[Bf-committers] Problem of Rendering Speed

Ruan Beihong ruanbeihong at gmail.com
Tue Aug 4 06:45:26 CEST 2009


I see. Maybe you should try 2.49 as others mentioned.

2009/8/4 Yaohua Xiong <xiongyaohua at gmail.com>:
> Hi Beihong,
>
> It's nice to see you too. The very reason behind considering using
> particle system is efficiency. What I'm trying to do is not visualize
> 2000 cars but 100000 cars. The 2000 cars problem can be solved using
> the first method(objects+ipo) with no problem, but the 100000 cars
> problem must use memory more efficiently, like creating cars on the
> fly when it's entering the road network ,recycling cars out of range,
> cache file swapping in/out memory, etc. To fulfill these needs, use
> the particle system should be easier than other way.
>
> As for the feasibility of the particle approach, I think it's not a
> problem. With jahka's latest add to particle system---external
> particle cache file---It's possible to control each particle's
> behavior once you have an external package to generate them. Current
> what is missing is how to give these particles some variation in
> appearance.
>
> On Tue, Aug 4, 2009 at 11:54 AM, Ruan Beihong<ruanbeihong at gmail.com> wrote:
>> Hi Yaohua,
>> It's good to see chinese (I'm one too) in the maillist. I'm an
>> undergraduate in Wenzhou Medical College.
>> I'm neither an artist nor a tech guy. I use blender mainly to edit
>> some DV clips.
>>
>> In my humble opinion, the particle system should not be used in this way.
>> Particles usually simulate something emits light and spreads randomly
>> on small scale but move regularly on large scale.
>> It's not proper either to control every particle in the system (I
>> wonder if it can be done even), nor to build thousands of
>> particle system in one scene. If I'm right (I'm not quite sure as I
>> hardly use particle system) you should give up the second
>> method.
>>
>> 2009/8/4 Yaohua Xiong <xiongyaohua at gmail.com>:
>>> Hi everyone
>>>
>>> I'm doing a traffic visualization project using blender
>>> 2.5svn(details:
>>> http://blenderartists.org/forum/showthread.php?t=161741&page=2) and
>>> having some problem with rendering speed.
>>>
>>> I represent the vehicles in blender using two methods.
>>>
>>> 1. every vehicle is represented as an individual object(sharing the
>>> same mesh) and use key frame move the vehilce
>>> 2. vehicles are represented as a particle system, one particle for
>>> each vehicle. The trajectory of particles are load in using external
>>> cache file. every particle render as an object.
>>>
>>> The basic scene is the same for these two methods(2011 vehicles, 600
>>> frame) but the rendering speed is very different. The rendering speed
>>> for the first method is about 30 seconds/frame which I think quite
>>> normal but the render speed for the second method is about 15
>>> minutes/frame, it's 3000% slower!
>>>
>>> The real strange thing is that about 14 minutes of the 15 minutes is
>>> spend on the "Filling octree" phase of render, but the number of
>>> objects should be the same for both methods(2011 vehicles), what makes
>>> the difference in render time and how this can be improved? Anyone got
>>> an idea? Thanks in advance.
>>>
>>> --
>>> Yaohua Xiong
>>>
>>> Traffic and Transportation Engineering College, Tongji Univ.
>>> Cao An Road 4800#, Jia Ding District, Shanghai,  China
>>> _______________________________________________
>>> Bf-committers mailing list
>>> Bf-committers at blender.org
>>> http://lists.blender.org/mailman/listinfo/bf-committers
>>>
>>
>>
>>
>> --
>> James Ruan
>> _______________________________________________
>> Bf-committers mailing list
>> Bf-committers at blender.org
>> http://lists.blender.org/mailman/listinfo/bf-committers
>>
>
>
>
> --
> Yaohua Xiong
>
> Traffic and Transportation Engineering College, Tongji Univ.
> Cao An Road 4800#, Jia Ding District, Shanghai,  China
> _______________________________________________
> Bf-committers mailing list
> Bf-committers at blender.org
> http://lists.blender.org/mailman/listinfo/bf-committers
>



-- 
James Ruan


More information about the Bf-committers mailing list