This is an open invitation to all our wonderful users to help us decide how to develop Crowdrender. We're working towards releasing a beta, but we need your help!!
We'd like you to comment and vote on this thread, here's how.
If you have used Crowdrender, and don't yet use it for your real projects, tell us what we need to do to make our software good enough for you to use for real work. Write a comment below that explains what features you need or problems you face that need to be worked on.
If you see a comment that you agree with, like it and please also expand on it by replying to that comment. This helps us tell which issues are important to everyone!
We'll use what you tell us to help us decide what to work on next. From now on, we'll be releasing frequently, with smaller updates so you can get what you need much faster.
Thanks to everyone for taking part, we can't wait to see what you post :)
Crowd Bakery... rendering is fine (cycles (eevee doesn't work for me)
but Physics sim baking still takes forever because it is single threaded.
I would like to spin up multiple CR-docker instances on my dual Xeon server... that way would at least see more than 2/56 hyperthread cores in use...
Not sure Blender can give this kind of low-level access to an add-on but if possible, a "Join render when ready" option could be a true time saver.
The other day I had trouble syncing my machines and it ended up taking more time than if I just rendered locally.
It would be amazing if there was a possibility to start a render even if not all machines are synced and whenever they are, they would take part of the remaining area to render.
Hi James, it's been a while but I'm back on using CrowdRender! Just posted in General questions btw.
To me the most useful next feature would be support for all linked libraries. So if a path only exists in the client machine, CR would generate a new path on the server machines, upload the files there recursively and modify the blend files to change where libraries point to. This would enable a non-destructive workflow on the client. I assume it's not an easy one since it requires to recursively modify blend files on the server machines, but I'd like to know what you think of it in terms of feasibility and possible limitations.
Cheers!
Step 1 - Needed implementation. Viewport Shading button is gone, would be great if there was a setting like "use local computer for Viewport Shading" and settings for what HW type to use, like Optix, Cuda, Metal and so on.
Steg 2 - Would be great !! Make is so that the tiles rendered on the CrowdRender network was sent back as soon as they were done. And maybe make the tiles small, and we have a competitor to Cinema 4Ds Team Render. Step 3. This is the big one, the dream one, the üiber one, that would very much put you guys at the centre of the map. Make it so that you can use an IPR or Viewport Shading with the power of CrowdRender.
Hi James, maybe crowd render that works with external file..... that's gona be awesome....cheers :)
Yes, absolutely! Distributing via buckets rather than a percentage of the frame would be fantastic. I'm trying to get our organisation to switch to using blender in production but having a network rendering solution is crucial to being viable.
We currently use vray on another software which works reasonably well and uses the bucket allocation method. I thought this was standard so it took me a while to get the method currently used which I think (sorry guys) is not as effective. Though much better than manual rendering on each machine! I would be so happy for this to be implemented.
why one of my pcs utilizes only 28% of the cpu? only with crowd render
Better sync.
I'd like to be able to specify a "root folder" for each node, correlating to a mounted drive, or cloud storage of some form, where all paths to blend files, textures, etc. are always the same on each node relative to that root.
I already have a shared asset library of models and textures that I use across multiple scenes, having to pack these into each and every blend file is insanely wasteful both in terms of storage usage and sync times.
Hello everyone.😎
I would like to raise the previous idea of each connected computer doing its own frame during animations.
Similar to the 32 slices or nodes presented for an individual image, but in this case for frames of animation.
I have seen previous discussion of this, and I am wondering how close this is to being realized?
Thanks to the whole blender community, and special thanks to the CrowdRender crew.
We love you, E:)😍
Great work, can't wait to see this grow. Here's few suggestions: 1.clear way to "turn on/off" CrowdRender so it doesn't do stuff in the background. I don't know if it does or not. Maybe that's the problem. So on-off-switch. 2.an option to render "one frame per computer", so that two computers don't work on the same one frame (For animations this would prove better since other computers wont have to wait for the slowest one to finish, though I'm only guessing) 3.clear, under 5 minute tutorial on how to set things up quick. (including installation on each pc, how to search ip and device name, which settings to use..) I guess these have been mentioned before but I wanted to give hobbyist blenderer's perspective on what I'd like to see. Cheers!
Centralized upload and download server/point, now it looks like it upload and download to each "slave" ? I have bad internet at home an it would be awesome to slim down the internet need.
Ability to connect to render slaves over internet. Like from home to the office. Bonus if it would work as Teamviewer,, so no wierd ports needs to be open and such.
Hey @Mike B and @Eddy :D, we're releasing the next version of Crowdrender, hopefully next week, we'll have full support for Optix, I recently was able to test this on a new PC we built, see https://www.instagram.com/p/CEWB-4nDjsb/?utm_source=ig_web_button_share_sheet
is there now an official version which supports optix via network?
So is there now an official version which supports optix via network?
As an IT admin I would love the ability to run this as a service. There would be a central management app on a server we could setup on the network. The service on each client would be setup to report back to the management server by IP or Computer Name. The management app would have a queue of blender jobs. The management app would be responsible for distributing frames to all the computers connected. Since networks like these have many computers that are used by users, the management app could be set to only send jobs between two times. It would also be responsible for each frame getting done. If a computer fails to do a frame, that frame could be resent out to another computer on the network. Computers may even drop off the network only to appear again later. It would be nice if the management app could quarry the client to see if it is idle or being used by a user. If no one is logged in, and it is withing the time set by the admin, the management app could add that client to the list of available pc's and start sending jobs to it. Email notifications would be nice to have but not needed.
Have you tought about creating some sort of rendering nodes market? I mean, many people have powerful hardware that can be used for rendering by other people, but they don't want to share their computers with strangers for free. It would be great to be able to rent out computers in public cloud, where other people could get them for rendering using your system. For 3D artists it would be great too, because it should lower costs and make cloud much bigger and more accessible for all.
Hi there! I've just tried the optix version of the addon.. there are some strange behaviours.. first of all, I can always select optix even if my second computer doesn't have it. Then, if I right click on the new node, I have some blender options, unrelated with the addon which is weird. But above all, the render has been done 100% by the optix machine, while the slower machine set to cpu render continued rendering even after the frame was completed.. I also tried to launch the frame render a second time and again was completed 100% only by the faster machine, while the status of the second slower one was still stuck at "7%" of the previous render. I tried to manually set the load balancing setting "1" for the slower machine, still the same behaviour, but now the frame was empty. The last weird thing I want to point out is that when it renders, it switches render engine from Cycles to an empty one. Once the render is finished, it goes back to Cycles.
I will!
I'll go to sleep now, it's 1 a.m. here! I'll try the build tomorrow! Thanks!