V3Digitimes Scene Optimizer
marcelle19
Posts: 159
I had to go back to my old graphics card because my new card failed, so I've been using V3Digitimes Scene Optimizer in order to deal with large amounts of textures, lights, etc. Sometimes it works great; others, not so much. I have a Nvidia GeForce GTX 1660 Super, and it only has 6 GB VRAM.
My question is, for anybody experienced with this Scene Optimizer, what settings are best for renders with multiple assets like 4 or 5 characters, a building, at least 4 lights, and perhaps a few other things? These scenes were no problem for the card that died, but for the GTX, they overwhelm it, and rendering with the GPU assist doesn't turn out good renders, if it renders fast enough to call a render at all (seriously, I've had it go several minutes between rendered layers). So far, I've only messed with the most basic settings in the Scene Optimizer.
Any help would be appreciated. Thank you.
Daz 3D is part of 
Connect
DAZ Productions, Inc.
224 S 200 W, Suite #250
Salt Lake City, UT 84101
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
As for your case, 6GB VRAM.. Assuming you only have Daz Studio opened, the windows / application will consume appr. 0.5GB VRAM, iray render engine will consume appr. 2 - 2.2GB, so there'll be around 3GB VRAM left for rendering.
Since you have 4 - 5 different characters in the scene, you have to roughly estimate the VRAM consumption in terms of texture maps specifically. For instance, a typical G8F character（non Base Figure）has around thirty 4K maps, one 4K map will consume 64MB VRAM （roughness, normal etc. won't consume that much）, so more or less one figure will consume 0.8 - 1 GB VARM after texture compression.. If you have 4 figures, even you make them undressed, you will run out of VRAM, let alone other objects / wearables in the scene...
Therefore first you may use Scene Optimizer to just decrease the size of all 4K texture maps, i.e. divide them by 2 or 4, to see if Ds can still use GPU to render. If the render is rolled back to CPU, you may further lower the Render SubD level. Anyway you probably have to give it a couple of tries by using different settings in SO...
Besides, for the figures who're far from the camera or out of the focus pane (if you use DOF), you may further divide their tex. maps by 8 then see the results... or use Resource Saver Shaders products, even Billboard figures if possible...
If it still does not work, I could only sugget you just have 1 - 2 figures in one scene and try not to use high definition / high details wearable / environments, etc.
Thank you. That was very comprehensive.
Also in render settings > optimization:
Also you may connect the monitor to the mobo and in windows set the graphics settings to power saving for daz studio, this way the viewport will use the integrated gpu and the 1660 is reserved to iray for the full 6G.
BTW, I just wonder why your new card failed...
So do I. It worked for around 3 months, then went blooie! and almost took my computer with it. I'm getting read to try it again, after making sure nothing loose in the computer (what I heard could have been a small, loose part, like a screw, that fell into the fan and caused all the trouble) and a memtest to make sure the memory is working right.
The monitor is connected to the mobo by default, is it not? I mean, the mobo is where the GPU is - not sure what you are saying, here. I get the instancing optimization and ray tracing low memory, but the last part, not really sure?
When you install a video card you can connect the monitor to the card, so windows will use the card as output, and 3d apps will use the card for the viewport. If you connect the monitor to the mother board the cpu integrated graphics will be used instead.
On 6GB GTX card, there isn't much space left for Iray rendering.
W10 takes 1GB (may have come down since last summer), DS the application takes some, the scene takes some, rendering Iray needs a "Working space" in VRAM, which by default is 1.9GB's on newer versions of DS - All of this comes to about 3.5-4GB's of VRAM, but non-RTX card users are then hit with an additional 1GB loss for emulation of the RTX functions, making the total in 4.5-5GB range...
When the scene starts filling the available VRAM, DS can lower the amount of reserved "Working space", but when it gets below 1GB it starts slowing down the rendering speed.
Using the motherboards integrated GPU to drive the monitor(s) has been suggested as an alternative, but nobody so far has shown any evidence of this freeing up VRAM on the GPU used for rendering and thus allowing one to render bigger scenes.
Here's my evidence. I get 0.0/6.0 GB vram used with daz studio on the 1060, since I use the ryzen apu for the viewport. The whole 6 GB are reserved to iray and it's zero until I render. Same when I work with blender and cycles.
I have one question ~ You're using ver 4.15 ? Any problem when rendering with driver 526.98 ?
The Task Manager is not that good in giving accurate information GPU-Z is better. With evidence, I meant proof at different stages of the process - VRAM usage before opening DS, after opening a scene, during rendering and the amount of VRAM used for textures and geometry as it's reported by the DS 4.15 log, showing that the amount is higher that would be possible with a GPU that's driving the monitors.
