Does anyone have advice on a reasonably priced external PCI-E X16 expansion box? The ones I could locate thru Google cost more than a new computer. I only have one slott for the time being, and I don't think I will be buying a new computer anytime this year (my car's transmission just went out).

I'm not an expert on this, but others have indicated it is better to get a new motherboard that supports multiple cards (maybe MSI Z77A-G45 @ about $100?) than an expansion box.  You can use a single graphics card for rendering and Windows display - it just means Windows runs a little slowly when rendering heavy scenes.  You can also use you on-board graphics for Windows display and the graphics card for rendering (my current setup) and they works very well.  The best way to check is to run the demo version of the plugin and see how you current system performs.

Attached Link: http://www.fxguide.com/featured/octane-render-realtime-ray-tracing/

Attached Link: http://render.otoy.com/forum/viewtopic.php?f=5&t=27920

Attached Link: http://render.otoy.com/forum/viewtopic.php?f=5&t=26689

I literally was like :O when I saw these renders over the weekend, while gazing through the Otoy forums.

Amazing work.

He's very very good!

Smith Micro should hire him when they start on a new high res model for Poser.

I'm running the Poser octane demo with 2 GTX 580 3GB cards, my question is does the renderer uses both cards and the 6GB of memory.  

Also I can load two gen 4 characters fine and it renders instantly but a Stonemason set never comes thru, the screen stays black.  I've seen videos showing Stonemason renders in Octane so I wonder if I need to setup things different.

What about using a third cheaper card for my monitor and leaving the 2 GTX 580 for Octane, will that help with Octane performance?

There are 3 limiting factors: VRAM, texture slots and number of GPU's.

A GTX580 has 64 RGB slots and 32 Grayscale slots and an additional GTX580 will double that (so 128 and 64). A texture connected to a diffuse channel takes a RGB slot and a texture connected to transmap/bumpmap will take a grayscale slot.

Octane (or actually CUDA) needs all texture memory available on each card. So your total with 2 cards remains 3GB. If it is the only card, it will share that memory with the OS. So having an additional (cheap) videocard for the UI of the OS will make more memory available to Octane. So, highly recommended.

Octane/CUDA distributes the load over all the GPU's available, so adding additional cards (with the same VRAM) will speed things up.

Now the Stonemason scenery. Good example is Urban Sprawl 2. If you take the complete scene and let the plugin convert, it will use up (If I remember correctly) 106 RGB slots and over 80 grayscale slots. So this will not load in your configuration. The plugin however contains a few actions which will let you strip bump maps from the whole or parts of the scene. You can do this with items far away or out of view. Another thing you can do is convert some of the grayscale slots to RGB slots to bring the number of grayscale slots down. One caveat - you need to do that for all references of that particular image map otherwise it will keep taking a grayscale slot.
The other thing you can do (usually easier) is just hide the parts out of view in Poser, this will take the total number of used slots down as well.
The actual VRAM used in that scene is not very high and it will fit in your configuration (I think it was around 800 or 900MB)

I hope this helps

Thank you for the detailed reply.  I still have to learn a lot to be able to maximize the Octane use in my system.  My motherboard is the sabertooth x79 from Asus and I currently have the two top PCIE x16 slots populated with the GTX 580 cards.  I have the last PCIe 3.0 x16 slot free so I'm guessing I couls stick my third graphic card there and use it for my monitor and the OS.  Problem is that third PCI slot says in the motherboard manual that it gets used at x8 mode instead of x16 mode so I'm wondering if my system performance will diminish.  Right now Windows 7 gives me a 7.8 rate because my i7 3930K CPU is being ran at stock, I know if I overclock just a little it can get to 7.9 but if I use a weaker card for my system and in x8 performance may dip in OpenGL viewports I guess, unless the two GTX 580's keep being used even though they don't connect to a monitor.  Many questions, like I said I still have a lot to learn.

I have my GTX580 as the card for the OS (monitor connected) and a GTX680 (which has more GPU's, 4GB VRAM and 144 RGB slots) as the card for Octane. So I do not have a performance problem. But I guess it depends on what you use the system for. If you have a reasonable videocard, OpenGL will work just fine. If you play games however, that might not be enough (but you may be able to use a monitor switch)

In my setup I have the option of rendering faster with additional GPU's by activating the second card as well. But it has 1.5GB VRAM and then has to share it with the OS - this means it can only be done for small scenes.

