In the past, x86 processors were the go-to choice for high performance solutions. However, with the rise of smartphones and tablets, it became apparent that x86 processors were not suitable for these devices. ARM processors, on the other hand, proved to be a better fit due to their performance per watt, cheaper design and deployment costs, and the ability to add IP cores like modems and graphics accelerators.



Large companies like Qualcomm and Apple have even developed their own custom ARM processor cores for high-performance systems. And while Microsoft's failure in the smartphone and tablet market opened the door for ARM processors to take over, Intel attempted to enter the market with Android OS but was unsuccessful due to the risks associated with using x86 processors in smartphones, the higher power consumption, and the final price for the consumer.

As ARM processors continue to dominate the mobile market, there is increasing talk about whether they will eventually replace x86 processors in data centers. With ARM's better performance per watt and lower design and deployment costs, it could potentially dethrone x86 as the go-to choice for high performance solutions. However, only time will tell whether this paradigm shift will happen.

Using ARM processors instead of x86 can result in significant cost savings. The initial infrastructure construction costs are reduced by 30-60% due to the compact form factor of ARM modules, their lower heat dissipation, and increased density per server. Additionally, ARM-based hardware is 20% cheaper and operating costs are reduced by 10-15%. Switching to a cloud infrastructure running ARM processors can result in up to 80% savings.

ARM processors are also better suited for specialized enterprise applications because it's possible to move software logic to specialized coprocessors, which work in tandem with the main ARM core. This approach can provide better results than optimizing algorithms on a general-purpose processor like x86. Amazon has developed its own AWS Graviton processors, which are based on original 64-bit ARM Neoverse cores and provide optimal value for money for Amazon EC2 cloud loads.

AWS Graviton2 processors are characterized by higher performance and advanced capabilities compared to the first-generation AWS Graviton processors. They are used for a wide variety of workloads, including startup application servers, microservices, high-performance computing, gaming applications, open-source databases, and in-memory caching. Snap Inc. and Netflix are just two examples of companies that have benefited from the transition to ARM-based AWS Graviton2 instances. Overall, as more companies look to optimize their IT infrastructure and reduce costs, we may continue to see a shift towards ARM processors over x86.

The adoption of ARM servers for enterprise applications may seem unclear at the moment. There are no clear gains compared to traditional x86 servers and independent benchmarks are not showing any significant improvement yet. Although we have heard claims that "all servers will be on ARM," this has not yet materialized, unlike the success Apple had with M1 processors in their devices, which cannot be directly compared to servers with ARM processors.

However, if ARM processors become more profitable for enterprises, it's possible that these servers will replace traditional x86 servers within the next 5-15 years. Enterprises aren't necessarily concerned with what microservices or standard applications they run on, as long as the choice is financially beneficial. It's important to note that technology evolves quickly, so it's difficult to predict the exact future of server adoption.

Since the release of ThunderX in 2013, there have been many claims about the growth of ARM servers, but the reality is that few options exist. While AMD has made a comeback with EPIC-based servers from various manufacturers, ARM servers are still scarce. It's also difficult to find a reliable workstation for ARM, which is necessary for debugging and software porting. Huawei offers servers with their ARM processor, which perform well and have good networking capabilities, but vendor lock and unclear prospects may hinder their success.

In order for ARM to gain more traction in the server segment, it would require significant support from a large hardware company, like NVIDIA, who could develop a line of CPU + chipset ready for OEM. Without this kind of support, it's doubtful that ARM will see widespread adoption in the server market. However, technology evolves quickly, so it's possible that we may see major advancements in the future that could change the landscape of server infrastructure.

Recent setbacks for ARM include the closure of the ThunderX3 project at Marvell and the discontinuation of the server direction in Qualcomm Nuvia. In reality, ARM's ambitions to compete with x86 in data centers have faced significant obstacles in terms of energy efficiency, scalability, and compatibility. More specifically, ARM processors can perform similarly to x86 processors but require a higher TDP, not all loads can be distributed across weaker cores, and they have poor performance in FP, among other issues. Additionally, software support for ARM is still unfinished.

Despite some enthusiastic advocates like John Masters, there hasn't been much progress for ARM servers as compared to x86. However, the technology industry is constantly evolving, and it's possible that advancements in ARM processors could address these issues in the future.