Hardware Decoder is indeed Intel -- Intel HD Graphics 630.

I use that to decode as it does a speedy-enough job, and keep my RX470 for "GPU acceleration of video processing" -- which I sometimes disable depending on project and stability.

I'd never noticed the RAM difference since we haven't frequently use the Hero 11 in our setups, preferring our older Osmo Action. The OA doesn't do *as* well in low light and has a few less features, but it's been extremely stable with lovely imaging otherwise, stability not being something I can say for the Hero lineup since Hero 5's.

Creating a proxy or DI for GoPro is a pretty standard recommendation. Unfortunately, formats designed for acquisition don't lend themselves very well to editing or delivery.

Oh -- I totally understand the need for proxies in some situations.
That's not this situation.

This post is simply to point out that using the Experimental HEVC Decoding has some serious RAM costs that can affect system stability.

We can edit the GoPro footage just fine on our newer laptop, and on my desktop when that setting is disabled. THAT said, if I had many H.265 cameras to work in our multi-cam flow, I suspect proxies would more likely be needed.

Thanks for clarifying.

Hardware Decoder is indeed Intel -- Intel HD Graphics 630.

I use that to decode as it does a speedy-enough job, and keep my RX470 for "GPU acceleration of video processing" -- which I sometimes disable depending on project and stability.

I'd never noticed the RAM difference since we haven't frequently use the Hero 11 in our setups, preferring our older Osmo Action. The OA doesn't do *as* well in low light and has a few less features, but it's been extremely stable with lovely imaging otherwise, stability not being something I can say for the Hero lineup since Hero 5's.

I noticed a huge ram usage increase in VP21.300 and up with QSV on my laptop which led to just about running out of memory with my 32GB of ram. I also noticed performance drops with MxCompound and QSV vs 208 with so4compound.

I am using NVDEC and MxCompound with 315 as it seems like the best combination at present for my systems and media. It uses VRAM vs system ram.

@RogerS For those following along -- which buttons enable MxCompound vs socompound? ;-)

Default file io decoding settings in VP 21.300+ use the MxCompound decoder. Check the legacy AVC/experimental HEVC box and you get so4compound. You can see yourself looking at the properties for media in VEGAS.

Hardware Decoder is indeed Intel -- Intel HD Graphics 630.

I use that to decode as it does a speedy-enough job, and keep my RX470 for "GPU acceleration of video processing" -- which I sometimes disable depending on project and stability.

I'd never noticed the RAM difference since we haven't frequently use the Hero 11 in our setups, preferring our older Osmo Action. The OA doesn't do *as* well in low light and has a few less features, but it's been extremely stable with lovely imaging otherwise, stability not being something I can say for the Hero lineup since Hero 5's.

I noticed a huge ram usage increase in VP21.300 and up with QSV on my laptop which led to just about running out of memory with my 32GB of ram. I also noticed performance drops with MxCompound and QSV vs 208 with so4compound.

I am using NVDEC and MxCompound with 315 as it seems like the best combination at present for my systems and media. It uses VRAM vs system ram.

Thanks for the tip on QSV. My Dell laptop with a RTX3070 Ti and Intel Iris Xe just completed a 1 day+ render, with memory utilization exceeding 85% (not a good thing) but it finished at a snails pace! Turns out, 17 minutes into the 2.5 hour show render, all I got was a black screen. :( Turned off the Iris decoding and re-rendering now. CPU is higher.. NVideo utilization way higher since I'm using it for decoding, but memory usage seems to be way lower (currently -- was 39% at start of render... currently climbed to 41%)

The reason for the Loooong render time is I set frame re-render to Optical Flow. Why? Two of our cams were shot 1080p60; one was shot 4Kp30. Rendering to 1080p60. In a test render, found the 4Kp30 footage looked best at 1080p60 when using Optical Flow... but mannn... it slows things down when that camera's clips are used!