iRay only I'm tired of it.

13468913

Comments

  • ColemanRughColemanRugh Posts: 511
    edited July 2016

    IRAY has obviously made life easier for vendors. Just like it's easier to rig in DS4 than Poser.  3DL customers will just have to decide if they want to buy every new thing coming out and fix all the textures themselves.

    For me, I don't need Pixar level quality. I'll go to each material zone and apply an all basic white shader I save. Then just add the vendors diffuse texture. If PBR makes it too bright or too dark... I'll then add grey in DAZ for too bright and take it into Photoshop for too bright and make the diffuse texture darker. use it without bumps, displacement, specular additions... and be choosey as hell to what I buy from here on out.

    I make comics - I don't need every render to be book cover quality.

    I don't own a $1,000 computer. Mine's $600. No IRAY for me. But I do support DAZ and the vendors as I can.

    Post edited by ColemanRugh on
  • ColemanRughColemanRugh Posts: 511
    edited July 2016

    This is a 3DL render with Photoshop. The render took 5 minutes. Postwork took 15. Jepes background effect prop is doing 90% of the work in this. It looks cartioonish but it's detailed. I could render 15 images like this a day and make a comic update for an ongoing webcomic. The numerous screw-ups in this image don't scream for attention, because it looks cartoonish. But everyone has their own style. I just don't need IRAY to accomplish this.

    sanpaku.jpg
    900 x 1200 - 242K
    Post edited by ColemanRugh on
  • GKDantasGKDantas Posts: 200

    I think a lot of people still use 3DeLight, thats why I updated my GIDome pack for Iray to work with 3DeLigt and even with a good settings its pretty fast http://www.daz3d.com/gidome-iray

    But yes you can see in the image that the result is very particular for every engine.

    UberIray.png
    989 x 881 - 1M
  • Ron KnightsRon Knights Posts: 2,196

    I'll drop in here for one last word concerning my attitude. I believe iRay is here to stay. Those of us who can only handle 3Delight will have to make do the best we can. I don't know if or when I'd ever get a more powerful computer that will better handle iRay. In the meantime I want to Make Art!

    It's a long, arduous journey for me to try and bring my ideas into "Reality." My Marlin's World creations have been waiting 15+ years to express themselves in their full glory. It's time to put those ideas to use. I'll use 3Delight until I can handle more. I want to thank Mec4D for providing both iRay and 3Delight materials for Unshaven2. You may know that many of my characters have beards!

    I have a new policy when it comes to forums. I usually say what I want, stick around to see if I'm heard. Then I quietly move on.

  • hphoenixhphoenix Posts: 1,335
    kyoto kid said:

    ...again Fibremesh hair is very memory intensive. It would pretty much require the memory resources of a Titan-X or even Quadro M6000 to hold a complete scene. especially with multiple characters.

    It is too bad Iray didn't have a similar system to Octane where it split the Geometry and Texture load.  It wouldn't be as lightning quick as pure GPU rendering but much faster than pure CPU rendering and wouldn't require a GPU with tonnes of memory at a heavyweight price..

    Actually, fibermesh hair isn't that bad.  While it is a LOT of geometry, geometry itself is a lot cheaper than image maps.  Geometry is around 32 bytes per vertex, plus Face lists (3 or 4 vertex indices per face, so about 14 bytes average).  This means that STATIC geometry (no morphs/bones) works out to about 96 bytes per face or less.  So a 100,000 face mesh would only occupy about 10MB of VRAM.  Compare this to a SINGLE 4k x 4k image map, which even at 50% compression in VRAM is 32MB.

    Compared to high-res textures, geometry is cheap.  So fibermesh hair is NOT that huge of a hit, memory wise.  However, the more polys in a scene, the more ray-tests and such have to be done, so it will slow down rendering.  Complex big texture maps are just table-lookups, so are pretty fast, even with decompression.  So fibermesh hair will slow your render down, but take up only a small amount of VRAM.  Strip-based textured hair will be faster to render, but takes up a lot more VRAM.  Trade-offs.  Of course, the moment you start using fancy shaders on the hair, including transparency, special specular calculations, translucency, and more, even strip-based textured hair will start to slow down the render as well.......

