Hello, a while back you added support for primitive "filenames" and "index" intrinsic attributes which meant that it was finally possible to time offset animated instances.
It works but i'm not sure that is work properly. Here you claim that the VRAM for one item should be about the same as for all instances, but i'm just not getting that.
VRAM usage is the same as when its unpacked BUT the loading times are there. It loads significantly faster than when it's unpacked.
Here you can see that a single item and instances have the same usage (2627mb)
Time offsetting yields the same as unpacked. (3565mb)
Am i missing something?
I've attached the file without the cashed geo since its larger than the maximum allowed size. You can cache it yourself.
Or you can download it from Here with the cached geo.
Instance Time Offset / Primitive intrinsic
Moderator: juanjgon
- Attachments
-
- animationOffset.zip
- (216.75 KiB) Downloaded 15 times
Win 10 64bit // GTX 4090 + GTX 3090 // 5900x // 64GB // SideFX Houdini // C4D
Hey galleon27!galleon27 wrote: Fri May 09, 2025 2:22 pm Hello, a while back you added support for primitive "filenames" and "index" intrinsic attributes which meant that it was finally possible to time offset animated instances.
It works but i'm not sure that is work properly. Here you claim that the VRAM for one item should be about the same as for all instances, but i'm just not getting that.
VRAM usage is the same as when its unpacked BUT the loading times are there. It loads significantly faster than when it's unpacked.
Here you can see that a single item and instances have the same usage (2627mb)
Screenshot 2025-05-09 155034.png
Screenshot 2025-05-09 155052.png
Time offsetting yields the same as unpacked. (3565mb)
Screenshot 2025-05-09 155114.png
Screenshot 2025-05-09 155253.png
Am i missing something?
I've attached the file without the cashed geo since its larger than the maximum allowed size. You can cache it yourself.
Or you can download it from Here with the cached geo.
Sorry for the delay in response!
So the thing is that octane only loads the geo of the CURRENT FRAME into the GPU, not the WHOLE frame range of the instances. So if every instance is at a different time offset there is no benefit to be had.
But as soon as multiple instances are on the same frame, then you get memory benefits.
Hope that makes sense!
Makes sense. Thanks for the info
Win 10 64bit // GTX 4090 + GTX 3090 // 5900x // 64GB // SideFX Houdini // C4D