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
Thanks for the info Cath and linvanchene!
You welcome Kevin , iray getting better with each new release , over one year ago the GPU scale with GTX cards was horrible but that was not done on purpose by Nvidia it was just some buggy engine
it was like 100/50/30/20 with iray 2015 but now it is already 100/50/50/50 and sometimes 100/75/50/50 with 2016.1.2 and it may only improve in the future catching up with the other better unbiased render , the fact that is a very young engine explain a lot , if the GPU scaling was as bad as in iray 2015 early release and never improved I would never go for more than 2 cards myself as that would be a pretty waste of money , I almost regret purchasing 4th card , but with iray 2016.1.2 everything is back in business lol and going to be only better with 2016.2
I should also mention that Iray is quite a bit slower than other commercial renderers because Iray doesn't have path/ray splitting -- that's more Vray or MentalRay I believe? Logically, it does make sense for rays to *not* split at a surface but it does mean Iray can't do dimly-lit scenes or glossy materials as fast...
Well I was not about to split the hair in half here but iray is a path tracer like octane and vray is a ray tracer and that have nothing to do with GPU speed but the fast calculation thanks to split of the secondary rays when the primary rays hit the geometry surface , iray in this case send secondary primary rays . vray need more time until you can see the rendering preview result but once it is done the rendering finish much faster , when you render with iray you can see the cleaner image much faster allow you to edit the light in interactive mode (real time view port ) but then it take little longer to finish , also dimly-lit scenes need less samples (iterations ) with iray when Architectural and caustic samplers are on but it take longer in the beginning to finish if you are not using portal lights that will tell iray exactly where is the light source is other way it will shot rays blindly and increase the time of rendering and as with vray if you really want increase the speed (calculation ) you need to pay attention to your rendering settings, materials and light as nothing will do it all at its own , wrong scene settings and materials will slow down all calculation no matter iray or vray , iray is also progressive render that allow you to render unlimited time until you like what you see where vray is not .
So we have different type of rendering here , one have one thing better the other something else.. but on the list of fastest unbiased GPU rendering iray is just slightly below Octane with iray 2016.1.2 (still in DS Beta )
so what you would prefer , render with iray and see faster preview result that allow you to make quick light changes before your final render or wait longer for your preview render and realize it looks not right using vray?
what is better , faster iray workflow to set the scene and wait slightly longer for the final render or waste time with preview renders not knowing the final result ?
the answer is simple , first vray will render faster your final render but to get there you will spend more time in preparations and second we don't have vray in DS so , not usable for me
Just a quick note. eVGA has b-stock Titan-X Hybrids (Maxwell, not Pascal) in the online store for $629.99 US...... also 980Ti SC for $309.99 US.
If you want them, act fast. These will be gone quick!
(oh, and 970 SSC Gaming cards, b-stock on the same page for $169.99....)
Linky: http://www.evga.com/Products/ProductList.aspx?type=8
Hehe yes :) I agree with all of this.
P.S forgot one more thing .... mic drop
I wish they do it little early ..I would saved myself some cash .. nice price , I guess people goign to buy it and sell for more on ebay in the next days
@ GPU scaling
Thanks for clearing it up that GPU scaling has been resolved.
I still am under the impression that something got mixed up with the time improvments per GPU.
Because this is a common missconception I provide some examples to illustrate GPU number to render speed relationship and how much that speed improvement costs.
Note that the render time is not cut in half further if you add the 3rd or 4th card.
Instead the simplified formula for cards with the same cuda core count is
- - -
Total render time for the scene with multiple GPU = Time needed with one GPU / number of GPU installed
- - -
Lets have a look at some examples:
Assumptions:
- each GPU has exactly 1'000 cuda cores.
- one GPU costs 1000$.
- 1 GPU using 1000 cuda cores needs 60 minutes to render the scene.
- - -
1 GPU = 1000 Cuda cores
Time needed = 60 minutes = (60/1 = 60)
Total money spent on cards: 1'000$
- - -
2 GPU = 2000 Cuda cores
Time needed = 30 minutes (60/2 = 30)
Total money spent on cards: 2'000$
You spend an additional 1'000 $ to gain a further speed improvement of 30 minutes. (60-30 = 30)
- - -
3 GPU = 3000 Cuda cores
Time needed = 20 minutes (60/3 = 20)
Total money spent on cards: 3'000$
You spend an additional 1'000 $ to gain a further speed improvement of 10 minutes. (30-20 = 10)
- - -
4 GPU = 4000 Cuda cores
Time needed = 15 minutes (60/4 = 15)
Total money spent on cards: 4'000$
You spend an additional 1'000 $ to gain a further speed improvement of 5 minutes. (20-15 = 5)
- - -
To phrase this differently:
With each new card you add you will observe less gain in render speed.