    But in general, geometry is a lot cheaper (memory-wise) but slower (render-wise.)

     

  • GatorGator Posts: 1,320

    I for one am glad Iray is here.

    I never used 3Delight, but it's probably similar to Poser's FireFly.  All these tricks with reflections and lighting and shaders, trying to mimic real lighting drove me bonkers.  Bonkers enough to buy Octane. 

    I find a PBR engine easier to use.  Simply use emitters where your lights are instead of trying to fudge them.  What I lose in rendering time I make up for in setup time.

    The world is going to PBR because consumer level hardware can handle it.  You can get a good previous generation Nvidia card cheap.

  • kyoto kidkyoto kid Posts: 41,925
    hphoenix said:
    kyoto kid said:

    ...again Fibremesh hair is very memory intensive. It would pretty much require the memory resources of a Titan-X or even Quadro M6000 to hold a complete scene. especially with multiple characters.

    It is too bad Iray didn't have a similar system to Octane where it split the Geometry and Texture load.  It wouldn't be as lightning quick as pure GPU rendering but much faster than pure CPU rendering and wouldn't require a GPU with tonnes of memory at a heavyweight price..

    Actually, fibermesh hair isn't that bad.  While it is a LOT of geometry, geometry itself is a lot cheaper than image maps.  Geometry is around 32 bytes per vertex, plus Face lists (3 or 4 vertex indices per face, so about 14 bytes average).  This means that STATIC geometry (no morphs/bones) works out to about 96 bytes per face or less.  So a 100,000 face mesh would only occupy about 10MB of VRAM.  Compare this to a SINGLE 4k x 4k image map, which even at 50% compression in VRAM is 32MB.

    Compared to high-res textures, geometry is cheap.  So fibermesh hair is NOT that huge of a hit, memory wise.  However, the more polys in a scene, the more ray-tests and such have to be done, so it will slow down rendering.  Complex big texture maps are just table-lookups, so are pretty fast, even with decompression.  So fibermesh hair will slow your render down, but take up only a small amount of VRAM.  Strip-based textured hair will be faster to render, but takes up a lot more VRAM.  Trade-offs.  Of course, the moment you start using fancy shaders on the hair, including transparency, special specular calculations, translucency, and more, even strip-based textured hair will start to slow down the render as well.......

    But in general, geometry is a lot cheaper (memory-wise) but slower (render-wise.)

     

    ...if you are stuck with only CPU rendering, those render times can become glacial  Also I need to make use of what I have as being on a fixed income, I cannot afford a whole new library of Fibremesh hair content to suit various different styles and lengths.  Persoanlly I like tools such as Garabaldi and LAMH as they give me the flexibility to create any style or length I need.  The drawback with Iray is they need to be imported as a .obj, and that really does pump up the polycount. So, with textures, specularity, and translucency, you get the worst of both worlds, more memory use and longer render time whreas in 3DL they render rather quickly.

  • Kendall SearsKendall Sears Posts: 2,995
    kyoto kid said:
    hphoenix said:
    kyoto kid said:

    ...again Fibremesh hair is very memory intensive. It would pretty much require the memory resources of a Titan-X or even Quadro M6000 to hold a complete scene. especially with multiple characters.

    It is too bad Iray didn't have a similar system to Octane where it split the Geometry and Texture load.  It wouldn't be as lightning quick as pure GPU rendering but much faster than pure CPU rendering and wouldn't require a GPU with tonnes of memory at a heavyweight price..

    Actually, fibermesh hair isn't that bad.  While it is a LOT of geometry, geometry itself is a lot cheaper than image maps.  Geometry is around 32 bytes per vertex, plus Face lists (3 or 4 vertex indices per face, so about 14 bytes average).  This means that STATIC geometry (no morphs/bones) works out to about 96 bytes per face or less.  So a 100,000 face mesh would only occupy about 10MB of VRAM.  Compare this to a SINGLE 4k x 4k image map, which even at 50% compression in VRAM is 32MB.

