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
© 2026 Daz Productions Inc. All Rights Reserved.
Comments
just went to the octane site v3 is there btw looking at the faqs while it says sli is no good it does mention having more that one card with sli disabled will improve speed strongly hinting at titans no mention of 1080/70s whether they update their info to include them when released who knows but they do compare cards slightly in performace quality again strongly leaning to titans
might try demo sometime see what its like some of my favorite artists here alternate between the different renderers and do amazing work with all of them out of the 3 I've got including 3delight and iray of course I'm more towards iray and buying content for iray like mec4d's shaders which are great so I kinda don't wanna go into another render app if I can't use them but I'd like to get the most optimised effect usuage out of them those promos are beautiful so whatever cards help to get the highest quality both detail as well as big complex scenes + speed I'd like the 1080s may have speed advantage and price but it's ability to render high quality detailed large complex scenes if they can't do that well well guess titans best from what few you saying
My apologies to Otoy - I should have checked before posting. Yes, indeed, V3 of Octane is out now but, if I'm not mistaken (again), the price has jumped considerably. Octane 3 including the DAZ Studio plugin is now $579 yet my somewhat suspect memory recalls a figure of about $100 less than that for the previous version. Seems to have shifted from the price of a GTX 1070 to that of a 1080?
Have just checked the GTX range pricing and I'm still confused. I live in the UK and NVidia seem to think that they can think up a dollar number and use the same (or close) number for GB Pounds. I probably shouldn't be looking at all - can't afford it anyhow.
I am curious to know how you are determing the size of the GPU memory you would need for your particular scenes. I assume you are not simply looking at the memory used by DAZ during rendering, as this is vastly more than the GPU needs. I have many scenes where DAZ needs around 10GB, but they fit comfortably inside 4GB GPU during rendering.
You still looking on the speed of the cards but the truth is total different
90% the users don't even use 100% of the GPU when rendering , Full load on GPU don;t mean the software is using all GPU
so I made for you diagram so you can see the difference in speed and optimal GPU usage in iray according to the rendering settings .
It is not about the speed but about how well the GPU is used and how well the light paths are rendered for photo real accuracy and why M6000 make so difference in quality of rendering vs GTX , and not talking about cleaning noises ...
Yes a lot better actually .. tonight I am going to test my 3 cards on optimal power with my new motherboard to see exactly the difference
but you have here 3 choices, you want to fast clean the noises or you want to use the full power of GPU you have what will slow down the rendering but the quality will be best at the end . It seems that the optimization activate the full potential of the GPU .
Also the benchmarks was done before the last patch in iray software so can be ignored but will test with 3 cards and see the full story more clear and see exactly how much GPU is used by iray from each card while rendering on the new rig tonight , just waiting for my new fans for CPU radiator and some cable to arrive
so glad didn't rush and buy something regret this is helping me a lot nailing down the setup want need afford which is something I would've done easily with my lack of understanding these things if it wasn't for these threads
Most of a scenes memory footprint comes from textures. And while high resolution textures are great for close-ups and portraits, the remaining elements in the scene RARELY need such high-resolution textures. Especially when you consider the memory limitations for GPU rendering on Iray. Basic rule is this: IF the on-render final size of the item is fewer pixels (on width and/or height) chances are you are wasting memory on pixels you will never see. If you are rendering at 4k x 2k resolution, and the figure is fully on the screen standing, the body texture only needs 2k x 2k resolution. The head texture probably only needs 1k x 1k. Eyes? 512 x 512 at best. Doing a portrait at that same res? Face can't be more than 2k tall., chances are it's a little smaller. 2k x 2k textures are probably fine. Especially if you're going to downsample later.
Now, Pascal has even better on-VRAM texture compression algorithms available (once the drivers are updated) so that 8GB of VRAM will go even farther than before. But take the time to resize your textures if it is clear the objects on the screen won't need that resolution. A single character in the background, taking up less than a quarter of the screen, with clothing, using 4k x 4k textures, is wasting on the order of 100MB - 300MB of VRAM on textures alone, maybe more. Props can have the same issue. In a complex scene with multiple characters and props.....you could be wasting easily HALF your VRAM on textures that are WAY too big for what appears on screen. While having that detail is nice when doing close-ups of a face, or a prop, or anything....when they AREN'T filling the image, they're wasting VRAM.
I recommend picking up something like ImageMagick, that has batch resizing capability. Just navigate to the folder, and run the resize batch. Make half-size and quarter-size versions. Then use those when you don't need that full gigantic texture.....just base it on the size of the item in the render (roughly) in pixels. Doing that, you'll probably find a lot of those scenes that were taking 6GB - 8GB or more VRAM take a lot less, with no loss of quality in the resulting render. And if you see some quality issues with specific items in a given scene, you can revert those to higher-resolution textures. Like if a portrait also shows the characters hand....the hand uses only a small portion of the whole body texture map, so using full resolution would make sense in that case. But not for the bracelet that's only about 1000 pixels wide and 200 tall on the final render.....
