Hi everyone i am rendering in houdini and in my task manager it says that only 1.5gb is been used out of 4gb. How do i allow it to use all RAM available ?
Any help appreciated
Anyone Know how to use all RAM available
8182 15 3- Lyonz
- Member
- 114 posts
- Joined: Oct. 2009
- Offline
- keyframe
- Member
- 1532 posts
- Joined: July 2005
- Offline
ha ha. It's funny that you ask that since typically we profile the renders and optimize as best as we can to REDUCE the amount of RAM required to complete the render.
Generally speaking, you want your renders to occupy as LITTLE ram as possible – if you are looking for a ‘go faster at the expense of using more RAM’ slider, I don't think such a thing is available.
Best,
G
Generally speaking, you want your renders to occupy as LITTLE ram as possible – if you are looking for a ‘go faster at the expense of using more RAM’ slider, I don't think such a thing is available.
Best,
G
- symek
- Member
- 1390 posts
- Joined: July 2005
- Offline
- keyframe
- Member
- 1532 posts
- Joined: July 2005
- Offline
- pbowmar
- Member
- 7028 posts
- Joined: July 2005
- Offline
- jason_iversen
- Member
- 12527 posts
- Joined: July 2005
- Offline
If you're raytracing at all, you can add a Rendering Parameter called “KD Tree Memory Factor” and increase it.
This is a general mem-vs-speed tradeoff.
This is a general mem-vs-speed tradeoff.
Jason Iversen, Technology Supervisor & FX Pipeline/R+D Lead @ Weta FX
also, http://www.odforce.net [www.odforce.net]
also, http://www.odforce.net [www.odforce.net]
- pbowmar
- Member
- 7028 posts
- Joined: July 2005
- Offline
Shame they aren't documented:
Houdini name vm_octreestyle
IFD name rendererctreestyle
Default (1)
Tunes ray-tracing acceleration structures. The default is usually reasonable and there is typically no reason to change the value.
Houdini name vm_kdcrossover
IFD name renderer:kdcrossover
Default (100)
Tunes ray-tracing acceleration structures. The default is usually reasonable and there is typically no reason to change the value.
Houdini name vm_kdcostratio
IFD name renderer:kdcostratio
Default (1)
Tunes ray-tracing acceleration structures. The default is usually reasonable and there is typically no reason to change the value.
Houdini name vm_kdmemfactor
IFD name renderer:kdmemfactor
Default (1)
Tunes ray-tracing acceleration structures. The default is usually reasonable and there is typically no reason to change the value.
Houdini name vm_octreestyle
IFD name rendererctreestyle
Default (1)
Tunes ray-tracing acceleration structures. The default is usually reasonable and there is typically no reason to change the value.
Houdini name vm_kdcrossover
IFD name renderer:kdcrossover
Default (100)
Tunes ray-tracing acceleration structures. The default is usually reasonable and there is typically no reason to change the value.
Houdini name vm_kdcostratio
IFD name renderer:kdcostratio
Default (1)
Tunes ray-tracing acceleration structures. The default is usually reasonable and there is typically no reason to change the value.
Houdini name vm_kdmemfactor
IFD name renderer:kdmemfactor
Default (1)
Tunes ray-tracing acceleration structures. The default is usually reasonable and there is typically no reason to change the value.
- keyframe
- Member
- 1532 posts
- Joined: July 2005
- Offline
- mark
- Staff
- 2597 posts
- Joined: July 2005
- Offline
keyframe
And there you have it.
Learn something new everyday!
So, errrr, anyone have ANY experience tuning those settings that might be able to shed some light on what they do, and in what context we might want to tweak them?
G
The documentation I found said: “The default is usually reasonable and there is typically no reason to change the value.”
I don't think you want to tweak them.
- Lyonz
- Member
- 114 posts
- Joined: Oct. 2009
- Offline
- mark
- Staff
- 2597 posts
- Joined: July 2005
- Offline
- Lyonz
- Member
- 114 posts
- Joined: Oct. 2009
- Offline
- keyframe
- Member
- 1532 posts
- Joined: July 2005
- Offline
- malexander
- Staff
- 5164 posts
- Joined: July 2005
- Offline
Using more memory for a simple object could have the side effect of overflowing your CPU's L2 or L3 cache (which is only 2 to 12MB) - in which case computation will be stalled by continuous memory requests. In that case, you'd actually want to reduce the size of the constructed data structures for better performance. It really is a difficult balancing act between memory usage and computation.
- Lyonz
- Member
- 114 posts
- Joined: Oct. 2009
- Offline
- heydabop
- Member
- 203 posts
- Joined:
- Offline
keyframe
You are right. Bigger buckets will undoubtedly take up more RAM.
Wouldn't necessarily go much faster though.
G
I've noticed small, but visible, increases with speed on multi-core systems, as with bigger buckets, less time is spent on cores waiting for another spot to work on, and more time is spent on rendering. Although if the scene is really complex it won't matter much, and as you said, it will use more memory.
-
- Quick Links