Maintenance and Troubleshooting

The DGX Spark benefits from regular software maintenance and proactive monitoring to ensure reliability, security, and peak performance. This section introduces recommended update workflows, health verification practices, and performance tuning guidance, and links to procedures that help keep your environment stable and efficient.

Maintenance and Troubleshooting

• System Updates
  • Overview
  • Preparation
  • Update Procedures
  • Validation
  • Troubleshooting
• Health Check
  • Overview
  • Preflight
  • GPU Health
  • System Metrics
  • Reporting
• Performance Tuning
  • Overview
  • CPU and Memory
  • GPU Optimization
  • Storage and I/O
  • Spark/RAPIDS Tuning
  • Validation and Profiling
• Known Issues
  • Use the supplied power adapter for optimal performance
  • nvidia-smi reports “Memory-Usage: Not Supported”
  • CUDA Support on DGX Spark
  • Guidance for reporting memory resources with unified memory architecture
  • Attached HDMI display goes into deep sleep mode after an extended period of inactivity and does not wake

Getting Support

If you encounter issues that are not covered in the troubleshooting sections above, or if you need additional assistance with your DGX Spark system, see Get the Right Support for Your DGX Spark for how to choose consumer or enterprise support, access the community forums, and understand entitlement requirements.

Before contacting support, please ensure you have:

• Reviewed the relevant troubleshooting sections in this guide

• Checked the Known Issues for any known issues that may apply to your situation

• Collected relevant system information and error messages

• Attempted the recommended troubleshooting steps for your specific issue