Server Processors: The Core of a Server’s Performance

In the modern world of IT technology, enterprises have grasped the necessity of initiating their digital transformation as soon as possible. The tools they choose are often high-performing, highly reliable servers—preferably ones with smaller, more eco-friendly carbon footprints. These servers are the driving force behind supercomputers and data centers. The core of their power comes from the server processors, also known as server CPUs, which are inside every server.

Learn More:
《Glossary: What is IT?》
《Glossary: What is Data Center?》
《Japanese Telco Leader KDDI Invents Immersion Cooling Small Data Center with GIGABYTE》

The server processor is the central processing unit, or CPU, at the heart of a server. Its role is to carry out “instruction cycles”—that is, work with the other components inside a server to perform the calculations that comprise the primary workload of a server. This is how servers complete a myriad of tasks, from storing data and hosting webpages to researching solutions for climate change and helping scientists explore outer space.

Learn More:
《Decoding the Storm with GIGABYTE’s Computing Cluster》
《Lowell Observatory Looks for Habitable Exoplanets with GIGABYTE Servers》

The CPU itself may not seem like much to look at; it’s a chip you can hold in your hand. Inside, there is a complete set of tools for running calculations, such as the program counter, the CPU register, and the arithmetic logic unit (ALU). They are used to perform instruction cycles, which can be separated into four steps: fetch, decode, execute, and write-back. For this reason, the instruction cycle is also called the fetch-decode-execute cycle.

Every time we run a program, the CPU “fetches” instructions and data from the random access memory, or RAM. It “decodes” the instructions to understand what it needs to do. Based on how the processor's instruction set architecture (ISA) is designed, the instructions may be in the CISC or RISC format.

Glossary:
《What is CISC?》
《What is RISC?》

At the “execute” stage, the CPU uses its internal components, such as the ALU, to perform the calculation and execute the command. Lastly, the CPU “writes back” the results into the RAM, thus completing the instruction cycle. A server processor may complete billions of instruction cycles in a single second.

On the matter of ISAs, it’s worth noting that whether a server processor follows the RISC or CISC architecture affects how the data it uses is structured and how the CPU works with the RAM. CISC processors support complex instructions that can be carried out across multiple clock cycles, which is why it can work with much smaller code sizes and doesn't require as much memory to store all the lines of instructions. On the other hand, the RISC approach uses simple instructions that can be executed within a single clock cycle, which may result in longer code sizes and require greater memory capacity. The performance of the two types of server processors are ultimately around the same, with RISC processors gaining an edge in power efficiency and heat dissipation due to its less complicated architecture.

The predominant type of CSIC processors is known as "x86", named after a series of Intel processors launched in the 1980s. Currently, the two movers and shakers in the x86 market are Intel and AMD. The most popular type of RISC processors is the ARM processor, which is widely used in smartphones and other mobile devices. Between 1985 and 2021, 200 billion ARM chips were shipped worldwide. ARM processors are making headways into the server market. "Fugaku", one of the world's leading supercomputers, runs on ARM processors.

Learn More:
《More information about GIGABYTE’s x86 Server》
《More information about GIGABYTE’s ARM Server》
《Tech Guide: The Advantages of ARM, From Smartphones to Supercomputers and Beyond》

Readers who are versed in tech may be wondering: is a server processor any different from a personal computer (PC) processor? Here are four crucial ways in which they are not the same:

● Number of Users and Amount of Data
The biggest difference is the amount of data and the number of users a server processor has to work with, compared with a PC processor. Servers work with terabytes of data and are accessed by hundreds, if not thousands of concurrent users. For this reason, server CPUs need to calculate faster and work with more data—to a level sometimes known as high performance computing (HPC).《Glossary: What is HPC?》

● Availability and Reliability
Reliable operations is a nonnegotiable characteristic of the server processor. CPUs in servers are working 24 hours a day, 7 days a week. “High availability” features are a basic requirement in most data centers because offline servers may inconvenience a huge number of users and cause an incalculable amount of financial damage. In comparison, as any PC user knows, PC processors may be prone to crashing, but it’s usually not a big deal.《Glossary: What is High Availability?》

● CPU Cache
A CPU cache is a hardware cache that acts as an intermediary between the RAM and the CPU. By storing frequently used instructions or data in the CPU cache, the time or energy it takes to retrieve information from the main memory is reduced, which improves performance. Server processors generally employ a hierarchy of cache levels, from level 1 (L1) to level 3 (L3), with a lower number denominating a smaller cache size but faster speed. A PC processor may also employ multi-level caches, but the requirements are not nearly as demanding.

