The Advantages of ARM: From Smartphones to Supercomputers and Beyond
Processors based on the ARM architecture, an alternative to the mainstream x86 architecture, is gradually making the leap from mobile devices to servers and data centers. In this Tech Guide, GIGABYTE Technology, an industry leader in high-performance server solutions, recounts how ARM was developed. We also explain the various benefits of ARM processors and recommend ARM servers for different sectors and applications.
ARM processors, which are used in just about every smartphone on the planet, have become a force to reckon with in the field of server solutions. For instance, Fugaku, one of the world’s fastest supercomputers, runs on processors based on the ARM architecture. If you are hearing about this decades-old technology for the first time, now is a good chance to learn about ARM.
Rewind to December 1st, 1981. The British Broadcasting Corporation launched the BBC Micro, a microcomputer (bet you didn’t know the BBC made computers!) designed and built by a company called Acorn Computers. Acorn would go on to launch the ARM1 processor in 1985. At the time, the acronym stood for Acorn RISC Machine; this would later be changed to Advanced RISC Machines. The nature of RISC meant that the ARM1 and its successors, colloquially known as the ARM processors, could offer better energy efficiency and total cost of ownership (TCO). Fast forward a decade or two, and it should come as no surprise that ARM has made the leap from microcomputers to mobile devices.《Glossary: What is RISC?》
In the last decade or so, high performance computing (HPC) and cloud-based data centers have experimented with ARM processors. In 2014, Japan’s Institute of Physical and Chemical Research (RIKEN) teamed up with Fujitsu to develop the successor to the “K computer”, their previous supercomputer. The new supercomputer, named “Fugaku”, was built with the ARM-based Fujitsu A64FX microprocessor. In June of 2020, the “TOP500” ranked Fugaku as the world’s fastest supercomputer. This shows that ARM has a definite role to play in the advancement of cloud computing, HPC, and other technological breakthroughs.
GIGABYTE Technology, a leading brand of high-performance servers, has years of experience in the development and manufacturing of servers. GIGABYTE has been a pioneer in ARM server technology since 2013. GIGABYTE works with Ampere Computing, a leading provider of ARM-based server processors, to incorporate Altra® and Altra® Max CPUs in GIGABYTE’s server solutions. ARM processors offer more cores (which translate to better performance), higher energy efficiency (which means fewer issues with heat dissipation), and better TCO. Because they are widely used in mobile devices, ARM processors are also “cloud-native”, which makes them suitable for cloud computing, edge computing, and AI implementation.
In the following sections, we will go over the development of ARM and delve into its many benefits, so you can decide if GIGABYTE’s ARM servers are right for you.
ARM: The Predominant Type of RISC Processors
Did you know? Computers talk to each other in their own language, like humans do.
Currently, the mainstream processor architecture is called “x86”; this is what you will find in a majority of personal computers. Architecture is shorthand for “instruction set architecture” (ISA). This is the abstract model used by computers to determine how everything is coded, and how commands are carried out. Processors designed by different companies may adopt the same ISA; for example, both AMD and Intel CPUs are based on the x86 architecture, which means any operating system or program that can run on an Intel CPU can also run on an AMD CPU.
The x86 architecture follows what is called a “complex instruction set computer” (CISC) design. Conversely, the “R” in ARM literally stood for RISC—“reduced instruction set computer”. Therein lies the fundamental difference between x86 and ARM processors. 《Glossary: What is CISC?》
From the 1950s until the 1970s, as computers became more popular, people envisioned a powerful ISA that could understand and execute a series of operations. This made it less time-consuming for programmers to develop software, and it lightened the burden on the limited memory capacity of the time. As the years went by, memory capacity began to catch up, thanks to advancements in the semiconductor manufacturing process. Compilers that could translate computer code from one programming language to another also appeared on the market. This paved the way for alternative CPU designs—ones that consumed less power and were cheaper to make.
This was the background against which Acorn created ARM, a processor that “speaks” the simpler, more energy-efficient RISC “language”. Other RISC processors have come and gone, but ARM has proven to be the most enduring. In the 36 years between 1985 and 2021, 200 billion ARM-based chips were shipped around the world, making ARM the most popular type of computer chip in human history.