The 8x and 16x thoughput does not matter too much for OpenGL (I think), it is only benefical for moving large chunks of memory to and from the videocard (games)

Once you get used to it how it works, it is pretty easy. If you modify a conversion for a particular figure or prop, you can save the Octane material in the Poser material. Save that in the scene, in the library or as a material/material collection and you can reuse it - saving you a lot of time. You can also save figure materials as Octane materials and reuse them for all your figures. Applying it has the option of retaining the texture maps which means you can reuse a saved material to any figure of the same type (gen4 for example). It will get all the SSS and other goodies you have setup with the new textures

There is one thing which I noticed however with skin textures. Your settings for those textures depend on the texture itself. Some translate very well, some need adjustments and some are not good at all. (but they did not look good in Poser either). So even though you can reuse them, you still may have to make adjustments depending on the lighting you use

I'm not a gamer, not even a casual one so my system was built specifically for rendering in CPU and GPU as I use both DAZ Studio and Poser, I also develop add ons for both using Adobe, Z-Brush and 3D Coat.  I use Luxrender too but I'm looking at Octane for realtime renders and animation.

 I thought you shouldn't mix Kepler with Fermi or that at least it wasn't recommended.  I also had been told that the GTX 580 was the fastest card for Octane so that is why I went for that one twice.  I hear that Kepler is coming thru now but I still see the developers using GTX 580 in the Octane forum.

To be honest, I haven't tried mixing the two with GPU rendering, so I am not sure it can be done.

Speed is a relative matter in this case. Yes, the Fermi is faster, but the Keppler has more GPU's. But for me the main reason to get a GTX680 was the amount of VRAM and the texture slots. I first worked with a single GTX580 with 1.5GB. And I kept running out of VRAM and texture slots. Now I hardly ever have this problem anymore (except for very large scenes). An image with 5 V4's, clothing, hair and props is about the max. After that you need to economize on texture size (the 4000x4000 texture are eating all the VRAM). The plugin has a script to reduce texture size

The realtime rendering is a difficult one: Yes, the render speeds can be incredible, but every time you add more realism, the renderspeed goes down. So getting a very realistic render (image) can add up - especially of you have lots of specular material in it (glass, water) with low light conditions. For animation you can get by with a lower quality, but not with stills

I was able to load a Jack Tomalin set into the Poser Octane demo just now, its the "Casa del Diablo" castle.  I tried to move the lights in Poser to see the changes in realtime but it doesn't do anything, maybe I need to setup something in Octane or the demo is limited this way.

Quote - I was able to load a Jack Tomalin set into the Poser Octane demo just now, its the "Casa del Diablo" castle.  I tried to move the lights in Poser to see the changes in realtime but it doesn't do anything, maybe I need to setup something in Octane or the demo is limited this way.

There are 2 lighting models: Daylight and HDRI. In addoition to that you have light emitters. There is a script in the plugin which created light emitters from poser lights. If you want to use those, run the script.

In the daylight model the first infintie light is taken as the sun. You can move that light in Poser and have immediate changes in the render. If you want the emitters to react in the render, you have to enable the automatic refresh. Camera changes are always seen in the render viewport. No need for the automatic refresh

For HDRI lighting the IBL image is taken (if it is available), otherwise plug one in.

Both daylight and HDRI models can work with emitters

Disclaimer: This is what happens in the actual version of the plugin with 1.02 octane. Not sure what version the demo is.

I have to go, so any other question will have to be answered later today

I ran the light script and it created the emitters but I move the lights in Poser and can see Poser viewport reacting to this but the Octane windows stays the same.  The demo version says 1.0.

I have to go, so any other question will have to be answered later today>>

I understand.  Thank you so much for sharing all this information.

Hi Simon (and Wim)

A few comments....

1. 2 x 580 cards should render incredibly fast (and yes, at the moment they seem to have the performance edge on the 600 series cards).  The only downside is you have fewer texture slots that the 600 series cards.

2. If the Stonemason scene is rendering totally black, check to make sure there isn't a skydome blocking all the scene light.

3. The demo you are using is 1.00, so there have been a couple of fixes and many enhancements since then.  With the light emitter script, the movnig of lights now propagates to the Octane scene.  Plus the scene infinite light now moves the Octane sun in the Octane scene.

4. If possible - use your on-board graphics card as your Windows display adapter.

Paul

Thanks for the information and confirming about the light, excellent work on this plugin and on Poser Phisics. Easily the most exciting plugins for poser right now.

I have to correct myself at one point in the explanation: I am not so sure anymore that the texture slots can be added up if you add 2 cards. The documentation nowhere explititly mentions these things, but apparently all textures must be present in the videocard memory in all cards. It now makes more sense for that to mean that it also uses up all texture slots.

