I noticed that this approach also improved performance on the Pi4, so I'm also using it there. Looking further into this, I finally got the time to look into how NV12 data is arranged. I got a vague idea based on reading the forum/code comments, but visually seeing them finally made it click for me.
The overclocking section in the documentation notes that the core clock is now 910Mhz up from 400, so this might be a fast way to transpose 4K HEVC frames as well. Given "Pi5 has scaled back the VPU significantly" mentioned earlier, there's no way to use this, correct?
The overclocking section in the documentation notes that the core clock is now 910Mhz up from 400, so this might be a fast way to transpose 4K HEVC frames as well. Given "Pi5 has scaled back the VPU significantly" mentioned earlier, there's no way to use this, correct?
Statistics: Posted by dividuum — Mon Mar 03, 2025 10:03 am