If you pay attention to world news, you will know that the ARM saga is still ongoing. In 2016, the Japanese conglomerate SoftBank Group purchased ARM Holdings (later renamed Arm Ltd.), the spinoff of the original Acorn Computers and the primary designer of ARM processors. In September of 2020, Nvidia Corporation announced its intention to acquire Arm Ltd. for US$40 billion. The idea of a GPU giant taking over a primary mover of CPU technology sent government regulators into a frenzy. As of the writing of this article, it has been announced that NVIDIA’s proposed acquisition has fallen through, and there are plans for Arm Ltd. to go public.《Glossary: What is GPU?》
The Unique Benefits of ARM
Out of all the different kinds of RISC processors, what makes ARM so special? What are the unique qualities that have made it not only the most popular RISC processor, but the most widely used computer chip of any kind? We’ve narrowed it down to four reasons:
● Higher core count, better performance per watt
While the maximum number of cores in each CPU is not wholly dependent upon the architecture, mainstream x86 CPUs tend to house a smaller number of larger, more complex cores. ARM CPUs compete by having more cores inside a single processor; for example, the Ampere® Altra® Max series can contain up to 128 cores in each CPU. The workload is distributed to a large number of smaller, more efficient cores, rather than having a few powerful cores handle every task. As a result, the cores of an ARM processor can offer better performance per watt of power.
● Superb energy efficiency, better heat dissipation
RISC processors are generally more energy-efficient than their CISC counterparts, because they use simple instructions that can be executed within one clock cycle. Not only do these “reduced instructions” require fewer transistors, pipelining between different sets of instructions is also possible. Because of this, ARM processors are incredibly energy-efficient and generate less heat. This is crucial for mobile devices such as smartphones, since battery life cannot be too limited, and the device must not get too hot! These attributes are also beneficial for server products, since optimal heat dissipation can ensure higher performance and stability.
These four benefits are the reasons why ARM processors have become the most popular kind of computer chip on the planet. Originally, it was the mobile devices, such as smartphones and tablets, which flourished because of these advantages. What we are learning now is that servers can also benefit greatly from ARM processors.
● Lower TCO, better performance per dollar
ARM processors are generally cheaper to design and manufacture. Dollar for dollar, you are getting better performance when you use ARM, which is why they played a pivotal role in the proliferation of inexpensive mobile devices. What’s more, the fact that ARM processors are more energy-efficient and easier to keep cool also goes a long way towards reducing the overall TCO of ARM-based devices. Especially in a server room or server farm, where electricity bills can really eat into the bottom line, switching to ARM servers can help keep costs low.
In recent years, ARM advocates have put a lot of focus on this selling point. ARM processors are inherently cloud-native, because they are the predominant type of computer chip used in mobile and edge computing. As more and more devices (whether it’s a smartphone or self-driving car) connect to the cloud, employing the same architecture in data centers could streamline the process. It may also reduce latency when the devices “talk” to one another in the same “language”, so to speak. In other words, while the x86 architecture has largely failed to make the leap from PCs and servers to mobile devices, ARM is poised to stage a counterattack and gain a significant foothold in data centers.
ARM and x86: Competition or Coexistence?
Before we move on to GIGABYTE’s line of ARM servers, it’s worth devoting a few paragraphs to the oft-debated topic of “x86 vs. ARM”. As we’ve mentioned, x86 servers are still ubiquitous in today’s data centers and IT infrastructure, thanks in part to Intel and AMD’s long years of investing in the PC market. Since x86 processors are based on the CISC design, they generally offer blazing-fast response time, and they excel at tackling complex workloads through something called multithread processing. What’s more, since x86 has been the go-to-choice for IT managers for decades, it has a very complete ecosystem, with software and tools for just about every application.
ARM servers have a lot of potential in the newer realms of cloud and edge computing. In fields of application where the number of cores matters greatly, or where power efficiency and scalability are key, ARM processors have a definite role to play. This includes AI and deep learning applications, as well as 5G. Although ARM’s ecosystem is less complete than that of the x86 architecture, it is catching up quickly. In fact, in new fields of application where there is no existing ecosystem, ARM’s shortcoming in this regard may be a moot point.