For Poser users with large scenery the texture slots are important because they use up a lot of 3rd party meshes each with their own textures. In the 3DMax like world, most meshes are under the creators own control and are "standard" materials and do not use a lot of slots, so they will not run into this problem as much as the Poser/Studio users.

So for Poser/Studio I think it is better to have the 600 series of cards.

If so why do Octane developers still prefer to use multiple GTX 580s? Perhaps Paul can clear this up.

One explanation is that they mostly use procedural textures, so they do not run into this problem of lack of texture slots very often. Another reason may be is that support for the kepler gpu was only introduced a couple of months ago and of course the fermi is faster.

But Paul may know of other reasons as well

I actually researched this a lot before diving in for the GTX 580's and even though I read in the Octane forum that the Kepplers are coming thru in performance I don't know if they will ever match the 580's, I could be wrong but I've read that Nvidia made many compromises to get the Kepler cards optimized for games and energy consumption that resulted less performance for graphic computing applications not only Octane but After Effects and others who use the GPU.

http://www.tomshardware.com/forum/351813-33-kepler-fermi-kepler-noticably-effects

I was going to go for a 690 and was told by a beta tester of the DAZ Studio Octane to stay away from Kepler, and that the Octane programmers were disapointed on how incredibly slow they were compared to Fermi.  I suspect that the comment that Kepler had finally come thru means that the speeds are closer to what a top of the line fermi can produce.

You mentioned there is an addon that automatically reduced texture resolution.  I wish there was a way to use the system ram to bypass the GPU ram limitations.

Regarding Kepler vs Fermi. Yes, the Fermi was a bit faster (I did use the 580 before I bought the 680). But if I can't get the scenes into Octane, it is of nu use to me at all

And the village Courtyard exactly shows the problem.

It has multiple parts. Each part has its texture, its bump map and many of them (plants) transmaps. All transmaps and bump maps are greyscale - so you can have a max of 32 on the 580 (and 72 on the 680), so you are running out of slots here.

In this case you can render it without the bump maps. Go to the material editor, right click on the scene line and select remove bump maps. You will see that 600 something bump maps (actually references) are removed. Now the image statistics say you only use 10 grayscale maps and you can render it

The limit refers to unique textures (so many references to 1 map counts as 1). Stonemason uses for each prop a different map. In this case it is a problem since there are just too many maps

If so why do Octane developers still prefer to use multiple GTX 580s? Perhaps Paul can clear this up.

The 500 series cards are my personal preference, because I like the speed, and I don' t use the Stonemason Cityblocks or other very heavy Poser props.  But because of these props, I /have/ the recommend the 600 series cards.  I think most non-Poser users would still prefer the 500 series cards (since Poser it's really the only 3d app which has such heavy texture requirements - ie. it's the only app with which you can click a button and load a huge scene like the Forest figure). If you are using these pre-fab scenes, it's a good idea to unclick the visibility box on any components that you don't need (ie. not in the camera shot)  - which will save you load time and texture slots.

Paul

I have to correct myself at one point in the explanation: I am not so sure anymore that the texture slots can be added up if you add 2 cards.>>

Paul can you confirm if the texture slots can be added up if you have two cards, for example in my case two GTX 580 each one with 3GB?

Paul can you confirm if the texture slots can be added up if you have two cards, for example in my case two GTX 580 each one with 3GB?

They are not added.  When you have 2 cards, the scene (simultaneously) gets loaded into both cards, and they both contribute to the render (effectively doubling your render speed).  So in your case, you have 3GB VRAM (which is plenty) and 64 RGB and 32 Grayscale texture slots.  The 600 series cards are capable of unlimited texture maps, however in order for Octane to be backwardly compatible with older cards, Octane has a limit on the 600 series cards of 144 RGB and 68 grayscale (plus another 20 HDR images) - so you can see the 600 series cards offer much more to the Poser user /IF/ you are working we very large scenes.  If your standard render is a couple of figures plus clothing plus a few props, you are unlikely to go near the 64 RGB limit - it's really the big environment props which use 100+ individual texturemaps where you will need to make the out-of-camera shot elements invisible to get into the 64 RGB texturemap limit.

Paul

Quote - In this case you can render it without the bump maps. Go to the material editor, right click on the scene line and select remove bump maps. You will see that 600 something bump maps (actually references) are removed. Now the image statistics say you only use 10 grayscale maps and you can render it

Thanks to this tip I was able to load the scene though with scenes this big I cannot rotate the camera while Octane renders, I need to pause the renderer, rotate the camera in Poser viewport and hit the play button to resume/update the render.

