Hello.
I'm a Maya user and I want to use ICE lagoa to simulate various items.
I'm currently learning ICE with tutorials from Digital Tutors, and I've noticed that my simulation runs slow.
I'm using example files from the tutorials and when simulating, my task manager shows an average 30% load on my cpu system.
I'm on a powerful AMD Ryzen Threadtripper 1950X 16- core, 64Gb memory, and I have 2X Nvidia GeForce GTX 1080 Ti.
How can I utilize the full power of my system?
Thank You for Your advice!
(I hope this is the correct place to post this question?)
Roger Gihlemoen
Utilize full CPU potential
Re: Utilize full CPU potential
Yep, correct place.
Lagoa is a little old, but if it helps, turn off "Index" of the harddrive, making the write speed faster - and also "Cache" the particles. Playback will be realtime after caching. Calculating first time round will take a little while,but afterwards all good for scrolling. Another trick is.. don't use spheres, but cubes for the particles (to first calculate) then push them later to spheres. Also, go for a lower density first to see how it roughly works - a simple ICE switch helps, then up the density on render/caching.
I think.... if things also may help, try Mootzoids plugins, now all free. He has some AMAZING tools that are unparrellel anywhere and they all mostly work in ICE like Lagoa. Maybe your particle effects can be rendered realtime on GPU via his systems, or you get better topology/UVs on the liquids or whatnot. I have not had the luxery to play with them though... but I do know with his ICE upgrades you've got good shelflife on ICE skills.
Also, previous tip, when I turn off the compressing and indexing, I get things going a lot faster in most software, including Unreal Engine running x5 times faster than usual (Windows 10).
Also, turn off Facerobot when you can and use wireframe view or Shader display.
Lagoa is a little old, but if it helps, turn off "Index" of the harddrive, making the write speed faster - and also "Cache" the particles. Playback will be realtime after caching. Calculating first time round will take a little while,but afterwards all good for scrolling. Another trick is.. don't use spheres, but cubes for the particles (to first calculate) then push them later to spheres. Also, go for a lower density first to see how it roughly works - a simple ICE switch helps, then up the density on render/caching.
I think.... if things also may help, try Mootzoids plugins, now all free. He has some AMAZING tools that are unparrellel anywhere and they all mostly work in ICE like Lagoa. Maybe your particle effects can be rendered realtime on GPU via his systems, or you get better topology/UVs on the liquids or whatnot. I have not had the luxery to play with them though... but I do know with his ICE upgrades you've got good shelflife on ICE skills.
Also, previous tip, when I turn off the compressing and indexing, I get things going a lot faster in most software, including Unreal Engine running x5 times faster than usual (Windows 10).
Also, turn off Facerobot when you can and use wireframe view or Shader display.
Re: Utilize full CPU potential
Thank you Draise for all those tips and tricks.
I went through your comprehensive list and had a look at them all. Appreciate it!
Although this does not address the issue of ICE Lagoa not beeing able to utilize my CPUs full potential. It runs at about 30%.
My simulations should be able to sim 3X times faster.
Is it common that when simulating with ICE Lagoa, people only see a 30% load on their CPU?
I went through your comprehensive list and had a look at them all. Appreciate it!
Although this does not address the issue of ICE Lagoa not beeing able to utilize my CPUs full potential. It runs at about 30%.
My simulations should be able to sim 3X times faster.
Is it common that when simulating with ICE Lagoa, people only see a 30% load on their CPU?
Re: Utilize full CPU potential
the only thing i recall that can be 100% parallelized is a fully unbiased rendering (after initialization stages) . The rest of cg stuff efficiency should be never measured by cpu/gpu load.
Considering lagoa, it has a lot of built-in ice nodes that are just not suited for this kind of calculations (like get_closest) and overall particle cloud primitive type in xsi is not something than can carry heavy sims. Imho it was more marketing stuff from AD rather something usable and comparable to realflow\houdini
Considering lagoa, it has a lot of built-in ice nodes that are just not suited for this kind of calculations (like get_closest) and overall particle cloud primitive type in xsi is not something than can carry heavy sims. Imho it was more marketing stuff from AD rather something usable and comparable to realflow\houdini
Re: Utilize full CPU potential
By the way, Maya parallel is able to show around 80-90% of all CPUs during playback, and good bit of GPU as well. It scales almost linearly, on 8 cores is almost two times faster than on 4. Of course I don't know what under the hood belongs to that 80-90%, anyway, Maya playback speed is what I'd expect from these numbers.
Re: Utilize full CPU potential
Thank You both Mr.Core and Mathaeus.
When I run Maya parallell (example Nparticles with springs), I usually only get a 30 percent load also.
So I guess theres no way around it.
When I run Maya parallell (example Nparticles with springs), I usually only get a 30 percent load also.
So I guess theres no way around it.
Re: Utilize full CPU potential
AFAIK Maya parallel is all about SRT, transform nodes, rigs and such, don't believe it has anything with nParticles. Anyway I'm afraid we are a bit off topic with this parallel thing
Who is online
Users browsing this forum: No registered users and 69 guests