1

I upgraded my system from a Ryzen 1800x to a Ryzen 5600x, while at the same time upgrading from an NVidia 1060 (6GB) to an NVidia 3070.

Before the upgrade I benchmarked some encodes - the first was with DaVinci Resolve and the speedup was dramatic (it's easily 3-4 times faster), the second was a simple ffmpeg nvenc encode of h264 video and the performance of this encode is shockingly poor.

On the upgraded system nvidia-smi shows 100% encoding load during the ffmpeg encode, but CPU utilization is similar to single-thread levels. It's 30-50% slower than the old configuration! This is consistent regardless of which input file I use.

My command line is:

ffmpeg -i input.mov -c:a copy -c:v hevc_nvenc -b:v 1800k -maxrate:v 2400k -preset slow -sn output.mkv

Does anyone have any insights?

2
  • I assume the 3070 has the same or more VRAM than the 1060?
    – Ramhound
    Jan 5, 2021 at 14:49
  • Yes - the 3070 has 8GB while the 1060 has 6GB of VRAM Jan 6, 2021 at 23:55

1 Answer 1

1

According to a commenter on the nvidia site this is by design.

The 1060 uses the Pascal generation of GPU, which is tuned for speed over quality, whereas the 3000 series uses the Ampere GPU generation and these are tuned for quality over speed (this is also true for the Turing generation as well).

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .