. The instructions of server CPU are generally RISC (Reduced Instruction Set). The advantage of this design is that it is more targeted and can be optimized specifically for different needs and is more energy-efficient.

2. Different cache

Cache also determines the performance of the CPU. Due to the high computing performance requirements of server CPUs, server CPUs often apply the most advanced processes and technologies and are equipped with one, two, or three caches, running more powerful. Server CPUs have used level 3 cache for a long time. Ordinary CPUs have used cache technology only in recent years.

3. Different interfaces

Server CPU and ordinary CPU interfaces are often different. Server CPU interfaces are mostly Socket 771, Socket 775, LGA 2011, LGA 1150 compared to ordinary CPU interfaces. Server CPU equipped with the motherboard usually no graphics card slot, because the CPU comes with a core graphics card to meet the demand, and its CPU bus bandwidth is higher than the home CPU.

4. Different stability requirements

Server CPUs are designed to work stably for long periods of time and are basically designed to work continuously all year round. Server CPUs are vastly different from home CPUs in terms of stability and reliability, and servers are generally run 365 days a year, with only occasional downtime for maintenance, which requires high stability.

Ordinary CPUs are designed to work continuously for 72 hours.

5. Multi-way interconnection support is different

Multiplexing is a technology used in servers. For example, a server motherboard can have multiple CPU slots at the same time, allowing multiple CPUs to be installed at the same time, which is CPU multi-channel interconnect technology. This technology is only supported by server CPUs, and ordinary home computers, where only one CPU can be installed on a motherboard, do not support multiplexing.

6. Different prices

Multiplexing is a technology used in servers. For example, a server motherboard can have multiple CPU slots at the same time, allowing multiple CPUs to be installed at the same time, which is CPU multi-channel interconnect technology. This technology is only supported by server CPUs, and ordinary home computers, where only one CPU can be installed on a motherboard, do not support multiplexing.

Ⅲ Top 10 Popular Server CPUs 

1 Intel Xeon E5-2678 v3

2 AMD Skyline 7763

3 Intel Xeon E5-2680 v2

4 Intel Xeon Silver 4210R

5 Intel Xeon Gold 5218

6 Intel Xeon E5-2650 v2

7 Intel Xeon E5-2680 v4

8 Intel Xeon Silver 4210

9 Intel Xeon Gold 5218R

10 Intel Xeon E5-2670 v2