GPU Instance

Definition

A cloud virtual machine equipped with one or more Graphics Processing Units (GPUs) for accelerated computing workloads, ideal for AI tasks.

Use Cases

• OpenAI: Large Language Model Training — Uses thousands of GPU instances with NVIDIA A100/H100 GPUs for training foundation models (Trains models with hundreds of billions of parameters in weeks rather than years)

Provider Equivalents

• AWS: P5, G5 Instance Families
• Azure: NC, ND Series VMs
• GCP: A2, G2 Machine Types
• OCI: GPU Shapes (BM.GPU)

Frequently Asked Questions

When should I use GPU instances vs CPU instances?

Use GPU instances for parallelizable workloads like ML training, video encoding, 3D rendering, and scientific simulations. CPU instances are better for general-purpose web applications and sequential processing.

How much do GPU instances cost?

GPU instances are significantly more expensive than CPU instances. For example, an AWS p5.48xlarge costs around $98/hour on-demand, but spot pricing and reserved instances can reduce costs by 60-90%.

Category: compute

Difficulty: intermediate

Related Terms

• GPU
• Machine Learning
• Virtual Machine
• High-Perf Computing
• AI Accelerator

See Also

• Compute Engine
• High-Perf Computing
• Bare Metal Server
• Web App Hosting
• Media Live Streaming