I have been playing around with the latest Scatter and Disperse and here are my findings.
Note: I also use Forest Pro and MultiScatter, but since both actually are bad for painting instances I hope Disperse (and Scatter) will fill in the gap.

Attached are renders from FP and MS.
The FP version has 1M instances and renders in 7m30 (of which 6m55 is real rendertime); RAM use 15GB
The MS version has way more instances (there is no way to read out the exact number) and renders in 10m (of which 8m30 real rendertime); RAM use 7GB.
The trees are VRay proxies of meshes with 750K-2M poly. System: Win7, Max2018.4, i7--3930K, 64GB RAM.

The scene has a BW mask for the distribution.

When trying to do this scene in Scatter, it generates around 450K instances. On render it starts with "updating instances", which takes 20 min. after which I killed the Max process. So clearly there is something wrong.

Trying to paint these kind of landscapes with these kind of trees/proxies (after converting the meshes into Sini Proxies) with Disperse, results in an extremely slow interface and "updating instances" is taking very long already at 10K (!) instances. So clearly something wrong here too.

Wishes:
- drastically reduce pre-render preparation time
- a display mode that is better for handling large number of instances (n-th, max. nr. of instances, etc.)
- a display mode that is better for detailling a part of the distribution surface: using a spline to only show all instances within this spline (only for display, not for rendering)

I also noticed that when using heavy poly meshes and large numbers of instances, the last bucket when rendering takes ages to finish: while doing so the used memory is slowly reduced (from 24GB to 2 GB takes 6 minutes!)

As it is now Scatter (and Disperse) are not useable for large meshes/numbers of instances. I also have been using GroundWiz for painting objects/scattering and recognize the long "updating instances" times with this plugin too. Probably the reason why it is not developed anymore.

Until 15 Years ago I used Lightwave on 32bit systems that easily did billions of instances with a plugin called HD_Instance (by converting meshes into a volumetric engine, so only 1 instance of each mesh is copied to memory). One only had to spray or generate particles around to indicate where the instances should be generated. Hopefully one day painting these kind of instance numbers in Max will be that easy. Until then I paint B/W-masks that drive density and scale in FP or MS.

Hi Erik,

The tech support team are off today and should be able to help tomorrow.
In the meantime, I can say this seems to be an isolated case, as our other users are not having issues.

Both Disperse and Scatter are capable of rendering billions of polygons and have been Q&A tested to ensure both perform as expected.
As you will appreciate Scatter is in BETA and will likely be improved before it is officially released. Your input and feedback are valuable to us to improve our tools.

May I ask a favour. To help the tech guys resolve your issues, would it be possible to send us the problematic 3ds Max scene so we can test ourselves?

It would be handy to also understand the following;
- 3ds Max version and release.
- V-Ray build.
- Graphics card and RAM.
- Other plugins in 3D scene.

Thank you.

Nigel

Of course! I'll send everything you need to get this solved! Will prepare a clean and lean scene showing the issues Monday or Tuesday.

Thanks Erik,

I hope you enjoy the rest of your weekend.

Regards,
Nigel

Just opened support ticket with files attached. I'm curious about your findings! If you need more info just let me know.

Hi Erik,

Thank you for the file. It's proved very helpful. What a load of trees!

The development team have identified the issue and will be working with the team at Chaos Group to fix it for the next release.
As this is the first beta release, it's good to know we have great support from artist like you, who are willing to push the product to the extreme.
Reminds me of Elon Musk's first attempt at launching a rocket but in this case almost a trillion polys (so the team tell me!).

We'll keep you posted on progress and let you know when the next version is ready to test.

Thanks again for your support and feedback.

Kind regards,
Nigel