Paul, it is not just "very large scenes". If you take a default V4 figure it already takes 8 RGB maps and at at least 6 grayscale maps, add a hair piece and add at least 2 RGB and 2 transmaps, add skirt, shoes and shirt and another 3 of each. So 2 clothed figures take at least 30 RGB maps and 22 grayscale maps. Add a few props and you run out of grayscale maps on a 64/32 GTX 580. And 2 figures with a few props is not a large scene by any means in Poser. When I was testing the GTX 580 I was unable to load 90% of my scenes without removing the bump maps

Quote - > Quote - In this case you can render it without the bump maps. Go to the material editor, right click on the scene line and select remove bump maps. You will see that 600 something bump maps (actually references) are removed. Now the image statistics say you only use 10 grayscale maps and you can render it

Thanks to this tip I was able to load the scene though with scenes this big I cannot rotate the camera while Octane renders, I need to pause the renderer, rotate the camera in Poser viewport and hit the play button to resume/update the render.

That is because you use the OS and Octane on the same video card. Add a low end videocard for the OS UI and that will not happen anymore

Quote - Paul can you confirm if the texture slots can be added up if you have two cards, for example in my case two GTX 580 each one with 3GB?

They are not added.  When you have 2 cards, the scene (simultaneously) gets loaded into both cards, and they both contribute to the render (effectively doubling your render speed).  So in your case, you have 3GB VRAM (which is plenty) and 64 RGB and 32 Grayscale texture slots.  The 600 series cards are capable of unlimited texture maps, however in order for Octane to be backwardly compatible with older cards, Octane has a limit on the 600 series cards of 144 RGB and 68 grayscale (plus another 20 HDR images) - so you can see the 600 series cards offer much more to the Poser user /IF/ you are working we very large scenes.  If your standard render is a couple of figures plus clothing plus a few props, you are unlikely to go near the 64 RGB limit - it's really the big environment props which use 100+ individual texturemaps where you will need to make the out-of-camera shot elements invisible to get into the 64 RGB texturemap limit.

Paul

I mostly want to use Octane to render animations of characters so I guess the 580s will be good enough.  What type of card should I use if I want to use a cheaper card for the OS, since I don't have onboard video.  I have a Zotac GT 430 2GB DDR3 with a measly 96 cuda cores that I can add to my third PCI-e slot, will that work? 

Quote - Paul, it is not just "very large scenes". If you take a default V4 figure it already takes 8 RGB maps and at at least 6 grayscale maps, add a hair piece and add at least 2 RGB and 2 transmaps, add skirt, shoes and shirt and another 3 of each. So 2 clothed figures take at least 30 RGB maps and 22 grayscale maps. Add a few props and you run out of grayscale maps on a 64/32 GTX 580. And 2 figures with a few props is not a large scene by any means in Poser. When I was testing the GTX 580 I was unable to load 90% of my scenes without removing the bump maps

 

 Your 580 is a 1.5 GB of ram which is a very small amount, I gather 3 & 4 is more appropriate but does that extra 1Gig between the 580 3G & the 680 4G makes such a big difference?

A GTX580 has 64 RGB slots and 32 Grayscale slots and an additional GTX580 will double that (so 128 and 64). A texture connected to a diffuse channel takes a RGB slot and a texture connected to transmap/bumpmap will take a grayscale slot.>>

Is the information above the numbers for the GTX580 1.5GB or the GTX 580 3GB?

I have a Zotac GT 430 2GB DDR3 with a measly 96 cuda cores that I can add to my third PCI-e slot, will that work?

Yes.  The display adapter doesn't need any cores! 

Your 580 is a 1.5 GB of ram which is a very small amount, I gather 3 & 4 is more appropriate but does that extra 1Gig between the 580 3G & the 680 4G makes such a big difference?

I would have thought it would be very rare to use up 3Gig, so 4Gig seems like a marginal difference.  I have 1Gig, and that's fine for the renders I do, but I have troubles loading some of the scene people send me into that space.  You probably can't buy 1Gig cards anymore!

