Problem
DaVinci Resolve playback drops frames because decode and GPU processing compete for limited VRAM.
Solution
Root Cause / Diagnostic:
4K/8K H.264/HEVC decoding and advanced DaVinci Resolve GPU color grading nodes (such as Temporal Noise Reduction, Face Refinement, and Optical Flow) share the same dedicated VRAM pool. When high-resolution inter-frame decoding saturates onboard VRAM, GPU memory allocations fail. Resolve drops playback frames, freezes CUDA/Metal processing, and displays 'GPU Memory Full' warnings.
Actionable Fix:
1. Lower timeline playback proxy resolution to Half or Quarter resolution via Playback -> Timeline Proxy Resolution.
2. Switch source media decoding to generated Optimized Media or Half-Res Proxies (ProRes Proxy or DNxHR LB) to reduce VRAM decode footprint.
3. Node-cache intensive GPU nodes: right-click heavy temporal NR nodes and select 'Node Cache -> On' to pre-bake results to disk cache.
Pro Tip:
Configure DaVinci Resolve Memory and GPU preferences to allocate 75% of system RAM to Resolve and set GPU Configuration to 'Auto' or select high-performance discrete GPUs explicitly.
4K/8K H.264/HEVC decoding and advanced DaVinci Resolve GPU color grading nodes (such as Temporal Noise Reduction, Face Refinement, and Optical Flow) share the same dedicated VRAM pool. When high-resolution inter-frame decoding saturates onboard VRAM, GPU memory allocations fail. Resolve drops playback frames, freezes CUDA/Metal processing, and displays 'GPU Memory Full' warnings.
Actionable Fix:
1. Lower timeline playback proxy resolution to Half or Quarter resolution via Playback -> Timeline Proxy Resolution.
2. Switch source media decoding to generated Optimized Media or Half-Res Proxies (ProRes Proxy or DNxHR LB) to reduce VRAM decode footprint.
3. Node-cache intensive GPU nodes: right-click heavy temporal NR nodes and select 'Node Cache -> On' to pre-bake results to disk cache.
Pro Tip:
Configure DaVinci Resolve Memory and GPU preferences to allocate 75% of system RAM to Resolve and set GPU Configuration to 'Auto' or select high-performance discrete GPUs explicitly.