Renderfarm Particles

A good question. There are a few possible causes:

- The particles aren't baked and odd stuff happens. If you submit particles, they must be baked.

- Windows / Linux / OS X machines render them a bit differently, different processors render them differently

- A bug at our end

- An errorenous results gets past the validator

I've noticed that this is mostly related to hair particles.

How do you bake hair?

You probably don't, unless it uses dynamics. If you have animated hairs, then the cache menu -> bake does the trick. But the issue is clearly elsewhere in the case of your recent Chrome session.

Yeah...for particle stuff, Renderfarm should just send sessions to one type of OS. I decided to just render the thing myself. 

Agreed, that may be a good step aheads. We have been planning a lot on using only Virtual Box clients at some point, but that's still in distant future. Like I mentioned earlier, at BURP Janus has split the rendering of workunits so that only machines of one operating system are rendering it (for at least the BBB scenes). Depending on our numbers, it might be reasonable to do this for all of our sessions. There are some possible issues though, such as a small chance of some .blend being absolutely impossible to render on Windows while it renders fine on the other operating systems. I'd like to investigate these problems in more detail at some point to figure out what sort of differences there are between Linux distros, processors and the bitness of the system.

That would definetly be nice. Can't wait until this is implemented. Hopefully, never again will I have to render on both my computers for two days because of some flickering grass. :)

It definitely would make sense to do this for sessions in the future.

I'm running some experiments right now. There haven't been any bugs, but obviously my testing is flawed. Please send me some bugging hair and add "[Hair Bug Hunt]" to the title. That way I will accept them if nothing else is using the service.

- Keep the files as homogenous as possible (nothing redundant), but try to make them into something that you might use in a real scene

- I'd like to have some grass as that's what's bugging the most

- Fur seems to be another issue too

- This is a hair particle test as I haven't observed any problems with point particles

- Not too many frames. Less than 200 please

- Short render times if possible

Alright, it would seem that the culprit might reside in child particles.

Test #1 featured static particles without any physics or baking. They worked perfectly.

Test #2 featured hair dynamics with the hair being baked.  This also worked perfectly.

Test #3 featured a collision, hair dynamics and child particles. Here we see some issues, but whether or not these are actually related to the particles remains a bit unknown. The problems we see in the video could be clipping issues (Blender not knowing which of two objects being rendered is on top of the other one), but they also seem like they might be our standard particle problem in action.

Test #4 was conclusive. We have a static, grassy field with no hair dynamics or baking. The problems are clearly visible: particles jumping to odd locations at some frames. Originally Janus proposed that the issue at our end in regards to the BBB sessions is our Linux client, but looking at the results for test #4 we can see that there were all sorts of machines returning those frames which were bugging. In fact, the first erroneous result was returned by a 32-bit Windows machine, while the second one was a 32-bit Linux machine. Earlier in the animation, 32-bit Linux and Windows machines returned correct results.

What can we suggest here? Baking seems to help, and child particles were involved in both bugging sessions. Not one client is responsible, which would have made things easier (splitting fur-based animation for only one operating systems). Once there aren't any user animations running I'll launch one more test which will have a grassy field done 100% with parent fur particles - no children at all.

Oookay... since I have more particles by hand, Ill post some more ;P

I've rechecked http://www.renderfarm.fi/animations/3720 and http://www.renderfarm.fi/animations/3698

Athena/renderfarm was really flawless. It had the "usual" hair-particles. Even with a bit of combing. Tough rendering went really fine

"Kitchenmates" had indeed lot's" of errors. Tough due to the nature of that shot (total-shot) the effect kept rather subtle and I think I can hide it even more within postproduction.

The only difference here is: The Athena-animation has no "kink" involved in the fur particles. However, "kink" is a very important setting for fursetups.

Now I will upload "The Garden" again. I added another fix (re-established the particlesystem of the grey guy completely) and introduced a rather simple postcompositing enviroment (Oh yeah). Uploading a short file for testing purposes does not make too much sense IMHO. It is just to huge (230MB) for that and I think the first 182 frames will come out in a usable fashion anyway. It is a total-shot again.

The test/trick starts at frame 183 where the camera goes for a portrait shot. Fur artifacts there = very bad. I hope for a nice outcome tough.

If you wish you could have a look at that session. If we have unbearable artifacts showing up after frame 183 it can be stopped (while keeping the frames 01-182 alive for download tough ;)