Quote - > Quote - Paul, it is not just "very large scenes". If you take a default V4 figure it already takes 8 RGB maps and at at least 6 grayscale maps, add a hair piece and add at least 2 RGB and 2 transmaps, add skirt, shoes and shirt and another 3 of each. So 2 clothed figures take at least 30 RGB maps and 22 grayscale maps. Add a few props and you run out of grayscale maps on a 64/32 GTX 580. And 2 figures with a few props is not a large scene by any means in Poser. When I was testing the GTX 580 I was unable to load 90% of my scenes without removing the bump maps

 

 

 Your 580 is a 1.5 GB of ram which is a very small amount, I gather 3 & 4 is more appropriate but does that extra 1Gig between the 580 3G & the 680 4G makes such a big difference?

 

A GTX580 has 64 RGB slots and 32 Grayscale slots and an additional GTX580 will double that (so 128 and 64). A texture connected to a diffuse channel takes a RGB slot and a texture connected to transmap/bumpmap will take a grayscale slot.>>

 

Is the information above the numbers for the GTX580 1.5GB or the GTX 580 3GB?

I ran out of texture slots before I ran out of memory. And you can reduce texture size, remapping props and figures is not as easy.

I gave you the limitation of using a GTX580, if you think that is not a problem in your case, then all is fine. I just want you to be aware of it before you buy your second video card

Is the information above the numbers for the GTX580 1.5GB or the GTX 580 3GB?

All 500 series cards have 64/32 texture slots.

Quote - I ran out of texture slots before I ran out of memory. And you can reduce texture size, remapping props and figures is not as easy. I gave you the limitation of using a GTX580, if you think that is not a problem in your case, then all is fine. I just want you to be aware of it before you buy your second video card

 

But I already have two 3GB GTX 580s.

Quote - Is the information above the numbers for the GTX580 1.5GB or the GTX 580 3GB?

All 500 series cards have 64/32 texture slots.

Hmm so the card memory doesn't has to do with texture slots then.  Again, I still have a lot to learn.  :( Thank you guys.

Then 64 rgb slots sounds like a very few slots, 144 sounds much better.

So for renders that use more than one skin layer like wounds on top of a skin texture map those would be occupying two RGB slots?

Quote -  

Quote - I ran out of texture slots before I ran out of memory. And you can reduce texture size, remapping props and figures is not as easy. I gave you the limitation of using a GTX580, if you think that is not a problem in your case, then all is fine. I just want you to be aware of it before you buy your second video card

 

 

But I already have two 3GB GTX 580s.

Then go ahead with that. If you do run into the texture slot problem you can always consider getting a 600 series for larger scenes. But do get that extra videocard (which ever it is) to offload the OS UI to that card to improve your UI responsiveness.

Quote - Then go ahead with that. If you do run into the texture slot problem you can always consider getting a 600 series for larger scenes. But do get that extra videocard (which ever it is) to offload the OS UI to that card to improve your UI responsiveness.

Now I wonder if that extra card, being a lower end card than the 580s, will lower my computer performance in other graphic software like Adobe or 3DCoat and Z-brush, viewport response in Poser and DAZ Studio or the dual 580's still get used by them even though the monitor is not connected to the 580's?  How does windows 7 deals with two different drivers? 

So for renders that use more than one skin layer like wounds on top of a skin texture map those would be occupying two RGB slots?

IMO, for this situation you are better merging the maps in Photoshop/Gimp/etc and using one map with the skin and wounds.

The thread at http://render.otoy.com/forum/viewtopic.php?f=45&t=25352 is the FAQ for the plugin, and covers a lot of what has been discussed here.

Quote - So for renders that use more than one skin layer like wounds on top of a skin texture map those would be occupying two RGB slots?

IMO, for this situation you are better merging the maps in Photoshop/Gimp/etc and using one map with the skin and wounds.

The thread at http://render.otoy.com/forum/viewtopic.php?f=45&t=25352 is the FAQ for the plugin, and covers a lot of what has been discussed here.

It depends on how the layering is done. If it is driven by multiply you would need 2 RGB slots (and  memory), if it is driven by a distribution map (as in a blender node), you might need 2 RGB maps and a greyscale map. If the wound map is positioned by moving texture coordinates, you cannot do it in octane. The easiest (and most efficient) is to - what Paul suggested - do the blending in Photoshop and present one texture.

I have experimented with a tan layer on a skin in combination with scattering and an color change for the tanned skin and it does get complicated. (But you can save the result and appy it to any similar figure, retaining the texture maps)

Quote - > Quote - Then go ahead with that. If you do run into the texture slot problem you can always consider getting a 600 series for larger scenes. But do get that extra videocard (which ever it is) to offload the OS UI to that card to improve your UI responsiveness.

