NVIDIA Windows on ARM Porting Guide

To reach new levels of productivity, integrated gaming, and AI performance, the NVIDIA Windows on ARM platform combines high-performance, efficient ARM cores with a powerful integrated RTX GPU, all with unified memory.

Many existing x86 and x86_64 applications can run using the Microsoft Prism Emulator. Compatibility depends on the application and its dependencies, so validate target configurations. To extract native performance, optimize power efficiency, and use the integrated platform, refer to this porting guide as a starting point.

• System Overview
  • Key Specifications
• RTX Spark GPU Identification
  • PCIe Device Identification
• Porting and Emulation on RTX Spark
  • CPU-Specific Details
  • Windows on ARM on RTX Spark
  • The Differences Between ARM64EC, ARM64, and ARM64X
  • Porting from x86/x86_64 to ARM64EC or Pure Native ARM64
  • Emulation Tips, Do’s and Don’ts
  • Memory Ordering and Other Porting Issues
  • Compiler Flags Guidance for RTX Spark
• Power
  • Battery Mode Recommendations for Gaming and AI Workloads
  • Querying Battery Mode (System-Level)
  • Querying GPU Power State (Fine-Tuning)
  • Power State Definitions from NVML (GPU Power State)
  • Developer Recommendations
• Unified Memory Architecture
  • Usage with CUDA APIs
• Porting Existing Apps
  • Setup and Build Systems
  • Vulkan and OpenGL
  • NVAPI
  • CPU Instruction Extensions