| | Purpose | Access | |----------|-------------|-------------| | NVIDIA Mods (Modular Diagnostic Tool) | Internal board-level diagnostics for manufacturing/RMA | NVIDIA partners/OEMs only | | DCGM (Data Center GPU Manager) | Data center GPU health & diagnostics | Publicly available | | NVIDIA-SMI | GPU status, memory, temperature checks | Included with NVIDIA drivers | | NVIDIA PVA Diagnostic Tool | Programmable Vision Accelerator tests | Limited access | | CUDA Samples deviceQuery | Basic GPU functionality test | Public (CUDA Toolkit) |
: Primarily used to test the GPU chip itself, validating its logic and functionality. nvidia modular diagnostic software download
If you are experiencing issues with your NVIDIA graphics card, such as: clocks.current.memory --format=csv >
NVIDIA Modular Diagnostic Software is a useful tool for diagnosing and troubleshooting issues with NVIDIA graphics cards. By downloading and using this software, you can identify and potentially resolve issues with your graphics card, ensuring optimal performance and stability. ensuring optimal performance and stability.
# Comprehensive NVIDIA-SMI report nvidia-smi --query-gpu=index,name,serial,uuid,power.limit,power.draw,temperature.gpu,memory.total,memory.used,utilization.gpu,clocks.current.graphics,clocks.current.memory --format=csv > nvidia_gpu_report.csv
# After installing CUDA Toolkit cd /usr/local/cuda/samples/1_Utilities/deviceQuery make ./deviceQuery