You get the best render speed improvements for your money by just adding a second rendering card because only then you cut the render time in half.
The 3rd card you add will reduce render speed to a third of the original time of one card but you allready paid a relatively high price to gain a little more improvement.
- - -
Therefore a rig with a good render speed to money invested relation ship is:
1 display card (medium price range) or (low price range)
2 rendering cards (high price range) or ( medium price range)
- - -
Maybe from that perspective it is not that bad that Nvidia limits the number of Titan X per user to two...
- - -
I always say the same as you just did , for the money the best performance is to have 2 x cards and one for display , unless you don't care about the budget or run a business with heavy daily rendering
also having Optix ON with 2 or 3 cards will do not best with GPU scaling and the 3th card may be not even be used in the last General release so optix off for best GPU scale .. unless you run 4 cards , there was some issue with the GPU scaling when you used more than 2 devices on a small scenes .. I talk to Nvidia programmer one month ago about that as it was definitely something wrong , you expect things to get faster not slower with 12K cuda cores .. but now the problem is solved already so the next General release will have it all as long all your cards are from the same series with the same cuda cores count .. things will be different if one of your card have different architecture as the card required different driver to run proper and there can be conflict that twist the GPU scale the wrong way delaying the faster cards like for example running series 900 + 10 together or 700+900
I spent $4500 on my 4 cards however it did improved as well both rendering and viewport as bugs was fixed by Nvidia
I measure the speed with GPU monitor not counting sec in iray for better accuracy as it is always different and not accurate , sometimes the sec jump from 10 to 30 in one sec so that can't be trusted for any serious measurements , also the result will be different by each system , I build my rig specific for iray for the max power so everything is balanced between the GPU , CPU and RAM as that is also important and not just GPU as CPU is 50% involved in rendering process with just GPU
Well you can use 4 Pascal Titan X if you want soon .. no SLI no limitation with iray , however for games you really not need more than 2 Pascal Tiran X anymore at max
...thank you both linvanchene and MEC4D for the better clarification on this.
So let's just talk CPU mode for now as a new system to support the new GPU technology is out of the question for me. How does Octane CPU mode rendering speed compare with Iray in the following:
Direct Lighting vs. Interactive Mode?
Path Tracing vs. Photoreal mode?
Both using a 4 core (8 thread) non overclocked i7.
Not bothering with PMC as that apparently is for GPU rendering only.
Octane 3.1 will have CPU support according to the OTOY site. It is probably still in alpha. I quit looking for more info when I saw alpha.
https://home.otoy.com/octanerender-3-and-roadmap-update/
It has been a GPU based render engine for a few years now. The update page is the first time I've seen CPU render use mentioned. The use of system RAM to hold textures was a recent development.
That is correct .. I wonder if we will get this option ever with iray to store textures usign just RAM .. never say never
Then you go to spend money for the standalone and the plugin and it is not going to render fast with CPU.. not worthy , waste of money in this case , you could get yourself better a new graphic card or render with CPU in iray without additional costs
Yes, from the begining Octane has only used the GPU to render, it has been able to use out of core (GPU) RAM for textures since version 2. Otoy has devised a means to recompile their cuda code for Octane to run on non-Nvidia GPU's (AMD, Intel, etc.) as well as CPU's. They have made an announcement that this new technology will be included in Octane 3.1 sometime later this year. But as MEC4D already noted, CPU rendering with Octane may not be much different than CPU with Iray.
Just wondering, why can't you upgrade the graphics card in your current system? Since you have an i7, I would think you could easily upgrade to a GTX 960, a GTX 1060, or possibly even a GTX 1070 (the 10xx series cards are already usable with Octane). If it's because you only have PCIe 2.0, a 3.0 card will run fine, just at the slower bus speed. The only time the slower bus speed will have any sort of impact is during initial scene loading, but for rendering it shouldn't make any real difference at all.
