Learn how to install NVIDIA drivers and CUDA Toolkit in a Windows environment.
Installation target
| software | purpose |
|---|---|
| NVIDIA GPU Driver | Basic drivers to make your GPU work properly on Windows |
| CUDA Toolkit | Development tools required for GPU-based development (e.g., AI/deep learning) (nvcce.g., , cuBLAScuDNN , etc.) |
Installation Instructions (Windows 10/11)
1. Check if your NVIDIA GPU is recognized correctly
Win + X→ Device Manager- Check if the “Display Adapter”
NVIDIA GeForce RTX XXXXis displayed.
2. Install official NVIDIA drivers
NVIDIA Driver Download Page
https://www.nvidia.co.jp/Download/index.aspx?lang=jp
Procedure:
- Select your GPU model number from the link above (e.g., GeForce RTX 5080)
- Choose a Windows OS version (e.g. 64-bit)
- Download the latest Game Ready Driver or Studio Driver
.exeRun the file and follow the instructions of the installer- restart
Search example)
Comparison summary
| characteristic | Game Ready Driver | Studio Driver |
|---|---|---|
| Main applications | Gameplay/Development | Creative work (video editing, modeling, AI, etc.) |
| Update Frequency | High (to match the game release) | Low (stability-oriented) |
| stability | Slightly lower | Very high |
| Base Driver Configuration | identity | identity |
| Eligibility | Gamers, game developers | Video Producers, Researchers, AI Engineers, etc. |
Which one should you choose?
- 🎮 Game-Centric → Game Ready Driver
- 🎨 Video editing, AI development, 3D production, etc. → Studio Driver
- If you want to do both → If you value stability, you can use Studio Driver (GRD can also be used Creative)
*Dual use is not possible
・Only one driver can be installed at a time.
・If you want to switch as needed, a clean install is recommended.
3. Installing CUDA Toolkit
CUDA Toolkit Download (Formula)
https://developer.nvidia.com/cuda-downloads
Installation Instructions:
- Select the OS as “Windows” → version (e.g., 10, 11)
- “exe (local)” recommended (works offline)
- After running the installer, select “Fast” or “Custom” to choose the features you want.
- After the installation is completed, the environment variables are automatically set (
nvcc --versioncheck here)
Example of selection)
Confirmation of operation
NVIDIA Driver Check:
nvidia-smiIf you see the GPU model number, memory, and driver version, you are successful.
CUDA Confirmation:
Command Prompt or restart PowerShell to run the command.
nvcc --versionIf the CUDA version (e.g. 12.9, etc.) is printed, it is fine.
Example of execution results)
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2025 NVIDIA Corporation
Built on Tue_May_27_02:24:01_Pacific_Daylight_Time_2025
Cuda compilation tools, release 12.9, V12.9.86
Build cuda_12.9.r12.9/compiler.36037853_0
*“Only one CUDA version enabled by the current environment variable (PATH)” is displayed. You cannot see a list of multiple versions.
*To use multiple nvcc at the same time, you need to respond to “specify a separate full path” or “temporarily switch environment variables“.
Checking CUDA Environment Variables
Windows search window> system prompts> advanced settings> environment variables>CUDA_PATH
Reference: Official reference
- CUDA Toolkit Documentation – NVIDIA Developer
https://docs.nvidia.com/cuda/index.html - CUDA Installation Guide for Windows
https://docs.nvidia.com/cuda/cuda-installation-guide-microsoft-windows/index.html - NVIDIA official nvidia-smi documentation
https://docs.nvidia.com/deploy/nvidia-smi/index.html
Comments