● CPU Sockets
The thinking goes like this: if one processor is good, two must be better. While a desktop motherboard will rarely need to support more than one surface-mounted CPU, server motherboards usually come with multiple CPU sockets for multiple server CPUs. Common types of CPU sockets include the Pin Grid Array (PGA), which has pins on the processor and holes in the socket; and the Land Grid Array (LGA), which has pins in the socket. CPU sockets make it easier for processors to be swapped when necessary.

Learn More:
《Glossary: What is CPU Socket?》
《More information about GIGABYTE’s Server Motherboard》

From all of this, we can see that the gap in computing power between a server processor and a PC processor is to such a degree that a lot of their peripheral components and functions may differ as well. In the next section, we will dive deeper into the different elements that make up a server processor. You will see that in addition to CPU caches and CPU sockets, other factors such as the number of cores and threads and the clock speed also affect the performance of a server processor.

Glossary:
《What is Core?》
《What is Thread?》

 ● Core
The core is the hub of the CPU in which the instruction cycles are carried out. It used to be that a single CPU possessed a single core; nowadays, a single CPU can house a multitude of cores, each capable of carrying out its own instruction cycles. This boosts the server’s performance to unprecedented heights. Larger workloads can be executed by multiple cores simultaneously through a process known as parallel computing. The number of cores is an important deciding factor in the server processor’s performance, but the number of threads and the clock speed must also be considered.《Glossary: What is Parallel Computing?》

● Thread
Threads are the smallest sequence of instructions that can be executed independently within each core. Usually, an advanced CPU core contains two threads—it can be thought of as the core initiating a new instruction cycle before the previous cycle is fully complete. Through a process known as multithreading (or hyperthreading), a core can essentially double its performance. Many HPC servers fully exploit their cores and threads to reach the level of supercomputers.

● Clock Speed
Clock speed, also known as clock rate, is a common indicator of a processor’s speed. It refers to the frequency at which the processor generates the clock signals, or “pulses”, that synchronize operations within the processor. The measurement unit is “clock cycles per second”, or hertz (Hz). Since the clock speed of modern server processors are usually measured in gigahertz (GHz), this means they are running billions of clock cycles every second! It may be tempting to opt for CPUs with the highest clock speeds, but as with everything else about server processors, no single indicator is the be-all and end-all. Throughput between processors and RAM, the cache architecture, the nature of your workloads—all these factors must be considered when choosing the optimal processor for your server.

GIGABYTE Technology has a variety of server solutions that support some of the most advanced server processors on the market. Whether you are looking for x86 processors based on CISC, such as the AMD EPYC™ and Intel® Xeon® Scalable series of products; or ARM processors based on RISC, which are represented by products from Ampere Computing—GIGABYTE has got you covered.

Both RISC and CISC processors are excellent choices for modern servers. Broadly speaking, x86 CPUs are ubiquitous in today’s data centers and IT infrastructure because they profit from the complete hardware and software ecosystem that is the result of Intel and AMD's long years of developing the market; they also offer incredibly fast response time and excel at tackling complex workloads through multithreading. ARM processors specialize in amazing performance with lower power draw and heat dissipation; since the vast majority of mobile devices are based on ARM, ARM processors also advertise themselves as being “cloud-native”. ARM CPUs generally offer a higher core count, better performance per watt, superb energy efficiency and heat dissipation, and lower TCO. They are a force to reckon with in the server processor market.

