Slow Rendering Speed in DaVinci Resolve

Hello fellow forum members,

I'm currently encountering a persistent issue while rendering a feature film on DaVinci Resolve 18.6, utilizing a robust setup comprising an AMD Ryzen 9 5950X processor, 64GB RAM, and an NVIDIA GeForce 3080TI 12GB GPU running on Windows 11 Pro 22H2 with GeForce Game Ready Driver 537.42.

The film's timeline primarily consists of footage shot in RED RAW 4K-6K, with a few DJI ProRes 4444 clips. Initially, the rendering speed is normal at around 15-17 FPS. However, it drastically slows down to an unworkable 0-0.5 FPS when the GPU memory is nearly full (11.2GB/12GB) and RAM usage approaches its limit (40.7/64GB). This happens particularly after rendering light masking fusion clips in the timeline.

A similar issue occurred in the past with a previous GPU, the GeForce 3070 8GB, while attempting to render a shorter 12-minute film comprised entirely of RED RAW 6K clips with denoise nodes.

Interestingly, when I attempted to render the same content on MacBooks, including an M1 with 8GB and an M1 Pro with 16GB, the rendering proceeded at a normal and satisfactory speed, indicating a potential anomaly with my PC setup.

I would greatly appreciate any insights, advice, or troubleshooting steps to resolve this issue and optimize my rendering performance on DaVinci Resolve 18.6.

Thank you in advance for your assistance.

[Peter Chamberlain]

Could be GPU driver, but until u work that our, render the troublesome clips to and intermediate format using render in place and then try the full render

Does it always slow down in the same place?

Could be GPU driver, but until u work that our, render the troublesome clips to and intermediate format using render in place and then try the full render

Thanks for providing the solution. I made some pre-renders for fusion clips to avoid these problem. It worked fine. The render gets slow to 0-0.5 FPS only when the GPU memory is full.

Does it always slow down in the same place?

Not really. It slows down to 2-5fps at fusion clips in the timeline, which is normal. But it gets lower to 0-0.5fps at random non-vfx clips whenever the GPU memory gets full. The performance of m1 pro on mac is stable 5-10FPS for same timeline/project.

I thought it was just me who had a problem with the new update. It's too slow that I can't do anything.

Update: I came up a solution is to choose "native" encoder instead of "auto" when deliver quicktime/mp4 files.

You can take a lot of load from the GPU by rendering into DNxHR or CineForm and doing final compression separately.
Current Macs don't have that problem with their unified memory and hardware encoders.