Evolution of AWS EC2 General-Purpose: A 13-Year Performance and Price Analysis

Over the past decade, cloud computing has undergone significant transformations, with advancements in technology leading to improved performance, increased efficiency, and reduced costs. Amazon Web Services (AWS) EC2 instances have played a crucial role in this evolution, offering a range of general-purpose configurations tailored to various workloads.

Starting from the initial m3 series launched in 2012, AWS has regularly updated its instance offerings to cater to changing user demands. Today, users can choose from high-performance options like the latest m8g series (launched in 2025). This study takes a deep dive into the performance and pricing evolution of these general-purpose EC2 instances, spanning over 13 years. By comparing the capabilities and costs of various instance types, we aim to provide insights into the cost-effectiveness and scalability options available on AWS, helping users make informed decisions for their cloud deployments.

Pre-Reading Knowledge

To fully understand and appreciate the results presented in this study, readers should have a basic some understanding of the EC2 VMs.

Our benchmark compare 5 generation of EC2 instances with all the additionnal/customized sizes. All with the same base specifications of VM: 8vCPUs and 32GB of RAM. The observed differences will reflect the evolution of hardware and its capacity.

To quickly categorize VM flavors, it's handy to know the meaning of each symbol:

• i: Intel powered virtual machines
• a: AMD powered virtual machines
• g: Graviton powered virtual machines
• d: Extra local disks attached to VMs
• n: Extended network bandwidth compared to non-suffixed
• z: Specify high frequency processor. Only used in with m5zn
• flex: Ideal for workloads that not always fill the system

It’s also crucial to understand that all the tested virtual machines share the same operating system image, which means the values don’t accurately represent the performance of the 2010 kernel.