It goes without saying that you need a remarkable server product to fully benefit from these server processors. GIGABYTE Technology has server solutions that support Ampere® Altra® and Ampere® Altra® Max CPUs if you choose ARM processors, or Intel® Xeon® Scalable and AMD EPYC™ CPUs if you opt for x86 processors. Let’s take a look at their strengths and applications to help you make the optimal choice.

As mentioned, ARM processors are widely used in mobile devices such as smartphones and tablets because RISC revolves around simpler instruction cycles, which reduce the CPUs’ power draw and heat dissipation. When used inside servers, ARM CPUs can offer computing power on par with the best x86 processors. If you are looking for a supremely power-efficient server solution, or if you plan on working with ARM-based mobile or edge devices, GIGABYTE servers that run on Ampere® Altra® and Ampere® Altra® Max CPUs may be the answer you need.《Glossary: What is Edge Computing?》

GIGABYTE has been a pioneer in ARM server technology since 2013. One notable case study is the Graduate Institute of Networking and Multimedia at Taiwan University (NTU), which employed GIGABYTE’s G242-P32 server and the Arm HPC Developer Kit to develop a “high-precision traffic flow model” as part of an intelligent transportation system (ITS). Since the roadside devices used in the project are industrial PCs (IPCs) that run on ARM processors, the G242-P32 eliminates the “communication barrier” between the data center and the field devices, resulting in an efficiency boost of 200%. The single-socket Ampere® Altra® CPU in the G242-P32 has up to 80 cores in a single CPU, which makes it ideally suited for dealing with the massive amount of data the NTU team collected from the roadside sensors.

Learn More:
《GIGABYTE’s ARM Server Boosts Development of Smart Traffic Solution by 200%》
《Glossary: What is Intelligent Transportation System?》

Since x86 processors are based on CISC, they are capable of running complex instructions that can be carried out across multiple clock cycles. Therefore, even if the number of cores doesn’t quite compare to ARM processors, they are still capable of providing incredible performance by utilizing techniques such as multithreading. The scalability and compatibility of x86 CPUs are also nothing to sneeze at—there is a reason why x86 server processors still hold a lion’s share of the market, and many enterprises swear by their x86 CPUs.《Glossary: What is Scalability?》

Almost all of GIGABYTE’s server solutions: including the H-Series High Density Servers and G-Series GPU Servers for HPC, the general-purpose R-Series Rack Servers, the E-Series Edge Servers for edge computing, the S-Series Storage Servers for storage, and the W-Series Tower Servers / Workstations—they all support x86 processors. Many enterprises and institutions select x86 CPUs—whether it is Intel® Xeon® Scalable or AMD EPYC™—for their servers. For example, the University of Barcelona increased the capacity of their on-campus data center by building a computing cluster with GIGABYTE’s R182-Z90 and G292-Z42 servers. These servers are powered by AMD EPYC™ processors that have 64 cores and 128 threads per CPU—perfect for the data processing done inside a world-class university.

Learn More:
《The University of Barcelona Gets a Computing Boost with GIGABYTE Servers》
《Glossary: What is Computing Cluster?》

Another example is an Israeli developer of autonomous driving technology. To develop the self-driving algorithm, it needed to conduct repeated test drives in many different locations and pour an ocean of information into its database. It chose GIGABYTE’s G291-281 and R281-NO0 servers for their high availability and scalability features, as well as their Intel® Xeon® Scalable server processors, which are available in dual socket configurations. The CPUs can support up to 28 cores and 56 threads per socket. They can be optimized to support different applications, including enterprise databases, cloud computing, and challenging HPC workloads.

Learn More:
《Constructing the Brain of a Self-Driving Car with GIGABYTE Servers》
《Glossary: What is Cloud Computing?》

We hope this tech guide has been able to explain the basics of server processors, how they are different from PC processors, and what affects their performance. If you are looking for server solutions that support the most advanced server CPUs, GIGABYTE can help. We encourage you to reach out to our sales representatives at [email protected] for consultation.

Learn More:
《Read Another Tech Guide! What is Big Data, and How Can You Benefit from It?》
《Read Another Case Study! 0 to 100 KPH in 3.3 Seconds! NTHU Builds Electric Formula Student Race Cars with GIGABYTE》