Resolve Studio/Fusion 17 with .braw 6k Unusable Playback

Hope someone here can advise the best way to handle/resolve/deal with this. I'm on an almost new computer (within this year).

OS Catalina.
2.4 GHz 8-core Intel Core i9.
32 GB 2667 MHz DDR4 AMD
Radeon Pro 5600M 8 GB
Intel UHD Graphics 630 1536 MB

Shooting .braw at 6k (I have the project timeline resolution set to 1920x1080 HD), everything is pretty smooth until I enter Fusion. Once I start adding nodes, playback becomes laggy. Once I add in a second piece of footage to composite (also at 6k) it gets unusable. Shooting at 60fps, I might be getting 19-20 frames/second playback in the read out on the lower right corner of the screen.

If I stack the tracks in the Edit page and then create the Fusion clip so that I'm only compositing at 1920x1080 it's not *as* bad but performance is still not exactly smooth

What am I missing that I need to learn? I tried generating and using optimized media to no avail. Working off of the internal SSD was just as laggy as my TB3 connected 8-disk RAID enclosure (working off of the SSD), so I don't think it's a drive issue.

Increasing RAM allocation in Preferences doesn't really improve things enough to make this usable.

I don't need it to be smooth enough to leisurely watch a film in Fusion. I just need it smooth enough to keyframe to tempos for instructional guitar videos. Right now it's too laggy to even judge where downbeats should be, etc.

All of the practice examples in the training books/courses work like a champ. Playback is totally smooth when I'm working with the material provided by BMD in "The Visual Effects Guide To Davinci Resolve 17"

If you can help me out of this dilemma, I'll love you forever.

Sean

If I stack the tracks in the Edit page and then create the Fusion clip so that I'm only compositing at 1920x1080 it's not *as* bad but performance is still not exactly smooth

What am I missing that I need to learn?

The key factor here is that going into Fusion on a clip that's directly on a timeline will see the full native resolution of that clip. You can see that by looking at the top right of a Fusion viewer, where it shows the resolution.

So you have a 1080p timeline, with 6K clips on it. Edit and Color are seeing and processing those clips at 1080p resolution, which provides reasonable performance on your machine. But Fusion is seeing and processing those clips at their full 6K resolution. And your system is not nearly powerful enough for that. To be honest 6K at 60 FPS in Fusion would be a struggle for any hardware.

Making a Fusion Clip changes this: a Fusion composition on a Fusion Clip will see the media inside the Clip at timeline resolution; 1080p in your case. Doing that means Fusion processes the clips at the same resolution as Edit and Color are doing, and performance will be more acceptable.

The fact that it's still slow is down to:
- Your system not being that powerful; especially the GPU is quite weak, and 32GB is the minimum recommended for using Fusion in Resolve.
- 60 FPS is a higher frame rate than most people operate at. It's going to require twice the processing power to play at full speed as would 30 FPS, and twice the RAM per second to cache frames that have been processed. You mentioned getting up to 20 FPS playback, which is actually pretty good for 6K source media, even in a Fusion Clip. If your source media was 24 or 30 FPS, that'd be close enough to real time to be quite acceptable. 60 FPS is a hard ask for a lot of hardware.
- Fusion isn't hugely fast to begin with. It's a common situation to find that things are pretty smooth until Fusion is involved. That's just how things are at the moment. It's an amazingly powerful tool, but it's not all that fast by modern standards.

So if you're going to put a composition on your source media clip, definitely make a Fusion Clip as you have been doing. That will make Fusion usable. But as long as you're processing 60 FPS on that hardware in Fusion, there's always the chance of slowdown; how much depend on what you're doing in Fusion, of course.

To help improve things further you can:
- make 1080p proxies (not Optimized Media, which is a lot more hassle than Proxies). By creating proxies in a format like DNxHR or ProRes, you avoid the decoding and debayering work that the system currently has to do for every BRAW frame, and by making them at 1080p you also avoid the scaling work that happens when you have 6K on a 1080p timeline.
- make sure you have Render Cache on, and that you're caching the Fusion frames as they play. Choose a suitable Render Cache format, eg DNxHR HQ or SQ . Then every frame that's played will be written to disk, and won't have to be re-processed if you play that frame again (unless/until you change anything of course). Make sure Render Cache is writing to your fast local disk.
- render out your Fusion compositions once they're ready, so you don't have to keep re-processing them later. The new Render In Place feature can be great for that. If you already have Render Cache generated, Render In Place will use that and will complete very quickly. Be aware that this will hide your Fusion composition - you can get back to it with Decompose To Original, but you might also want to manually back it up first by dragging the Composition to the Media Pool.
- does your Fusion composition actually need to see the source media? For example if you're overlaying a title or a graphic on the source media, that could go on a separate Fusion Composition which renders the title/effect onto a transparent background. That Composition can then be placed on a higher layer in the Timeline, and Edit will overlay it over the media. With no source media involved, less work is required from Fusion.