 

Now I wonder if that extra card, being a lower end card than the 580s, will lower my computer performance in other graphic software like Adobe or 3DCoat and Z-brush, viewport response in Poser and DAZ Studio or the dual 580's still get used by them even though the monitor is not connected to the 580's?  How does windows 7 deals with two different drivers? 

I really can't say. OpenGL does not use HW Shading, so for preview it should not matter too much. For HW Shading in preview it might make a difference if it is really slow. (GPU won't matter at all). I have used ZBrush only with recent cards, so I can't comment on that

Windows only uses the card where the monitor is connected. The GTX 580s would only be used by the software which accesses it directly (Octane - and maybe Lux, I don't know about that)

Thanks again for your answers Paul and wimvdb.

I have bought a GTX TITAN. Great improvement, more than 2X faster then my previous 660 GTX Ti.

But I have a problem with memory with Poser plugin : only 2GB is recognize and not 6 GB.

I have also noticed that rendering in the standalone app seems to be faster. Or it is just the material conversion ? because exporting a poser scene and then import it to the standalone app is tedious for configuration of all materials.Would it be possible Paul to add an export nodes option in the poser plugin and then import it in the standalone ? and also to render a scene done with poser without poser launched.

I have bought a GTX TITAN. Great improvement, more than 2X faster then my previous 660 GTX Ti.

I'm am so jealous!

But I have a problem with memory with Poser plugin : only 2GB is recognize and not 6 GB.

I suspect it is using the whole 6GB, but the plugin is only displaying 2GB.  Does Standalone display the 6GB?  If you load 5GB into the video card, does the Poser plugin indicate there is still 1GB free?  I'd be interested to investigate this further, to sort out any issues there might be with the Titan - since it's going to be incredibly popular.

I have also noticed that rendering in the standalone app seems to be faster. Or it is just the material conversion ? because exporting a poser scene and then import it to the standalone app is tedious for configuration of all materials.Would it be possible Paul to add an export nodes option in the poser plugin and then import it in the standalone ? and also to render a scene done with poser without poser launched.

Small things like different "fov", "aperture" and camera positions will change the render speed.  Also, the Poser plugin builds up a node network where each prop and figure is plugged into a different "geometrygroup" node pin, which enables the live-geometry updating (which you cannot do in standalone).  So it's rare you will see the plugin and Standalone rendering at exactly the same speed - since they are not rendering the same node setup.  If rendering the same node setup, the plugin and Standalone render at exactly the same speed (assuming you are using the same versions for Octane for each).

Paul

The standalone is displaying correctly the 6 GB. I need to find a large scene on the computer to check the poser plugin memory display.

By the way thanks for the new test build 1.11t and bringing the 1.11 version improvements to the plugin ;)

The standalone is displaying correctly the 6 GB. I need to find a large scene on the computer to check the poser plugin memory display.

OK, it's just a display error then - the plugin will be using the full 6GB.  I'll address the issue in the next patch.

By the way thanks for the new test build 1.11t and bringing the 1.11 version improvements to the plugin ;)

New features are pretty cool - I've been playing with fog.

Paul

Just some thoughts.

If you are using a video card for display and a GPU render, the reported memory difference could be correct due to hardware overlays that are already in video memory prior to launching the render engine. You would be shocked at how many programs grab video memory as soon as you open them. They have too, they are coded to use hardware overlays for the display.

Not using the rendering GPUs for display purposes will also allow programs written to load the cores effectively to work as they should.

To be fully effective, each GPU core has to be fed as many things as it can handle at once. It is not easy to get a GPU to be fully effective and at 100% load, adding the GUI to it drops that chance to all but zero. Most GPU render engines are coded to be the soul thing running on the GPU, take advantage of all the GPUs cores, and feed each core what it can do at the same time.

Based on the speed that Octane renders, it must be optimized extremely well, and taking full advantage of ordering and loading of the GPU cores.

If you can use a different GPU for the GUI, by all means do so. It will render faster, wont interfere with the GUI, etc..

You can even set up different hardware profiles to boot too if you need to use your good GPUs for games, editors, etc.

If you can use a different GPU for the GUI, by all means do so. It will render faster, wont interfere with the GUI, etc..

Yes - I think I practice this is a mandatory requirement for using Octane - runnning your Windows display adapter from the on-board graphics and letter Octane have 100% use of the GTX video card.  Otherwise you need to pause the render if you are going to do something GUI intensive.

Paul

How many texture slots/limits does the GTX Titan have?

Quote - How many texture slots/limits does the GTX Titan have?

224 If I remind.