So I had a chance to test Crowdrender (CR) this morning and had a few things that came to mind.
Some if it you already mention in the setup video, but I wanted to give you my full set of findings. :)
- Live Tile Updates: When test rendering, it's so important to see the rendered tiles. I stop a lot of renders after a few tiles because I'm just figuring out how much noise is acceptable. If I'm waiting for the full image every time, I'm actually working slower. Updates before the tiles are done would be even better. Most of the distributed rendering I've done with other packages supports this, so it would be nice to see CR do the same thing. (There could be a bunch of extra options you could enable / disable depending on your network speed in the next points, "live" tile updates could also be one of them.)
- Custom Working Directory: Having CR dump everything in to my home folder on Linux is something I'd like to avoid. My machines have a custom scratch disk set up for this kind of thing. It would be nice if CR offered one of two options (or both). A) Be able to define a custom working directory for in in the Add-on prefs. B) Use Blender's temporary directory as a starting point.
- GPU Load Balancing: I'm sure this is something you guys are working on. ;) But I have one computer with three gpu's and one with two, yet the one with two seems to finish a lot faster every time. Maybe look into a benchmark scene you can optionally run the first time you setup CR that can give you some good info for load balancing? No issues mixing AMD and Nvidia so far though, so that's good! Or maybe even progressively smaller tile sizes to keep both machines working until the end?
- Background Sync / Auto Sync: I have a dedicated wired network for my work machines, so it wouldn't be an issue if I had the option to Auto Sync / Background Sync while I'm working. Maybe something simple like an interval setting, or a "Sync on render start" checkbox so I don't get half a pink render when I forget to sync before I render. Again, it's all about time-saving.
- DWAA EXR's: The VFX TD in me has a small aneurysm when I see lossy compression being used inside an EXR. I get that the file size is something to consider, but it would be nice to have the option the use lossless compression instead. For AOV's it's vitally important that the data doesn't get messed with. Another checkbox could do the trick to enable it. (Again, wired network etc.)
- Locked UI / Fullscreen Rendering: I generally render in fullscreen mode with a locked UI, but crowdrender doesn't update the top text then it seems, so there's no visible feedback. Not a huge deal, but might be something to consider.
This all being said, the Alpha runs pretty smooth so far for me. You guys are doing a great job! A good distributed render platform for Blender has been something that I've been looking for for quite some time now and what is here so far feels really nice already. :)
For reference I'm using two machines bot with multiple gpu's, one with Nvidia cards, one with AMD cards. Both running Xubuntu 16.04.3 and syncing over a local wired network.
Hi Daniel :) Yes, that is currently the expected behaviour, but the next pre-release for 0.1.3 will store the optimisation data so that each time you load the scene or resync, it will still be as fast as it can be :D. Stay tuned for the announcement as to when this version will be released to the early access testers. Should be in a week or so from now.
James