Skip to main content

Evaluate and adopt energy-efficient processor architectures

Description​

Applications are built with a software architecture that best fits the business need they are serving. Cloud providers make it easy to evaluate different CPU architectures, including alternatives to traditional x86-64 processors. Cloud virtual machines come with different capabilities based on their underlying hardware processors. Using virtual machines based on processor energy efficiency impacts hardware efficiency and reduces carbon emissions.

Solution​

Evaluate and adopt processor architectures with energy efficiency and execution performance in mind:

Evaluate Alternative CPU Architectures​

Consider CPU architectures beyond x86-64, evaluating them for:

  • Energy efficiency: Performance per watt for your specific workload
  • Execution performance: Actual application performance on the architecture
  • Cost effectiveness: Price-to-performance ratio
  • Workload suitability: Whether the architecture is well-suited to your application's computational patterns

Cloud-Native Processors (ARM-based)​

Modern cloud providers offer virtual machines with cloud-native processors engineered specifically for efficient, scale-out cloud workloads:

ARM-based instances provide significant energy efficiency advantages:

These processors are designed for cloud-native workloads and typically offer:

  • Better energy efficiency per core
  • Cost effectiveness (lower pricing for similar or better performance)
  • High core counts for parallel workloads

Specialized Accelerators​

Also consider specialized accelerators that cloud providers offer for specific workload types:

  • GPUs for parallel processing and machine learning
  • FPGAs for customizable hardware acceleration
  • ASICs for domain-specific tasks (e.g., AI/ML inference)

SCI Impact​

SCI = (E * I) + M per R Software Carbon Intensity Spec

Evaluating and adopting energy-efficient processor architectures impacts SCI as follows:

  • E: Energy-efficient processors reduce total energy consumption for the same computational work, directly lowering operational carbon emissions
  • M: By using processors with better performance-per-watt characteristics, fewer physical resources may be required, reducing embodied carbon

Assumptions​

  • The application framework can be executed on alternative CPU architectures and is optimized for them
  • The cloud provider offers VM SKUs based on efficient hardware in your deployment regions

Considerations​

  • Energy-efficient VMs tend to be cost-effective as well, offering both environmental and financial benefits
  • Consider SKU availability in your chosen deployment region, as not all instance types are available in all regions
  • Ensure your application and dependencies are compatible with alternative architectures (particularly ARM), including:
    • Runtime environments (Node.js, Python, Java, etc.)
    • Container images and base images
    • Third-party libraries and native dependencies
  • Performance testing is essential to validate that the architecture meets your requirements
  • Some workloads benefit significantly from ARM processors (web servers, microservices, containerized apps), while others may need x86-64 compatibility

References​