I know it's a lot easier to just click the preset and not bother with it. But then you're going to run into those VRAM limits.
(I still think the PAs should be including multiple resolutions of their textures, with presets for them, with their products. They say it would add a bunch of time and effort to the product, but I just can't see it. It isn't hard or time-consuming to resize an image, or to make a copy of an existing preset and edit the .duf file to use the lower-res textures and then rename the new preset.......should take less than 3 minutes a preset. 40 presets? That's less than two-hours. And with batch-processing, it would be even faster.....)
Would it not be better if iRay itself performed these optimizations? It knows how far certain objects are from the camera, and could downsize their textures before sending to VRAM. I recall one of the DAZ guys saying that 3DL does this kind of optimization, but iRay did something different (I do not recall what). I am certain that some form of work is going on, as a number of my scenes are close to the max capacity of the VRAM of my card, which I suspect is not a coincidence, as clearly the renderer knows how much memory it has and will try not to exceed that if it can.
It was actually in the Intel rigs thread I said that, but:
http://www.tomshardware.com/news/ekwb-gtx-1080-water-block,31857.html
Didn't take long. Of course, there's nothing unusual in seeing aftermarket people making water blocks for new graphics cards, but it's amusing that this one is for buyers who want to remove the expensive collectors shroud they've paid extra for. And when their shiny new card throttles down after four minutes of gaming, they will want to remove it.
Isn't IRay performing some optimisation with texture compression, as described in this thread? https://www.daz3d.com/forums/discussion/74718/iray-texture-compression-and-performance
If so, how does this differ from manually reducing the texture image size?
More control. Especially in mixed scenes. It's not based on distance from the camera, but size...so it will affect ALL the textures in a scene, while manually doing them, you can leave the close up elements large.
3Delight, optimizes textures by creating mipmapped tif images...and then uses the 'correct' one based on distance. It would be nice if Iray did something similar.
Thanks - further question though: how do I activate texture compression in IRay? I see (in Advanced) two Threshold values - one for High and one for Medium but how do I tell IRay to use High or Medium or any other kind of compression, for that matter? I don't see an "Enable" switch or such like.
@marble
Nvidia isn't the only vender that ignores exchange rates between the US and UK. Seems like a fairly common practice.
I recognize that's Not much consolation if you live in the UK.
Iray's 'compression' settings are somewhat confusing. The setting is basically telling iray at what size image to use what level of compression.
If the size is equal to or larger than the medium setting, but less than the high setting, it will use medium compression.
If the size is equal to or larger than the high setting, it will use high compression.
if the size is less than the medium setting, it won't compress the image at all.
The compression used in Iray is how the image is stored internally in VRAM, and decompressing images during rendering does take additional time (compared to non-compressed images.) But it's a fairly small hit to performance. Pascal chips actually have a few new ways to compress images that get a lot more compression than what is in Maxwell chips. But not sure if the Iray code has been updated to use them yet. But they'll give Pascal-based cards even more memory to work with, when they're working..
But even compressing the images doesn't change that too many 4k x 4k images is going to eat up your VRAM fast. And as @mjc1016 said, it's a global setting. But even a high-compression 4k x 4k image is going to take up more VRAM than an medium-compressed 1k x 1k....And it'll take a LOT more than a high-compressed 1k x 1k.
So if you DO downsize your images, make sure to update your compression thresholds so they still get compressed. Otherwise, you may not see nearly as much memory savings as you should.
Edit to add: One other note.....when the object gets smaller on screen, very high-rez bump maps and normal maps may (if they use very small details) become mostly useless. Downsizing them usually just results in a smaller image without details, or nearly invisible details. When it gets to that point, chances are you are better just not using the image map at all. Saves even more memory. These have to be evaluated on a case-by-case basis, though......which is why it would be better if the PA's took care of this for us.....they'll know which detail maps will lose too much when resized, and which will still be needed for larger details......
Many thanks for that explanation. I think I get it:
1. Reduce the image sizes manually before rendering (unless close-ups are required).
2. Reduce the thresholds in the Iray Advanced settings.
I guess that reducing both by half would be too simple, right?
Not at all. It is fairly simple. Just a bit time-consuming. But not bad. If the original textures are 4k x 4k, make 2k x 2k versions, save them with NEW names (don't overwrite the original 4k x 4k images!) Then load up the item in your scene, switch it to the new, smaller image maps. And lastly, update your compression settings the same way. You can also take it another step, and resize each step by half again (2k to 1k). That way you have settings/images you can use for close/large, medium/average, and small/far items in the scene.
Once you have them set up on an item, you can save a new 'preset' for that item, just add the text '_2k' or '_1k' on the end, so you know which size it uses. That way you can reuse it later too without having to mess around with changing all the maps......