...the i7 is first generation 930 (LGA 1366). The MB ( a P6T X58) has PCI 2.0, not 3.0 x 16 not sure if I can fit a dual width card in along with the current one (I would use the old GPU just to drive the displays). With only 12 GB of Physical memory I am not sure it would adequately support a high memory GPU like a 1070 or 1080 let alone a Maxwell Titan X if I could find one. I would at the very least need to swap out the MB to even get 16 GB (maximum memory supported by W7 Home Edition and recommended minimum for the 1070) as the one I have is a tri channel board with six, not four or eight slots. This in turn would also mean a new CPU (as LGA 1366 is obsolete) and memory kit. The recommended CPU is 3.2 GHz with 4GHz for optimal performance. . Mine is rated at 2.8. There is also a BIOS issue with older P6T boards. To upgrade the memory on the current MB, I would need a W7 Pro OEM.
Basically it would pretty much mean building a new system save for the PSU, drives, and case.
When I built the system I currently have, I took the time to research components to make sure there was 100% compatibility between them. The last thing I wanted was something going *poof* when I booted it up. For 3DL it is more than adequate (particularly with improvements to the render engine in 4.8). for Iray...mmmm....not so much. Were R4 not so buggy, I'd go back to Lux as at least I can render with the Daz application and scene closed to save memory resources even if it renders in geologic time.
@ CPU render modes
Otoy has a tendency to announce early where they want to put developer focus the following year. But this does not always mean that those features actually end up being integrated. So when thinking about using OctaneRender judge it by what is available now and not what may or may not come further down the line.
For now Octane is a (Nvidia) GPU only render engine.
I was also a bit surprised that a CPU render mode showed up on the 3.1 roadmap (again)
Most information you find about a CPU render mode on the Otoy forum was some one or two line remarks from developers.
- It actually seemed that they postponed that idea to further down the line together with out of core geometry.
--> CPU rendering may still come with Octane but based on the previous communications it is pretty unclear when and how.
- - -
@ render speed improvements
It is a long shot but maybe things will look completely differently in a year with new technology changing things up?
Imagination’s PowerVR Ray Tracing technology
-> Latest developments for OctaneRender 4 (!) seem to go in a competly different direction. They seem to have partnered up with Imagination to develop a new GPU hardware technology to work in tandem with OctaneRender 4 that can drastically speed up GPU rendering.
https://home.otoy.com/otoy-and-imagination-unveil-breakthrough-powervr-ray-tracing-platform/
- - -
with your rig you can run at max 1 video card with 6 GB vram .. for 8GB you need update your system Ram to at last 16GB it will still works fine with PCI 2 , iray do not consume the full power as with games , as my Titan X never get above 60% of power usage to max 70% so not even 200watts ... around 160-180 watts while rendering , to run 2 cards you need minimal 32GB , a simple shorter version of 1060 would improve your rendering time already so there is always solution
Kyoto Kid, you really should consider limiting what you render, and for bigger scenes... I know you don't like it... render separately then COMPOSITE! Bigger scenes with less cash spent on what Cath suggests, and you still get the final render you want.
You can always do a simple composite in Studio by rendering your background separately, and then render your characters interacting separately and then use the background as an actual background image and have your render of your characters as the foregound image. Then render out the two (or more) images as one which will go quickly.
For that things are iray canvases.. the best CGI productions are crested by rendering images in layers not as full scene , it use less resource ajd fast processing and accuracy , and all you have to do combat the all layers together in one full editable image , I render often in canvases without even setting up the light , I do it in raw camera PS or light room for faster workflow and perfect light without wasting time with tonemping and endless previews not to mention DOF and other settings like adjusting shadows etc.. it is a black Perl especially for those with less rendering power or time .
I use paintshop pro X7 is a lot like photoshop with pretty much all the same features and is way way cheaper and it's not a subscription thing so it's a great alternative to photoshop
GIMP also works beautifully as a free Photoshop alternative, and this is coming from a Paintshop Pro user myself... I also used Paint.net for many years, as the community developed a large number of plugins for it.
Combine that with the newly-free Nik Collection, and you have a pretty capable setup :)
Is this only for really big scenes? I have done one scene inside a room containing two characters with canvases and from my experience it rather uses more time/resources. One problem is the emissive surfaces render black when using canvases. Next when rendering the background with canvas I can't hide the characters in the scene (only in canvas) as otherwise there won't be any reflection of them on the surface and the light would act differently. So basically its the full scene to render with everything in it just not in the result image. To render the characters its the same I can't remove the background from the scene (only in canvas) as it will otherwise change how the light behaves. How is this supposed to reduce the resources needed if it needs all the elements for correct light behaviour?
...I stopped at X4 as PSP went to just a digital download after that. If I am paying money for software, I want a "hardcopy" as a backup. You have to pay extra now for a DVD burn.