    Compared to high-res textures, geometry is cheap.  So fibermesh hair is NOT that huge of a hit, memory wise.  However, the more polys in a scene, the more ray-tests and such have to be done, so it will slow down rendering.  Complex big texture maps are just table-lookups, so are pretty fast, even with decompression.  So fibermesh hair will slow your render down, but take up only a small amount of VRAM.  Strip-based textured hair will be faster to render, but takes up a lot more VRAM.  Trade-offs.  Of course, the moment you start using fancy shaders on the hair, including transparency, special specular calculations, translucency, and more, even strip-based textured hair will start to slow down the render as well.......

    But in general, geometry is a lot cheaper (memory-wise) but slower (render-wise.)

     

    ...if you are stuck with only CPU rendering, those render times can become glacial  Also I need to make use of what I have as being on a fixed income, I cannot afford a whole new library of Fibremesh hair content to suit various different styles and lengths.  Persoanlly I like tools such as Garabaldi and LAMH as they give me the flexibility to create any style or length I need.  The drawback with Iray is they need to be imported as a .obj, and that really does pump up the polycount. So, with textures, specularity, and translucency, you get the worst of both worlds, more memory use and longer render time whreas in 3DL they render rather quickly.

    LAMH FiberHair is significantly smaller than OBJ.  Using default settings as much as 40-60% smaller on average.  Some hair styles can get upwards of 80% smaller.  But FiberHair does require the full plugin and not just the free player.

    Kendall

  • kyoto kidkyoto kid Posts: 41,925
    The world is going to PBR because consumer level hardware can handle it.  You can get a good previous generation Nvidia card cheap.

    ...not one that can handle the load of the scenes I create.  The downside of not having a beefy GPU is unlike Octane, in Iray once the file exceeds GPU memory everything. geometry and textures, are dumped to the CPU and all those CUDA cores are suddenly worthless.I've had Iray render files reach upwards of 10 GB in size  Only one card will work, and it has been discontinued (besides, my older hardware won't support it anyway so I would have to build an entirely new system which is even more expensive).

    I agree, emissives in Iray make for much easier lighting than having to place point and spot lights everywhere. However, I've worked with 3DL much longer. I also worked in theatrical lighting so placing a light to mimic an effect is not a big deal. The one thing I wish we had in RL was the ability to view through the light to aim it, would have saved having to make a lot of scary climbs up into the flylines to manually position and aim a hot lighting unit.

  • kyoto kidkyoto kid Posts: 41,925
    kyoto kid said:
    hphoenix said:
    kyoto kid said:

    ...again Fibremesh hair is very memory intensive. It would pretty much require the memory resources of a Titan-X or even Quadro M6000 to hold a complete scene. especially with multiple characters.

    It is too bad Iray didn't have a similar system to Octane where it split the Geometry and Texture load.  It wouldn't be as lightning quick as pure GPU rendering but much faster than pure CPU rendering and wouldn't require a GPU with tonnes of memory at a heavyweight price..

    Actually, fibermesh hair isn't that bad.  While it is a LOT of geometry, geometry itself is a lot cheaper than image maps.  Geometry is around 32 bytes per vertex, plus Face lists (3 or 4 vertex indices per face, so about 14 bytes average).  This means that STATIC geometry (no morphs/bones) works out to about 96 bytes per face or less.  So a 100,000 face mesh would only occupy about 10MB of VRAM.  Compare this to a SINGLE 4k x 4k image map, which even at 50% compression in VRAM is 32MB.

    Compared to high-res textures, geometry is cheap.  So fibermesh hair is NOT that huge of a hit, memory wise.  However, the more polys in a scene, the more ray-tests and such have to be done, so it will slow down rendering.  Complex big texture maps are just table-lookups, so are pretty fast, even with decompression.  So fibermesh hair will slow your render down, but take up only a small amount of VRAM.  Strip-based textured hair will be faster to render, but takes up a lot more VRAM.  Trade-offs.  Of course, the moment you start using fancy shaders on the hair, including transparency, special specular calculations, translucency, and more, even strip-based textured hair will start to slow down the render as well.......

