> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/sdgm/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/sdgm/_mcp/server.

# How do I size compute and storage for my graph?

> Estimate source data size and choose CPU, RAM, GPU, and scratch capacity

Use the total size of the source Parquet files for an initial estimate, including every table in the graph. Warehouse-reported size, exported Parquet size, and materialized index size can differ.

The [GPU Sizing guide](/kumo-engine/get-started-with-kumo-engine/gpu-sizing) covers CPU, RAM, GPU, and local SSD scratch capacity for Python workflows. GPU memory also depends on the model, batch size, sampled neighborhoods, encoders, and task.

Measure peak memory, scratch usage, and runtime on representative data before scaling up. For the prerelease Kumo Studio Helm deployment, use its separate [sizing guide](/kumo-engine/kumo-studio-guide-coming-soon/preliminary-documentation/installation/helm-kubernetes/gpu-sizing).