OptiX???
kwannie
Posts: 870
in The Commons
What happen to OptiX in DS 4.12. I was doing a render after installing 4.12 and noticed it was sluggish which I usually was able to resolve by checking the OptiX box but discovered it is no longer there in DS 4.12.
Why did they remove it?
Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2025 Daz Productions Inc. All Rights Reserved.
Comments
I noticed stuff in the log file about it not being needed or used as of 4.11 so I dont think think Iray uses it anymore, so it was removed.
Matty, Do you know if 4.12 requires a certain GPU series threshold for the same acceleration OptiX yeilded, for example 2080 or above?
OptiX Prime acceleration is always active in the version of Iray that ships with 4.12 so long as you are rendering with a non-RTX GPU (RTX cards switch OptiX Prime acceleration out entirely for hardware based RTCore acceleration since that is much faster.) Hence the removal of that checkbox - there is no longer any non OptiX Prime/RTCore solution for the renderer to fall back on. For better or worse.
Oh, so that's probably why my renders have been so incredibly slow to start up since 4.12 came out, I have a 1080ti and sometimes it'll take a full half hour before a single iteration. I remember turning it off on 4.10 and my machine started rendering much faster, I didn't even think of looking at it with 4.12, I was thinking it was a windows or Nvidia driver issue.
Hm. That doesn't sound normal. My 1080ti renders as great as usual with 4.12. No 30-minute wait time here.
Something is definitely wrong. My 1080ti starts about as fast as before and rendering in 4.12 is much faster.
Yeah, that doesn't sound normal at all to me either. If you check the last column ("Loading Time") of this table, that is the range of time (<10 seconds) other 1080Ti owners have been reporting it taking their GPUs to fully initialize a scene. Obviously the relatively simple benchmarking scene these numbers come from is going to take a lot less time to load than a fully formed production scene. But ~30 minutes seems awfully suspect. Mind posting a sample log file from after one of your 30 minute loading time scenes is finished? Chances are that there'll be something in there of use for enhancing performance.
Oh dangit, I was hoping that this was the reason lol. Yeah, the next time I do a render and it does this, I'll post the log here. I might not be able to do anything the next day or two, but it'll happen the next time I render for sure. Thank you all for commenting and any help you can give in the near future!
I have a 980 TI and 4.12 renders, on average, 20% faster than 4.11.
RayDAnt, so if I'm tracking what you are saying is that DS 4.12 is designed to used with the RTX cards primarily, which do not need OptiX. In DS 4.12 OptiX is always on even if it may ultimately serve to to slow renders down. I actually have a just a regular 1080 and 32 gig of ram so I don't know if having no option to turn OptiX on or off hurts or helps. I have not seen a huge slowdown but still no increase in speed either.
From what I see, the main issue with Optix is that it grabs more VRAM. I haven't noticed 4.12 being any slower than 4.11 - if anything it may be slightly faster.
I still have the option in my 4.12, is this the public beta that removed it? I really hope that decision is reconsidered as Optix does increase VRAM usage and can cause renders to drop to CPU that could have rendered just fine without optix.
In that case Nvidia would have to reconsider, because they built it into Iray ... it's not Daz3d's fault.
As for 4.12 optix prime is used for gtx cards, while optix is used for rtx cards, and embree is used for cpu rendering.
As for speed 4.12 is faster but it requires more vram on gtx cards because of optix prime always on. That means that a scene that was barely fitting 4.11 will not fit 4.12 and will fall back to cpu, thus the slower render. Unless there are driver issues of course.
So I don't want to beat the same topic to death, but out of curiousity have much difference can be realized between the various GPU cards above a 1080? I have a 1080 not a 1080 ti so I have 8 Gig on my GPU not 11 Gig. how much difference can that extra 3 Gigs make when I cannot get an envirement (such as a club or resturant) and 2 characters in a scene without exceding the 8 Gigs. It seems like the only thing DAZ can handle is maybe up to 3 charcters and an HDRi in a scene at a time. Placing props or vegitation in a scene and you are bust. Why wouldn't it be an advantage then to use the work stion cards that have up to 24 Gigs on board?
The option is still there but the checkbox does nothing.
Based on testing so far, the rendering performance increase for the highest level consumer friendly Nvidia card out today (the Titan RTX) over a 1080 is approximately 300% - not counting RTX acceleration (which has so far been tested to bump that up to as much as 3600% depending on the specific scene.) So rendering speeds somewhere in the 3-36x range is what you can currently expect to see from cards above the 1080.
Something to keep in mind is that memory usage in Iray doesn't always scale linearly with the amount of content in a scene due to built-in features like texture consolidation and object geometry instancing. So although going from 8 to 11 gigs might not seem like all that much on paper, in practice it should net you a lot more flexibility in scene composition.
It absolutely is an advantage to those cards. Hence why I went and bought one (a Titan RTX) myself just a little less than a year ago now. And have yet to regret it.
That's why you need the scene optimizer addon.
1080ti
G8males
some scene optimisation
On mine I can pretty much guarantee a render will drop to CPU if check the box. I've tested a scene from a fresh restart of Studio, Optix off, renders GPU. Restart again, check the box, renders on CPU.
Check your log file. You will see that Iray is using OptiX Prime acceleration in both cases. You can further prove this by doing a test render and seeing how the device rendering times (also in the log file) are within margin of error of each other regardless of the checkbox's status. If OptiX Prime were actually being switched on/off as expected, there would be a significant time difference.
A question that keeps nagging at me: does it make a difference to VRAM when the figures are either smaller or further away from the camera? It certainly makes a difference to the render time - the more skin filling the frame, the longer the render time. The only way I can think of that it might make a difference is if IRay somehow further compresses the texture files for non-close-ups.
AFAIK iray doesn't support mipmaps nor adaptive subdivision so there's no difference as far as distance from the camera is concerned. As for rendering time, that heavily depends on the shaders being involved. So for example an advanced skin shader taking most of the frame will slow down rendering.
As a side note 3Delight supports both mipmapping and adaptive subdivision so it's better optimized for large scenes than iray.
If I am not doing a closeup portrait type render I usually reduce the 4k+ skin maps etc to 2k, to 1k or 512 for backround filler type people. Also adjusting the subd down too sometimes, no point keeping details you can't really see anyways. I am fairly sure nvidia keeps their renderer unoptimized on purpose hoping people will buy their crazy expensive cards lol.
When skin, and more importantly hair, is close to the camera the render gets more complicated and therefore takes longer. Amount of VRAM consumed seems to have little to do with how long a render takes.
My best guess is that on skin that is well away from the camera the various details of the skin, normal maps, subsurface scattering etc., are optimized away. Something similiar seems to happen with hair as well. You can establish this for yourself. Create a scene with 2 cameras. Put one camera to take a portrait shot of a character and then place the second farther away to get a full body shot. The full body image should render faster.
It doesn't seem Iray is 'smart' enough to compress textures further from the camera. Interestingly Iray for iclone gives users some more options for selecting exactly what gets compressed more. So you can manually select an item away from the camera to be more compressed and save memory. It would be cool if Daz Studio could adopt this.
That's interesting and probably the kind of thing I was trying to get at with my question.As you say, if it is possible in IRay, I hope DAZ get around to implementing it.