    But in general, geometry is a lot cheaper (memory-wise) but slower (render-wise.)

     

    ...if you are stuck with only CPU rendering, those render times can become glacial  Also I need to make use of what I have as being on a fixed income, I cannot afford a whole new library of Fibremesh hair content to suit various different styles and lengths.  Persoanlly I like tools such as Garabaldi and LAMH as they give me the flexibility to create any style or length I need.  The drawback with Iray is they need to be imported as a .obj, and that really does pump up the polycount. So, with textures, specularity, and translucency, you get the worst of both worlds, more memory use and longer render time whreas in 3DL they render rather quickly.

    LAMH FiberHair is significantly smaller than OBJ.  Using default settings as much as 40-60% smaller on average.  Some hair styles can get upwards of 80% smaller.  But FiberHair does require the full plugin and not just the free player.

    Kendall

    ..LAMH fibre? As I reacall LAMH isalso a strand based system (Like Garabaldi) which is why it also needs to be converted to .obj to render in Iray.

  • Kendall SearsKendall Sears Posts: 2,995
    edited July 2016
    kyoto kid said:
    kyoto kid said:
    hphoenix said:
    kyoto kid said:

    ...again Fibremesh hair is very memory intensive. It would pretty much require the memory resources of a Titan-X or even Quadro M6000 to hold a complete scene. especially with multiple characters.

    It is too bad Iray didn't have a similar system to Octane where it split the Geometry and Texture load.  It wouldn't be as lightning quick as pure GPU rendering but much faster than pure CPU rendering and wouldn't require a GPU with tonnes of memory at a heavyweight price..

    Actually, fibermesh hair isn't that bad.  While it is a LOT of geometry, geometry itself is a lot cheaper than image maps.  Geometry is around 32 bytes per vertex, plus Face lists (3 or 4 vertex indices per face, so about 14 bytes average).  This means that STATIC geometry (no morphs/bones) works out to about 96 bytes per face or less.  So a 100,000 face mesh would only occupy about 10MB of VRAM.  Compare this to a SINGLE 4k x 4k image map, which even at 50% compression in VRAM is 32MB.

    Compared to high-res textures, geometry is cheap.  So fibermesh hair is NOT that huge of a hit, memory wise.  However, the more polys in a scene, the more ray-tests and such have to be done, so it will slow down rendering.  Complex big texture maps are just table-lookups, so are pretty fast, even with decompression.  So fibermesh hair will slow your render down, but take up only a small amount of VRAM.  Strip-based textured hair will be faster to render, but takes up a lot more VRAM.  Trade-offs.  Of course, the moment you start using fancy shaders on the hair, including transparency, special specular calculations, translucency, and more, even strip-based textured hair will start to slow down the render as well.......

    But in general, geometry is a lot cheaper (memory-wise) but slower (render-wise.)

     

    ...if you are stuck with only CPU rendering, those render times can become glacial  Also I need to make use of what I have as being on a fixed income, I cannot afford a whole new library of Fibremesh hair content to suit various different styles and lengths.  Persoanlly I like tools such as Garabaldi and LAMH as they give me the flexibility to create any style or length I need.  The drawback with Iray is they need to be imported as a .obj, and that really does pump up the polycount. So, with textures, specularity, and translucency, you get the worst of both worlds, more memory use and longer render time whreas in 3DL they render rather quickly.

    LAMH FiberHair is significantly smaller than OBJ.  Using default settings as much as 40-60% smaller on average.  Some hair styles can get upwards of 80% smaller.  But FiberHair does require the full plugin and not just the free player.

    Kendall

    ..LAMH fibre? As I reacall LAMH isalso a strand based system (Like Garabaldi) which is why it also needs to be converted to .obj to render in Iray.