Indeed, but is there an opportunity here for an enterprising PA? If Studio also knows the distances of objects (from camera) in a scene, and that information is available, could a script be written to choose appropriate textures? Then Iray just does its own thing regardless.
That's actually pretty standard in a lot of industries. It has a little to do with UK/EU taxes and a lot to do with 500 year old policies of granting companies state-sanctioned monopolies on distribution. The exchange rate between US dollars and UK pounds may be sitting at 1.44, but a Nikoin D500 costs $1,999 in New York, and £1,729 in London, an "exchange rate" of just 1.16. What happened to the other 0.3? Instead of "pounds sterling", cameras are sold in the UK in "pounds Nikon", 25% more expensive.
That's what I do...saves a lot of time later. It may take a bit at first, making the new presets, but later on, there's an immense return on that investment.
Except that it's not simple, at all. If there's a component of the lighting setup behind an object, then a component of its shadow stretches toward the camera. So, as you back-project rays from the camera to distant scene objects and then back to the lights, you find that a particular object may have shadow components 2x, 10x, or more, larger than the object itself, and therefore, it needs high resolution textures.
If the object is semi-transparent, this effect is even greator.
A distant object may have a magnified reflection in a convex curved surface (eye, drop of water, car fender, etc) that also blows up the resolution required. You don't know what the resolution needed is until the renderer computes a coupled passes of forward and backward ray tracing. By then, it's far too late for any simple vendor-supplied script to fix the problem. You need to do it in the inner workings of the renderer.
OK, technically, that last step insn't "necesary", but dang, it would be great to get something like that instead of just a simple "out of memory" errror.
Textures don't matter for shadows. Normal and bump maps do NOT add detail to shadows.
While true that reflections may matter, it is only to a point, though...the effect of a lower resolution texture would be the same as blurring the reflections...so, not necessarily a bad thing. And short of doing accurate light study/architectural renders (which Iray really isn't designed for anyaway) super accurate/crisp reflections aren't all that important.
Aw shee-it ... just when I thought I understood what we were talking about.
It boils down to this...if you need absolutely, spot on, accurate reflections...you can't downsize the textures, because they will no longer be as accurate for reflections. Shadows don't matter...
Displacement maps do.
Some of us like displacement maps, a lot.
I find it to frequently be a bad thing. Curved relective surfaces frequently magnify reflections, even convex surfaces over other convex surfaces. Curved transparent objects also magnify things you're looking through, and I have a penchat for glass vases, spheres, bowls.
You hear that "thud, thud, thud" sound?
That's M. C. Escher turning over in his grave.
It is simple. What I was describing was setting up OPTIONAL lower-resolution maps and presets. Not describing any automated system by which it would be decided on whether or not to use them, or which level to use. That's an ARTIST'S DECISION. The artist knows what size render they are making, they know which items are where in the scene relative to the camera. THEY decide 'hey, this figure way in the back isn't very large on the render, maybe those 4k x 4k image maps aren't really helping it.....'
Options are a good thing. Knowing when to use them is good too, but takes a little experience or at least some thinking.
As @mjc1016 said, bump, normal, and texture(diffuse) maps aren't going to affect shadows. Only opacity and displacement. THOSE maps you may not want to use lower-resolution. Depends on the scene, but you can probably make a good guess as to whether or not any of its projected influence is going to cover enough of the render to warrant a large map resolution.
Magnification effects (whether from refraction or reflection) once again vary greatly from scene to scene. This isn't an automated thing, like texture compression is. This is something the ARTIST does, knowing what his scene and camera are set up for. Having those reduced resolution textures is a benefit. USING them is a matter of experience. Typically, I'd use the lower resolution textures, do a test render, and see if anything warranted higher resolution due to reflections or refracted effects.
Trying to have something like this 'automated' by the renderer is a major headache, as the decision-making algorithm basically has to render the scene and decide if the output shows 'blockiness' based on some threshold. But if it has to render a test frame anyway, you might as well do it yourself, as the human eye is going to be able to detect the 'blockiness' much more cleanly and quickly than anything we have now.
...I'm looking at rendering for creating gallery quality/sized prints, like upwards of 24" x 32". At that resolution, high quality textures even in the background elements is very important.
I'm with you there marble. I've lurked around threads like this for a while with a confused look on my face. It's starting to pay off though and I'm beginning to be able to follow a lot of it. At least on some subjects. I've learned some good stuff from this thread.
from reports so far for Australians so far a few retailers are pricing the 1080 at $1299
well now this is interesting while on google looking up any new info on the 1080 in Australia at the side noticed an ad from one online aussie store scorptec offering a 12gb superclocked titan x for $1599 not much cheaper than the 1080s pricing so far ok $300 is a lot to many but not the huge savings we were told but then that's the aus $ and gst taxes for you