Choose GIGABYTE to Bring the Wonders of ARM to Different Sectors
GIGABYTE Technology, a leading provider of server products that are used in a wide variety of sectors, began developing ARM-based solutions as far back as 2013. This was hot on the heels of the launch of AARCH64 (or ARM64), the 64-bit extension of the ARM architecture that offers even better performance than traditional 32-bit processors. Over the years, GIGABYTE has built a strong partnership with many advocates and pioneers of ARM technology. One of the crowning moments came in 2019, when GIGABYTE built the Customer Reference Board (CRB) for Ampere Computing’s Mt. Snow platform, a single-socket rack server powered by a single Ampere® Altra® processor containing eighty 64-bit CPU cores. Through the years, GIGABYTE has developed a common design of ARM-based servers, which means GIGABYTE can quickly incorporate a different configuration of components and create new server products for specific applications.
In the last year alone, GIGABYTE has launched over a dozen ARM servers for general-purpose use, GPU-centric workloads, HPC clusters and HCI, and edge computing. If you are intrigued by the benefits of ARM and want to work with a server solution that offers high performance, low TCO, superb energy efficiency, and cloud nativity, you need only to consider the type of computing tasks you want your new ARM-based solution to handle, and pick from GIGABYTE’s complete line of ARM servers.《Glossary: What is HCI?》
Whether you want to use ARM processors for edge computing, high-density computing, GPU-centric workloads, or general purpose data center applications, GIGABYTE has the matching server solutions. These products combine the benefits of ARM processors with GIGABYTE’s decades of experience in the development of server technology.
● R-Series Rack Servers for General Purposes
GIGABYTE’s R-Series Rack Servers are the go-to-choice if you want an optimal balance between performance, reliability, and versatility. By combining this tried-and-true general-purpose design with the prowess of ARM processors, you can benefit from an ARM server that is suitable for nearly every situation. GIGABYTE offers the R152-P30, R152-P31, and R152-P32 for customers who require a server that fits a 1U (one rack unit) form factor; and the R272-P30, R272-P31, R272-P32, and R272-P33 for the 2U form factor. These rack servers support a single socket Ampere® Altra® processor with sixteen DIMMs, two Gen4 M.2 slots, dual 1GbE LAN, a dedicated management port, and PCIe Gen4 expansion slots.
For special tasks optimized for distributed computing, the addition of GPGPU accelerators can serve as an incredible boost to computing power. GIGABYTE’s G-Series GPU Servers are designed for this kind of workloads; specifically, the G242-P31, G242-P32, G242-P33, and G242-P34 are a good choice if you want to achieve synergy between CPUs and GPUs through heterogeneous computing. To use the G242-P31 as an example, its 2U chassis can house up to four NVIDIA ® A100 PCIe Gen4 GPU cards and dual 80+ Platinum 1600W (240V) power supplies, in addition to a single socket Ampere® Altra® processor. It should also be noted that the G242-P32 is the server used in NVIDIA’s Arm HPC Developer Kit total solution.
● H-Series High Density Servers for HPC Clusters and HCI
The H262-P60 is GIGABYTE’s high-density Arm server. It features four nodes, which means it can support a total of eight processors in a 2U chassis. It offers eight DIMMs per socket for the 8-channel memory mode, and it supports the latest DDR4-3200. Inside each node, there are six SATA drives, a single M.2 slot, two half-length, half-height slots, an OCP 3.0 slot, dual 1GbE LAN ports, and a dedicated management port. There is also a CMC module to aggregate the information from each of the four nodes’ BMC for chassis-level management and multi-node monitoring. This allows remote management and control of a single server or a cluster of servers. The H262-P60 also supports the Ampere® Altra® Max processor, which can contain up to an astounding 128 cores per CPU.
GIGABYTE E252-P30 and E252-P31 bring the benefits of ARM to the edge by installing the processors in an even denser and more compact 2U form factor. The chassis of these servers have been trimmed down to 439 x 86 x 449mm (WxHxD); in other words, the depth has been decreased from the standard size of over 660mm to just 449mm, while maintaining the structural integrity and thermal control of a top-quality server. A front-access server chassis has been incorporated for easy access and maintenance in confined spaces, and a tool-less rail kit design enables convenient installation and maintenance.
GIGABYTE’s comprehensive line of ARM servers powered by Ampere® Altra® and Ampere® Altra® Max processors is a testament to GIGABYTE’s commitment to developing the most advanced and innovative server solutions that will “Upgrade Your Life”. We hope this tech guide has been able to explain the origin of ARM, its many benefits, and how you can use it in your servers. If you are looking to incorporate ARM in your data center or server room, GIGABYTE can help. We encourage you to reach out to our sales representatives at server.grp@GIGABYTE.com for consultation.