    Unlike Garibaldi, LAMH was written to support multiple systems.  LAMH FiberHair is a highly optimized geometry hair that renders in Vue, Iray, Maya, Lightwave, and others.

    EDIT:  I normally avoid talking about competing products especially GE.  However, this time it was necessary in order to highlight one of the differences.

    Kendall

    Post edited by Kendall Sears on
  • Oso3DOso3D Posts: 15,088
    edited July 2016
    Yeah, when I was converting LAMH to obj I gave up on it as infeasible. Fiberhair is -incredibly- better. (And I'm still kicking myself for taking so long before I figured out the fiber mesh conversion was different)
    Post edited by Oso3D on
  • SpottedKittySpottedKitty Posts: 7,232
    Yeah, when I was converting LAMH to obj I gave up on it as infeasible. Fiberhair is -incredibly- better. (And I'm still kicking myself for taking so long before I figured out the fiber mesh conversion was different)

    Agreed; I've done renders now of a good few of the LaMH-equipped critters, and converting to Fiberhair so I can render in Iray is much easier for me. Doesn't it also render faster as well?

  • kyoto kidkyoto kid Posts: 41,925

    ...well have to wait to get back into PC+ and there's an Alessendro sale, don't have the 49$

    So does LAMH play better with G3?

  • Kendall SearsKendall Sears Posts: 2,995

    An update that (hopefully) fixes all of the issues with G3 hit DIM today, I believe.

    Kendall

  • kyoto kidkyoto kid Posts: 41,925

    ...don't see it yet.  Only the 4.9 stuff (still on 4.8)

  • 3Diva3Diva Posts: 11,981

    As someone who's new to Daz Studio and 3D art, I didn't even know that 3Delight existed for the first couple months. I would see 3Delight images and just figure they had a filter or something used to give it more of a "toon" look. I like the look of both Iray and 3Delight renders - as both have their ups and downs. Iray obviously looks great for more realism renders, but 3Delight really does wonderful work as well and even, imo, looks better with non-realism renders. 

    If I where to pick a favorite style though it would be more realism and Iray (and that, perhaps, might only be that I only ever rendered in Iray for the first few months). But I don't think I'm the only person to prefer the look of Iray, though. I mean, you look at the Galleries and the vast majority of the most upvoted images are Iray. I think there is a reason for that. To me, Iray is a step forward - at least as far as reality based rendering. 3Delight does look better for non-realism renders though. Art styles that lean toward toon or comic art type definitely, to me, look better rendered in 3Delight. I do think that both have their merits and pitfalls and that there is room for both in most artists' toolbox. I really hope that vendors continue to support both 3Delight and Iray. 

  • Subtropic PixelSubtropic Pixel Posts: 2,388

    I too favor Iray.  3DL takes a long long LONG time to set up lighting.

  • Oso3DOso3D Posts: 15,088
    I'm finding I've refined 3dl lighting a lot. Most of the time I just need AoA distant or spotlight + AoA ambient (sometimes a second AoA distant or spot, but not very often) aaand... done.
  • DaWaterRatDaWaterRat Posts: 2,885
    I'm finding I've refined 3dl lighting a lot. Most of the time I just need AoA distant or spotlight + AoA ambient (sometimes a second AoA distant or spot, but not very often) aaand... done.

    Yeah, I just did a composite.  It took about the same amount of time to render each version (3DL/Iray)  I had 1 AoA ambient, 1 AoA Distant light and 2 AoA spotlights in the 3DL.  I dropped the Ambient for an HDRI and replaced the other AoA lights with their respective default versions in Iray.

    Each picture took less than 10 minutes to render.

    Granted, it's an outdoor immage. (Final postworked version attached.  The intent was not photoreal but painterly)

    Charlies new plane Artistic.png
    1200 x 900 - 2M
  • algovincianalgovincian Posts: 2,667

    As someone who's new to Daz Studio and 3D art, I didn't even know that 3Delight existed for the first couple months. I would see 3Delight images and just figure they had a filter or something used to give it more of a "toon" look. I like the look of both Iray and 3Delight renders - as both have their ups and downs. Iray obviously looks great for more realism renders, but 3Delight really does wonderful work as well and even, imo, looks better with non-realism renders. 

