Question about IRAY Rendering, Camera, Texture, & GPU?
zetsu6040_069bdd73da
Posts: 21
greetings again, DAZ community!
here are my questions guys:
1) does anyone know how IRAY computes the texture that it will use for rendering?
a) does it only compute those that are being captured by the camera (or the Perspective View by default)?
b) or does it include in its calculations all the texture that is present in a scene, regarless if it is visible through the camera or not?
i'm curious coz i noticed that my DAZ has been rendering at a slower rate after my Windows 10 upgraded to version 1703.
i am comparing a scene with almost the same setting.
and based on the size of the saved file, i'm actually having smaller file sizes recently.
but for some reason it is taking more time to render the new scenes.
so i'm wondering if the texture of those characters have an impact in the calculation of what is to be rendered, eventhough they are not captured by the camera.
it's just a Genesis 2 character by the way.
2) my GPU isn't powerful enough, so i'm using a combination of GPU and CPU always.
and the condition is the same as in #1; the same setting with almost the same content (textures that are present), just different in terms of what is being captured in the camera for rendering.
what i noticed is that there are times when my GPU would work, and that basically adds up or even doubles the percentage of what is being rendered (based on the number of iterations on the activity log file).
but most of the time, my GPU just won't work.
and the log file says that it wasn't able to allocate the required memory.
i usually do my renders upon turning on my computer, so i assume that my GPU is not that busy yet during this time.
a) so does anybody have an idea why there are times that the GPU can handle a required amount of memory, and why there are times when it can't?
b) and is there a way to ensure that the GPU would always work together with the CPU?
unfortunately, i was not making use of the activity log file before my Windows 10 upgraded to version 1703, so i can't tell if that has affected the way that my Nvidia GPU works.
Daz 3D is part of 
Connect
DAZ Productions, Inc.
224 S 200 W, Suite #250
Salt Lake City, UT 84101
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
If you haven't, get the latest driver directly from NVIDIA. The driver that Microsoft ships with the OS isn't always the most efficient.
Textures have to be loaded because even if an object is outside the view of the camera it could still be seen in a reflection or contribute to the scene in other ways (translucent and a light shining through it, it having an emission texture casting light into the path of the camera, etc).
The size of the saved file / scene has little to do with the required VRAM for rendering.
There is a way to render only what the camera is seeing, which is to use iray section planes to hide everything outside of the camera's general field of view. Snow Sultan made a really good tutorial and included a ready-made 3-sided plane that's for that purpose on his YOuTube channel.
Alternatively, V3Digitimes has a utility for sale here that cuts down on the size of textures https://www.daz3d.com/scene-optimizer
yes i'm considering this, but i am thinking of when, since my Windows 10 itself is also trying to download a NVIDIA update on its own.
this is a very good point. because of the lighting factor, shadows, and reflective properties of different materials, i understand now why the machine always need to consider the full scene before working on what is only shown in the camera.
thank you for reminding me about this. :D
yes, i just realized it yesterday.
i compared a setting which is utilizing the same number of objects (and textures of course), with the same file size.
i did 2 renders with different poses for the characters and different camera angles, and the render speed and conversion rate was totally incomparable.
so i guess the render speed is also dependent on the direction of the lighting and the properties of the objects that are being captured by the camera.
but in a way, the file size can still give the user an idea of how much objects and properties are being called by a particular scene once rendering starts.
great tip, i'll probably study this one.
thank you! :)
oh here's another question regarding the camera.
particularly about using the focus function or effect.
does it help in speeding up the rendering process, if you just focus the camera on the characters, and let the background be blurred??
There should be no "considering" about it. The driver that Microsoft publishes through Windows Update does not work with Iray. To use your Nvidia GPU in Iray rendering you need to be running the Nvidia published version of the drivers.
I haven't tested this in Iray, but I would expect a blurred background to be slower rather than faster.
well, the Windows Update said it's a NVIDIA driver update, so i allowed it.
and it actually matches a version from the official site.
after that, i downloaded the latest release from the GeForce website to override the update from Windows.
i'm not sure how the updates from different entities (like Windows and NVIDIA) actually work, but i did that to avoid Windows from forcing unnecessary updates to my driver.
hmmm?
i guess testing is the best way to learn and find out.
but is there anyone else who has already done a lot of tests about this??
just so i can have more reference.
@zetsu6040_069bdd73da "but is there anyone else who has already done a lot of tests about this?? just so i can have more reference."
There are enough variables that affect render speed that other people's testing may have little applicability to your results.
Such as specific graphics card, scene specifics, lighting set and setup, render settings etc.
sorry for the confusion.
i mean tests using the same scene of course, and just applying depth of field and different focal distances.
because even if users have different GPU or CPU power, i think that can still give an idea of what renders faster.
anyway, i did one test using the same scene.
the original render was plain, and the second utilizes the depth of field to focus on the characters and blur the background.
i set the time limit for both renders to see which has a better performance.
as for the number of iterations, the 2 renders are almost at the same range after it reached the time limit.
but as for the convergence ratio, the one utilizing the depth of field had a little higher rate.
not that conclusive, but that was for my initial result.