View Full Version : Speeding up translation time
10-26-2011, 08:59 PM
I am curious, are there some basic things that I can do to speed up translation time? Translation time took 24min while rendering took 2min for this scene and I am only using a fraction of the particles that I will need to use by the end of the project.
I am currently running on a Windows 7 64 bit computer using an Intel 7 Quad Core multi-threading 2.6ghz processor, and 6ghz of memory.
Not sure if anyone can help but I sure would appreciate some tips on this one.
10-27-2011, 12:23 AM
Maybe use a lower poly count particle instance? Nt sure though what are you doing?
10-27-2011, 01:22 AM
Sorry for not being more descriptive.
I was working with 6 particle systems that throw out stone down a shute and rendering performance was fine with these even thought they spit out about 100 particles per frame.
When I added a water system I created before to tweak, it is a super spray using meta particles for blobing and then using a water promaterial for realism, the render slowed down drastically. This system is spitting out particles at about 60 per frame and the size is about 2. No matter what I change for this system the translation period of the render takes forever.
I will see about posting a scene to show yall tomorrow when I get a chance.
10-27-2011, 02:01 AM
Its just strange. I try to jump to frame 377 in the animation and it takes 3ds Max about 2minutes to jump there but the CPU usage never spikes above 12% and memory is steady at 2.72GB.... Its like 3ds Max isn't using as much of my CPU as it should?
10-27-2011, 02:33 AM
Blob meshing particles is super heavy and only single threaded, plus your are using a single threaded particle system, which is why you are only seeing a 12% utilization. You can speed this process up significantly but it is going to cost you some coin.
10-27-2011, 10:36 AM
if vray is the render engine...the bercon metaball is free but i never used it.
10-27-2011, 11:20 AM
Well I am actually using Mental Ray as I cannot afford Vray currently. It took my compy 2hrs to render a 320x240 image and the lowest mr settings with the current metaball particle systems I have going on.
Any thoughts on if it is possible to change the system to use hyperthreading? Otherwise I guess I will need to use a different particle system that does support it or something that just isn't as processor intense.
10-27-2011, 03:33 PM
Does 3ds Max 2012 support hyper-threading or multi-threading for particle systems?
10-28-2011, 03:17 AM
I hate to say it but not a chance, at least not anytime soon anyway.
Super Spray is pretty lean and efficient I believe it to be the meta meshing that is your major crippler. It has been a while since I have used superSpray but i am pretty sure you can set it to use no shape and add a blobmesh object and see if you get any performance change. Also try a clay render, ie a standard grey material instead of your custom water shader.
The thing to do is identify exactly what is causing the major part of the slowdown and try and optimize it.
You can give a visit to MaxPlugins.de there are a few free meta meshers there you can try too, if it does indeed seem to be the root of the slowdown.
10-28-2011, 04:45 AM
Well I have learned a lot today. It seems part of my problem occurs due to using multiple separate objects for collision. I combined them into one object and saved some time. Also I played around with turning off hyper-threading so there were only 4 cores instead of 8 and here are some interesting results.
Without Hyper-threading frame 277?
Render time: 21sec
Translation time: 6:32
With Hyper-threading frame 277
Render Time: 16sec
Translation Time: 5:44
With Hyper-threading and single deflector object frame 277
Render Time: 10sec
Translation Time: 3:21
So it seems the hyper-threading is still a tad faster (assuming I did render 277 without hyper as I am not positive).
I still could use some help optimizing the scene if anyone is willing to give suggestions as 3:21 is way too long for my timeframe to render 780 frames. I have attached two photos, one showing the scene and the other showing a glimpse of my particle view. Let me explain a bit about what is going on in the scene.
Basically there are 4 different types of rocks being dispersed throughout the chutes displayed. There are non wanted rocks and wanted rocks which separate by a process. To show this I created 2 lines per chute to make the rocks follow with a follow icon that is constrained to the path. There is one particle system per chute so a total of 9 particle systems that each create about 1500 particles throughout the lifespan of the scene and are rendered out as 4 different types of rocks.
Then there are 9 supersprays with instanced geometry, a pro-material for water, and collisions to bounce off of the surfaces of the chutes.
Hope this explains a bit of what is going on. If anyone knows a way to optimize the scene or a better way to build the particle systems I would be grateful.
10-28-2011, 10:12 AM
I looked at your file but I'm not the right person for this type of problems, however:
I think there are more particle systems (and redundant operators) and sprays and this is the first cause of a slow down.
a second bad thing for 3ds max is the omniflect deflector.
third problem is the blobmesh compound object you need to mesh the particles...it is a dangerous trap!
If it is fine for you, I solved at my manner some questions but for meshing particles you need a good plugin (pwrapper or, better, frost or free pieces of code from max users).
I hope really this can help you at least a bit.
10-28-2011, 12:25 PM
I thought the amount of systems might have been the issue but I really am not sure how to condense them. What you see in my file is actually my "attempt" to condense them where I saw it might be possible but I don't think I did a good job.
I was unable to open your file. 3ds Max gave the following error. "File Open Failed: File Location". I am currently using 3ds Max 2010 if that makes any difference (If your using a newer version perhaps?).
Thanks for trying to help with this!
10-28-2011, 12:57 PM
ok, try again...now is a max 2010 format.
I don't know if my method symplify eccessively your flow but it can give you an idea about a more light solution.
10-28-2011, 01:36 PM
All I can say is that you rock! That is pretty ingenous how you handled spawning the particles at multiple locations. It definitely makes the system a whole lot smaller and I feel like an idiot now for doing it the way I did it. But live and learn right? Thanks again for your superb help with this! And thanks to everyone else who were helping too.
Edit: Just out of curiosity, did the scale of things have anything to do with the fix as well Savat? I noticed the units are different than I was using and when I try to move the camera around to view different parts of the scene it jumps around like crazy like the objects are really small, though comparing them to the grid they seem bigger.
Edit Edit: Just out of curiosity, is there a way to make the water come out more like a wide spray along the Y axis (I think)?
10-28-2011, 03:42 PM
scale is different because, opening your file, I've converted it in the scale units I use...not worry, you need only to adapt to your scale.
regarding a more spray effect (speed operator) you can increase the speed divergence or use a random 3d direction (and decrease the speed value).
last thing, I've supposed you need an exact percentual of any of stones you've instanced (the 3 spilt operators) but you need to know that a single shape instance operator can select groups of objects and split them automatically.
look at this more compact solution and good luck :)
10-28-2011, 04:16 PM
Very cool. yeah I was trying to spawn like 25% overall particles as one type and then the rest as the others. As it turns out I actually had to change the one to 15% so the split works, but this is definitely a cool way to look at doing the rest as I could get rid of a few of the other splits and just group the other ones. Thanks again for your help! I am learning a ton about particles.
10-28-2011, 04:17 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.
vBulletin v3.0.5, Copyright ©2000-2013, Jelsoft Enterprises Ltd.