    If I where to pick a favorite style though it would be more realism and Iray (and that, perhaps, might only be that I only ever rendered in Iray for the first few months). But I don't think I'm the only person to prefer the look of Iray, though. I mean, you look at the Galleries and the vast majority of the most upvoted images are Iray. I think there is a reason for that. To me, Iray is a step forward - at least as far as reality based rendering. 3Delight does look better for non-realism renders though. Art styles that lean toward toon or comic art type definitely, to me, look better rendered in 3Delight. I do think that both have their merits and pitfalls and that there is room for both in most artists' toolbox. I really hope that vendors continue to support both 3Delight and Iray. 

    You'll be rendering in 3DL soon enough ;)

    - Greg

  • Oso3DOso3D Posts: 15,088
    My arc was 3dl, frustration, Iray when it came out... frustration, 3dl... as I've said elsewhere, Iray is easier at first, but as you gain more skill and mastery the two renderers start evening out. Unless you have very particular goals that favor one or the other, I think there are good reasons to use both.
  • RAMWolffRAMWolff Posts: 10,352

    I love iRay, it's a breath of fresh air but there are some weird things about it I hope in time will be ironed out.  I mean we have only had it for, what?, a year now?  Already we have folks discovering ways to bring special FX effects that we were told could only be done in 3Delight so there is hope. 

    One kevetch I have is using older Environments, even after conversion sometimes brings my renders to a crawl and takes upwards of 2 minutes to even see the preliminary render on screen.  Usually with a character, dressed with hair and a few props the render is visible in a few moments but with a small environment loaded and converted to iRay I find the wait to be concerning and then it's there (when I come back from washing up a few dishes in the sink, LOL). 

  • MattymanxMattymanx Posts: 6,997

    I find I can get nearly the same look in IRay as I did in 3DL but with far more realistic lighitng.  After having to do promos in 3DL for the new car and addon kit I was reminded why I prefer Iray.

  • kyoto kidkyoto kid Posts: 41,925

    ...Iray seems to be "made" for shiny items (like cars, planes, high tech gizmos stuff with lots of lights, etc).  I'll see a great pic of a classic car or bike on the Car and Bike Lovers Thread which looks pretty darn real, only to find if a character is included, it (the character) doesn't seem to visually fit in.

     

  • kyoto kidkyoto kid Posts: 41,925
    RAMWolff said:

    I love iRay, it's a breath of fresh air but there are some weird things about it I hope in time will be ironed out.  I mean we have only had it for, what?, a year now?  Already we have folks discovering ways to bring special FX effects that we were told could only be done in 3Delight so there is hope. 

    One kevetch I have is using older Environments, even after conversion sometimes brings my renders to a crawl and takes upwards of 2 minutes to even see the preliminary render on screen.  Usually with a character, dressed with hair and a few props the render is visible in a few moments but with a small environment loaded and converted to iRay I find the wait to be concerning and then it's there (when I come back from washing up a few dishes in the sink, LOL). 

    ...my railway station scene takes over 40 min before anything shows up in the render window.

  • MattymanxMattymanx Posts: 6,997
    kyoto kid said:

    ...Iray seems to be "made" for shiny items (like cars, planes, high tech gizmos stuff with lots of lights, etc).  I'll see a great pic of a classic car or bike on the Car and Bike Lovers Thread which looks pretty darn real, only to find if a character is included, it (the character) doesn't seem to visually fit in.

     

    lots of people are doing good character renders with Iray.  But organics are more difficult to do then metals and such

  • kyoto kidkyoto kid Posts: 41,925
    edited July 2016

    ...again Hair seems to be the biggest issue as most ready made hair content is strip based and transmapped. If I ever can get the funds up I'll look into getting the full LAMH since it can convert to fibremesh. I wonder if the new UHT shaders will work with that.

