dansky

Also keyframe handling are fiddly. Several years-long unresolved usability issues are in the forum like keyboard-selecting transitions etc.. The support forum is full of annoyances but BMD has mostly other priorities. I rarely complain as Resolve is still better than the rest for me

I had a few OFX plugins and maybe had the browser running which may have impacted RAM. Depending on how much RAM you can keep free and the amount of grading on that 8K video you may be OK with the 24GB of the M4 Pro base model but yeah - more is always better though with Apple it's painful to add more of anything .. on a budget.

It's a comparison of two CPU/iGPU combos I have on my desk with similar power draw. Those iGPUs are most power efficient for video editing as I like QuickSync from Intel.

The i5-1135G7 (17W TDP) has 2 Media Engines which I use for proxy generation in parallel for example and pretty versatile so I use it daily (64GB RAM..).

Still, I think it's a notable achievement to get 4x performance with the M4 for video at similar wattage of the i5. I don't have an M4 MacBook but I guess the M4 would perform similar to the one in the Mac mini.

Thank you!

I expected a fast M4 package but still was mind blown to see the video editing performance. After all these video renders run for many hours.

My 2 year old i5 laptop - even with 64GB RAM and 2x2TB SSDs upgrades - was around the same price like the base M4 Mac mini / uses similar Power. The PC surely is way more versatile with these specs and expandability.

Staying mostly in X86-land due to affordable RAM & storage, nothing I currently have comes close to the M4 performance per Watt - and now even performance per $/€ - in my video-editing use case.

I just arranged a selection of 4K H.264/H.265 clips in a 2x2 grid on a 8K timeline in DaVinci Resolve.

Playback works well - up to 60fps. However, export to H.265 creates a lot of Swap. Rendering went with 15-18fps. All videos on a SMB network drive but the GPU was the bottleneck for rendering.

Swap was even around 24GB with 5 videos which I tested first. Using 4x4K it went 9 GB before stabilizing at around 2GB. No effects or grading whatsoever - plain 4K60 SDR videos.

One single SDR 4K clip renders to 8K at 25fps. Using Superscale 2x makes that 0.5-1fps.

For 8K rendering you may be better off with 32GB RAM minimum or trying the M4 Pro model maybe with 24GB. For 4K/6K editing the base 16/256 M4 Mac mini seems sufficient when all video storage will be on external drives or network.

Edit: added single 4k->8K rendering performance

Right now I mount up to 7 HDDs to the Mac via SMB, have some Streamdeck / Pedal and the necessary external SSDs for fast storage connected. I will see if the SMB mounts come back OK after sleep (my laptop acts as server) but the Streamdeck and HDDs wake up randomly so overall it's easier to switch everything on and off depending on usage.

The 120GB/s memory bandwidth of the unified memory helps especially with video, I guess. The M4 CPU isn't really stressed out most of the time. Only multicam and HLG conversations it maxes out.

Once I patched my old Dell T1650 BIOS for ReBAR support yet the iGPU of the i5 1135-g7 had similar GPU performance for video as the Intel Arc A380 in the desktop PC. The old PCIe3 speed limited its performance. I heard others reported a smoother replay experience with Apple silicon compared to even a RTX4090.

I get some delays when fast scrubbing through a 9 multicam 720p timeline and just 360p proxies. Still impressed compared to what I was used to. Video editors may be surprised about the performance for the price.

The Mac mini M4 performance is around 4-5x in DaVinci Resolve for me - compared to my HP laptop (i5-1135G7).

Rendering HDR video was around 12fps there on the i5 - the same project in the Mac mini gets 60fps.

The M4 10 core GPU seems on par or better with a mobile RTX3060(65W) for video tests (NR / Deflicker) so I'm also impressed about the M4's efficiency. A lot of power per Watt.

It's becoming a dedicated video rendering machine for me where all the SMB auto mounting issues with macOS seem solvable. Pretty happy so far with the base model price even in the EU. The power button placement is an annoyance for me, though.

Thanks for making the dataset available. As I recorded several city walks with multiple cameras simultaneously (usually 3-6 Osmo Pocket in 4K60) I'd like to create Gaussian splatting datasets from many such walks. What costs and hardware requirements can be expected for city-wide GS datasets like the one of Prague?

I'm wondering if I can create them myself or should donate the many TB of multicam recordings.

I guess 180 would work. There is 360/VR content on YouTube yet it seems not mainstream yet. I also have a 360° cam that I rarely use. BTW, half of all my viewers watch this type of content on TV according to the stats.

I guess subtitles, HDR and decent editing are more important for now so I'm looking for AI help in the editing part. Subtitles are already supported by GPTs