@crosswind Yes I keep 4.15 to have the iray materials and ghost lights working with old assets, since 4.21 broke them. No problems with assets updated by the PAs but not all of them are. I keep 4.21 as beta so I can use both. No issues here with 526.98 drivers but it may also depend on the card, a 1060 is old and thus probably more stable and less demanding in terms of drivers than a 4070 for example.
@PerttiA I agree nothing of what you say. To me it is evidence enough. The used vram is zero until I render with iray.
@Padone A friend of mine has the similar situation and wanna keep 2 versions of DS after upgrading the driver. Now it sounds there'll be no big issue. Thank you for the answer !
Too bad, since that would have been interesting information to give to ones strugling with low VRAM GPU's
This is a chart I made when I still had my old 8GB RTX 2070 super, and I wanted to know how much RAM and VRAM was used at which stage and when would the rendering drop to CPU.
Case a) just one lightweight G8 figure with lightweight clothing and hair
Case b) four similar G8 characters with architecture
Case c and d) started increasing SubD on the characters to see at which point the rendering would drop to CPU
"RAM/GB" and "VRAM/MB" taken from GPU-Z, "DS Log/GiB" taken from DS Log, no other programs were running but DS and GPU-Z
The "DS Log/GiB" is the sum of Geometry usage, Texture usage and Working Space - After Geometry and Textures, there should still be at least a Gigabyte of VRAM available for the Working space => In my case, Geometry + Textures should not exceed 4.7GiB
Note; Case c) was already using 38GB's of RAM, even though the rendering was done on the GPU, Case d) when rendering on CPU the RAM usage went almost over my 64GB's
Tests made using RTX 2070 Super (8GB), i7-5820K, 64GB's of RAM on W7 Ultimate and DS 4.15
Your case is different you use the 2070 for the viewport. Otherwise, again, you have zero vram until you render. Other numbers are meaningless since they depend on the OS version, drivers version, daz studio and iray version, used test scene and rendering parameters. What does matter is that you reserve the card to iray so windows and the viewport will not use it.
But until someone has tested and provided the results of what is happening at different stages of the process, we don't know how much of the VRAM is actually available and/or used during Iray rendering or what is the limit when using a GPU that's not used to drive the displays.
I would do it, but I'm still using W7 and I'm not sure my 750W PSU could handle adding the RTX 2070 Super next to the RTX 3060 12GB, as the system is quite loaded already even without putting an other GPU in it.
And I never use motherboards with integrated GPU's, so using one is not an option.
For others, the problem is lack of information in the log, as Nvidia removed the information about the amount of VRAM used by textures and geometry in some version after DS 4.15.0.2
Even if we had proof of what is the maximum amount of textures and geometry, which such GPU can handle, that would already prove that the consept does work and it would give some idea of how big of a scene such GPU could handle, but so far nobody has bothered testing and/or providing such information here on the forums.
There was one important test case missing in your table - Empty Scene.... from Idle > Ds started > Preview or Render Empty > Stop preview or Close render window > Ds complete shut... Only with this test, you can clearly tell - the VRAM consumption from Iray Engine itself, and when this part of memory usage can be completely released, etc. other than mixing them up.... Again, the test results still depend on those factors as above-mentioned...
I ever organized nearly 50 people to test all these by using varieties of hardwares / Ds Ver / Driver Ver, etc. then I was more confident in giving my online tutorials of Scene Optimization ... Btw, pls DON'T fight back, that was my way and just tell you the truth...
Ok, that stage was missing, but with W7 + DS 4.15, the amount of reserved RAM and VRAM was released completely once DS had finished it's shut-down process.
On DS 4.15, using the last W7 supported drivers, there has never been a problem with VRAM or RAM not being released, not between different scenes and not between renders, I can render scenes as many times as I want without having to close DS between renders.
My chart does show that even after opening a new scene, some VRAM was still reserved, but when a new render was started, the old reservation did not prevent the new render from taking all the necessary VRAM it needed
NP~ these points were almost as same as what I ever observed. PS: If there're multi-instances of Ds (no matter GR + PB, or GR[0], GR[1]...), iray engine will 'eat' VRAM when render in each instance.. I often work with multi-sessions but still have to shut instance when rendering big scenes, even if there're 48GB on my display card.. Anyway, there'll be always some resoltuions leading to happy rendering ~
Wow, this thread took on a life of it's own.
I don't even know if my motherboard has onboard video. Pretty sure it doesn't, though.
Okay, I see what you are saying now. I've always had my GPU connected to the monitor; I never connected directly to the motherboard. That would pretty much defeat the purpose of even having a separate GPU.
Thank you for you responses.