MI300X GPU Status Report | Generated by AI
Here’s the quick GPU status:
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
MI300X GPU Status — rocm-smi
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
GPU Utilization: 100% ← fully saturated
VRAM Used: 63.5 GB / 192 GB (33%)
Power: 747W / 750W cap ← running at limit
Junction Temp: 75°C
Memory Temp: 50°C
SCLK (GPU): 1351 MHz
MCLK (Memory): 900 MHz
Fan: 0% (passive/server cooling)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Key commands for future reference:
rocm-smi # quick overview (like nvidia-smi) rocm-smi -a # all details rocm-smi –showuse –showmeminfo vram # GPU% + VRAM rocm-smi –showtemp –showpower # thermal + power rocm-smi –showclocks # clock speeds watch -n1 rocm-smi # live monitoring
GPU is at 100% utilization, 747W power draw (near the 750W cap), junction temp 75°C — all normal for a sustained training workload. VRAM is only 33% used, so you have plenty of room to scale batch size or model size if needed.