Private BETA Build Feedback (GPU Decode only)

Comments

3 comments

  • Avatar
    Steven Yaeger

    The "System Cached Frames Memory Limit" is for OpenCL and software image pipeline processing.  

    Previously "System Cached Frames Memory Limit" was also used for the CUDA image pipeline.  But when GPU R3D decoding was added we also moved the playback frame cache for CUDA GPU rendered frames from system memory to GPU memory.  Thus the "System Cached Frames Memory Limit" no longer effects frames rendered for viewing by the CUDA image pipelines.

    In the beta test version the GPU Cached Frames slider replaces the "System Cached Frames Memory Limit" slider for frames processed through the CUDA image pipelines.

     

  • Avatar
    rand thompson

    Thanks for the explanation Steven! I thought I wasn't doing something correctly. I think the limitation for my Laptop's playback performance is it's Nvidia Geforce GTX 980M(Mobile) GPU. Plus "8" seems to be my limit for the 980M

  • Avatar
    rand thompson

    I played back the project full of different 8K footage with different Frame Rates and Compressions from the post above. This time I played it back without using the Windows Game DVR, which records the screen. It seems Windows Game DVR was utilizing some of the GPU. Now without using the Windows Game DVR, instead of 11 to 13 Frames Per Second for a 23.98 clip, I'm getting 21 to 23.98 Frames Per Seconds in a 1/2 quality playback setting. Unfortunately I can't show you that because I would have to use Windows Game DVR to record it.

    It also seems I was able to get this new performance by changing the "System Memory Limit" down from 16GB to 6GB from an available 32GB leaving the " Number of cached Frames" to 8, which seems to be my limit before negatively affecting performance according to Redcine-X.

Please sign in to leave a comment.