    I agree, skin is getting better.  When Skin Builder Pro3 comes out then I'll have something to work with as I believe it will also have Iray settings. Until then I'll still be behind the curve as I cannot afford a bunch of G3 character morphs just to get updated skins. I've pushed skin texture as far as I can on my own without getting so wrapped up in it that I never get any other work done (part of the reason I dropped from the Fiddling With Skin in Iray thread, just got a little too technical for me, figured I'd let those with the know how deal with it and maybe we'll see something filter down into the store).

    Still to really make good use of Iray and not risk waiting days for a render to complete, will need a new system with more up to date components.  Again would be nice if they could work out a hybrid mode like Octane uses, then I could get by with a 250$ 6 GB 1060. instead of a $$$$ 16 GB Titan P.

    Post edited by kyoto kid on
  • kyoto kid said:

    Still to really make good use of Iray and not risk waiting days for a render to complete, will need a new system with more up to date components.  Again would be nice if they could work out a hybrid mode like Octane uses, then I could get by with a 250$ 6 GB 1060. instead of a $$$$ 16 GB Titan P.

    If I understand how Iray works correctly, the reason CPU mode is slower than GPU mode is because they are emulating CUDA in software; hybridized mode may not gain as much speed as it could for that reason.

  • outrider42outrider42 Posts: 3,679
    kyoto kid said:
    The world is going to PBR because consumer level hardware can handle it.  You can get a good previous generation Nvidia card cheap.

    ...not one that can handle the load of the scenes I create.  The downside of not having a beefy GPU is unlike Octane, in Iray once the file exceeds GPU memory everything. geometry and textures, are dumped to the CPU and all those CUDA cores are suddenly worthless.I've had Iray render files reach upwards of 10 GB in size  Only one card will work, and it has been discontinued (besides, my older hardware won't support it anyway so I would have to build an entirely new system which is even more expensive).

    How can you be certain you'll go over the VRAM that much? The ram monitors are not 100% accurate. I routinely exceed 2 gb and still pump out renders on my 2gb card. 

    We all have to deal with limitations. I'd love to have a Titan or two myself, and I'd probably go nuts with them if I had them. But that's simply not happening. Part of the art form is figuring out creative ways to get around hardware limitations. I would guess that even if you did fit your entire scene into a Titan, it would still take forever and a day to render, far too long to be practical. The stuff you are wanting to do is beyond any single computer. So no matter what the VRAM outcome is it would still be best to find a way to render the scene in parts and post them together. I've had images of a single figure with no back ground take hours to render, even though my GPU was trucking away at 100%. That VRAM didn't help me, my CUDA count was the issue (a GTX 670.) The 670 is old, but it still has a decent CUDA count.

    The fact is Iray is still extremely difficult on hardware, even high end hardware. And that is the problem. It needs to be better optimized so badly. It needs to be able to split duty with standard RAM and delegate tasks between GPU and CPU. It needs to understand when pixels are not visible do not need to be rendered! Like a car with an engine in it, even though the hood is closed, that engine is still taking up memory. So to render the car faster, you need to delete or hide that engine. What nonsense is that? With VR taking off, look at how Nvidia worked to build a better VR experience. The 1000 series cards are so much faster at VR, way faster than their actual specs suggest they should be. Why? Because of better software. When the 1000 series runs a VR app, it takes into account which pixels are not visible, and it dumps these pixels. It does this for every single frame, which is over 90 frames per second. Now I know gaming tech does not transfer directly to Iray, but my point here is if Nvidia can figure out VR like this, they need to apply a technique like this to Iray. When pushed, Nvidia is capable of delivering an answer.

    So if Nvidia ever gets off their sorry butts to fix Iray, then we'll see dramatic improvements in render times for everybody. Iray is more than just spec, it is software. My hope is that somebody will come along and do it better than Nvidia. Until then, I really wish Daz could work on their own rendering engine. That way they would have full control over it and not need to depend on Nvidia. I just think its a bad idea for Daz to depend on Iray so much.

Sign In or Register to comment.