Compute Services
Amazon EC2 (Elastic Compute Cloud)
What it is:
Amazon EC2 provides resizable virtual machines (instances) in the cloud. You can choose from a wide range of instance types optimized for compute, memory, storage, or GPU-based workloads.
Why it matters:
- Offers full control over compute resources
- Supports AI/ML training and inference using GPU instances
- Scales from small experiments to high-performance distributed training
Typical Use Cases:
- Running deep learning frameworks like TensorFlow or PyTorch on GPU instances
- Hosting custom-trained ML models for inference
- Performing large-scale simulations or model training
- AWS Trainium Instances (Trn1)
- Accelerated Computing P Type Instances
- Accelerated Computing G Type Instances
- Compute Optimized C Type Instances
AWS Trainium instances use a custom-designed machine learning chip engineered for high performance with low power consumption, reducing the carbon footprint of training large-scale models.
Key Features:
- Up to 25% more energy efficient than comparable accelerated computing EC2 instances.
- Specifically designed for optimal performance per watt for deep learning workloads.
- Lowers environmental impact compared to other instance types.
Typical Use Cases:
- Large-scale deep learning training.
- Organizations prioritizing sustainability and energy efficiency in AI workloads.
Why it matters:
They are the most environmentally friendly choice, helping companies meet sustainability goals while training complex models.
Accelerated Computing P type instances are powered by high-end GPUs like NVIDIA Tesla and are optimized for maximum computational throughput.
Key Features:
- Delivers high GPU performance for ML and HPC tasks.
- Not designed with energy efficiency as the primary goal.
- Consumes significant power.
Typical Use Cases:
- Heavy ML model training and inference.
- High-performance computing (HPC) workloads.
Why it matters:
Best when raw GPU power is needed — less suitable for energy-conscious workloads.
Accelerated Computing G type instances use NVIDIA GPUs for graphics-heavy applications like gaming, rendering, and video processing.
Key Features:
- High computational power for visual workloads.
- Not optimized for ML training or energy efficiency.
Typical Use Cases:
- Real-time rendering, video processing, game streaming.
- Graphics-intensive applications.
Why it matters:
Excellent for graphics tasks but not the best choice for minimizing environmental impact.
Compute Optimized C type instances provide high CPU performance for compute-intensive applications.
Key Features:
- Maximizes raw compute power for CPU-bound workloads.
- Not specifically designed for energy efficiency like Trainium.
- Suitable for high-throughput applications.
Typical Use Cases:
- Web servers, gaming backends, scientific modeling.
- Applications needing maximum CPU power.
Why it matters:
Ideal for compute-heavy tasks but less ideal for organizations focused on lowering carbon footprint.