If you're the guy who posted about showing guitar tab alongside a video of the music being played, then that could certainly go on a separate Composition. No need to see the media in the Composition, at least not from a Fusion perspective. You can still hear the audio from any Timeline clips when in Fusion.

In that scenario, you wouldn't have a Fusion Clip. Just a Fusion Composition, which will automatically be at timeline resolution.

Final point: if you're using still images (PNG, JPG, etc), do not drag them into the composition from the Media Pool, which creates a MediaIn. Instead add a Loader node and load them directly from disk. Using MediaIn for still images can be significantly slower than a Loader node, as MediaIn does not cache still images properly; it handles them like they're video footage that change every frame, when it could just cache them in RAM once.

If I stack the tracks in the Edit page and then create the Fusion clip so that I'm only compositing at 1920x1080 it's not *as* bad but performance is still not exactly smooth

What am I missing that I need to learn?

The key factor here is that going into Fusion on a clip that's directly on a timeline will see the full native resolution of that clip. You can see that by looking at the top right of a Fusion viewer, where it shows the resolution....

Just a very quick follow-up response (I deleted most of the quotation to make my reply more noticeable) to extend my sincere thanks to you for the very detailed and helpful explanation. This will help a *ton* and demystifies the issues for me. I'm coming from an audio world where I primarily use real instruments (not many virtual ones) into Pro Tools as a musician of 30+ years. I quickly realized it's a blip on the radar when compared to video.

I'm shooting at that high of a frame rate (combined with a ridiculous amount of set light to accommodate a 45-degree shutter angle), in order to demonstrate a lot of very complex material that is *clean clean clean* when slowed down even to 25%.

Your answer now explains why those training files are so smooth and my system is crapping out under the conditions of what I'm asking it to do currently. If I'd just been able to shoot at 24fps with a 180 degree angle (maybe not even at 6k, which I chose to keep things clean when zooming in post), it sounds like this would have been considerably smoother. I think based on your suggestions that I'll find a way to get what I want at 60fps but it helps me to know that it's what I'm feeding Davinci and asking it to do on my system.

I am the guitar player who was asking about the animated notation/TAB. The original idea was to shoot everything literally 5 times. Even after 30 years of doing this (15-20 professionally), the fact that I *can* play things that perfectly 5 times in a row didn't mean I *should,* and it was literally a deal breaker to play it once, then mime to the playback *perfectly* to shoot the right hand pick stroke (now demonstrating RH technique), then re-record it again in slow-motion. I realized it would take days and multiple takes of everything to make it perfect. Far better to play something in the moment and have clean footage for everything, when zoomed, when slowed, when full tempo, etc. Hence, I went to a pair of BMD cameras and a ton of light.

I'll be revisiting your generous response as I dial in the final details to just get started (I spent 2 years writing 1500 pages of manuscript and the past 6 months just learning how I *will eventually* shoot the videos. Which spiraled into taking the self-training online courses in Resolve).

Thank you again. I really appreciate it.

Sean

- make 1080p proxies (not Optimized Media, which is a lot more hassle than Proxies). By creating proxies in a format like DNxHR or ProRes, you avoid the decoding and debayering work that the system currently has to do for every BRAW frame, and by making them at 1080p you also avoid the scaling work that happens when you have 6K on a 1080p timeline.

Hi Tom, thanks again for all of your help the past few weeks on here. I was going through this note again late last night and messing around with Project Settings while testing performance. Proxies obviously help a lot but I'm confused on how to make them at 1080p.

In Project Settings under Optimized Media and Render Cache, I can only see resolution options for "Original," "Half," "Quarter," "One-Eighth," and "One-Sixteenth," along with "Automatically."

I *did* find a video that says "choose automatically" is supposed to create the Proxy at the timeline resolution, which I have set to 1920x1080. However, when I do that, the resolution of the Proxy file is showing up to actually be 3072x1280

Last, with RCM on in Project Settings, the proxies look terrible when viewed in Finder but I don't think that matters in this case since I'm not using these to collaborate - and in the Davinci viewer, the basic contrast is normal. it's just when viewed in Finder that the contrast is way off and the image is unusable. I've double checked that it's really working through since when I go into the Color page and change the .braw settings on the clip level, there is no change so long as "Use Proxy Media" is selected. But when I render the file, it renders whatever changes I make at the clip level. I'm not making many (if any) changes at the clip level in general, but I did that as a way to confirm that my system is indeed using the proxies in the Davinci viewer, which looks fine (with Color Management enabled) as long as I am in Resolve.

If I don't engage Color Management and just leave it at "Davinci YRGB," the proxies in Finder look the same as the flat, raw footage in Davinci. .

I started using Color Management after going through the Color Grading course for a couple of months but am also fairly comfortable now if I had to grade without it. In this case, does it really matter? What is the protocol for generating proxy files with color management on? Probably more importantly, should I just leave the resolution set to "Choose Automatically," and accept that it is generating the Proxy at 3072x1280?

Thanks as always,

Sean