Working with GH6 5.7K video using RTX 3060 12GB

Hi,

Whether a GPU with 12GB VRAM would be sufficient or not to do basic grade GH6's 5.7K (5728x3024) was not immediately clear to me when buying my GPU, especially given Puget System's recommendation of 20GB for 4K and beyond. It turned out that the answer was "it does work, albeit without spectacular performance", and I'll like to give a bit more details for anyone who owns a GH6/GH7 and contemplating buying a new GPU to do basic grading with Resolve.

My hardware is Intel 12600K CPU, 128GB RAM, RTX 3060 12GB GPU. On the software side, I'm using Resolve 19 on Arch Linux, with up-to-date drivers. I recorded in MOV format, HEVC 10-bit, log, at 5.7K resolution. My timeline in DaVinci Resolve is at the same 5.7K resolution, and I'm exporting 10-bit HEVC MOV again, using NVENC H.265 with ultra high quality slower preset. During the encode, I'm also running a KDE Wayland session which also uses a bit of VRAM.

I do my basic grading using 10-15 nodes in the DaVinci wide gamut space, and I use temporal noise reduction with 2 frames with best/largest settings. Enabling Spatial NR works with fastest/basic/enhanced with any radius, but when I switch to UltraNR, I run out of VRAM.

With 2 frames temporal NR only, my encoding works and the speed is around 4.5FPS, and Resolve uses around 10.3GB VRAM. Total system VRAM usage was around 11.4GB.

I hope this information will be useful to someone!

Edit: I'm not soliciting any advice or opinions on my work flow itself. This post is for informational purposes on VRAM requirements when working with 5.7 video and timeline.

The problem here is that the GH6/GH7 can only record 5.7k in LongGOP H.265/HEVC, which is much more processor intensive than an ALL-Intra codec. If you record 4k All-Intra you will see a significant difference in performance.

Why are you using a timeline in 5.7K? There is no device to screen that.
If you need NR, downscaling (try different algorithms) will already reduce noise. In a 4K or UHD timeline you will need less VRAM.

Unless you're working on a transfer of analog-to-digital assets and need uncompressed 444 color for final transfer to film, there's absolutely no reason to be working in that non-standard frame/color space. You're creating an overload situation for resources and processing power. A standardized 4K workflow is more than sufficient in a 422 10 or 12-bit workspace.

I shoot GH6 too, use as the " full stage" framing for theatre shoots. So 5.7K, 60P, Vlog. Project though is 1920x1080 60P. With scale to fit in project settings. So just a little crop makes the 17x9, 16x9 but gives the opportunity to zoom in and/or pan if the other cameras miss something important. The multicam edit is done on the Studio Max as it has hardware decode for the files from GH6 and GH5S. However I often use the PC for final render if there is a lot of noise reduction used or Fusion titles. The Threadripper can manage decode for one track at a time and the 4070Ti is much more powerful than the Mac. Can be 4 times as fast in a render. The 4070Ti has dual encoders not in the lower level NVIDIA GPU's.

Apologies for not saying this up front: I'm not a professional like most people here, and I'm just recording private family videos. I like the 5.7K (at 60 FPS) resolution since it allows me to capture most pixels on a GH6, which I can crop later on to fit my needs.

CPU-wise, DaVinci uses just two cores, so single-core performance seems to matter, but it seems it really comes down to GPU.

I'm happy with the fact that it works at all. I'm planning to upgrade to a an RTX 50 series GPU soon for the performance, but it's not crucial to me. I rarely do encoding, and when I do, I just leave the computer do its thing over night.



Nonetheless, thanks a bunch for the comments!

Please do yourself a favour and get familiar with the scaling and cropping options in DR, they are both in the project settings and in the Inspector (for the clip at hand). DR will use the full resolution of your sources where needed, but you can still set your timeline to UHD or even HD if the machine is getting slow.

The secret to a great editing experience using Resolve is PROXIES, and doing as much of your edit in an HD timeline as possible, before changing (via project settings) to a UHD timeline for final export.

Thanks for your inputs, but truly, I am not soliciting any advice.

As I mentioned in my original post, my purpose is to provide information about what is possible with 12 VRAM when working with 5.7K videos in a 5.7K timeline for anyone else out there who may need this piece of information.


I'm well aware of the scaling and cropping options in Resolve (it's not clear to me why you chose to assume otherwise), and although this is not my day job, I have been using various photo and NLE video tools in the past 30 years.


I prefer to keep all the information (i.e., pixels) unless it can be readily discarded/cropped (which is not the case most of the time for me), without any distortion that is inevitably caused by scaling (I'm well versed in scaling algorithms, having implemented them in various programming languages over decades).

Hello,

Rtx2070 here and gh6 6K60 shooter here. Upgraded Cpu to 12700, but ud dus not work out for large edit.

So I work with proxies doe now.

If RTX5000 cut through 6K60 H264 like butter in DaVinci I would be very happy to hear!

For your info I am not a professional either, film then video has been a hobby of mine for best part of 60 years. What we are trying to say to you is you do not need to edit in a 5.7k timeline to get a 5.7k output in the end. Resolve can edit in one timeline and export into another. So you can use your source 5.7K , like I do, edit in a 1920x1080 timeline and then render to 5.7K. Just change the project settings. We are also saying that when you watch this on a normal screen/TV ( there isn't a screen for 5.7K that I know of) it will be scaled to the resolution of that screen by the PC/TV scaler etc. Rather than let some uncontrolled scaler make that change, most of us prefer to do that in Resolve where we have more control. Can then make outputs for 1920x1080, UHD etc with or without HDR. You can do this all from one edit done in 1920x1080 which is much easier for a PC to manage. My PC has no problem with one track of GH6 5.7K Vlog 60P. The PC just cannot manage that track and a C4K Vlog 60P track from the GH5S in multicam. The Studio Max can do this. That is because of the hardware decode in the Mac. When it comes to rendering things are different and the PC is often a lot faster because the 4070Ti is a much faster GPU than in the M1 Mac.