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
© 2025 Daz Productions Inc. All Rights Reserved.
Comments
.
Edited to combine quote and reply
Again, did you find a "GiB available" number, like I posted? That's what shows what's actually available for rendering.
Decisions, Decisions. I currently use Windows 7 pro have a GTX 1080ti GPU. Support for Windows 7 will expire in 2020 (less than 3 years) and I only have until the end of December to get Windows 10 Pro free. Should I upgrade and sacrifice the VRAM that Windows 10 will take or should I keep Windows 7and worry about no more security updates after 2020?
Any recommendations?
.
OTOY, the developer of Octane Renderer has a thread on their forums, and there is a similar thread on the Microsoft TechNet forums about this. And nVidia has been involved in the Microsoft forum discussion, from what I remember.
Windows 10 fall update has new GPU stuff in Task Manager.
https://blogs.msdn.microsoft.com/directx/2017/07/21/gpus-in-the-task-manager/
It's a bit weird, I can render at 100% utilization in Iray and Task Manager still shows around 45%. I think it's because it factors in other stuff that the GPU could do besides 3D. I have 3 more graphs there, Copy, Video Encode and Video Decode. Other than that though, VRAM numbers seem to make some sense. I start rendering, Daz Studio shows (under Details, need to add the GPU columns yourself) as 5gb dedicated vram. Then there's 540mb Desktop Window Manager and 225mb some Client Server Runtime process. Few other things take a little bit, like Chrome but nothing dramatic. So then the total dedicated vram is 5.4gb and corresponds pretty much with what GPU-Z shows (5660mb). When I exit the render window, Daz Studio now shows 2.5gb dedicated vram still. When I restart Daz Studio with the same scene it's 266mb. Now when I render again it's 3.8gb.
So the difference is, the first time Daz Studio had been open and rendering for a quite a while. I guess some stuff just remains on the GPU for efficency. Either that or it's just completely leaky/bugged.
davegv: Dude, seriously, you should read the thread if you want the facts. It's mentioned in the Microsoft Developers' Network forum, in various professional renderer forums, and so on.
If you have texture compression running,, which most people tend to do, it could be that the reported Size of the textures loaded to the GPU is BEFORE compression come render time.., I have had scenes which report, for example 7.65 gb of textures loaded plus other elements yet they have rendered on a card with 6GB of Vram.
S.K.
First, Please do not refer to me as "Dude" Sounds childish. I am 62 yrs.
I have read this entire thread and I will repeat I am satisfied that it is not as big a issue as YOU make it out to be.
The only rendering forum it is mentioned in is OCTANE forum which is not and should not have any bearing on the way IRAY handles it's memory consumption.
NVidia which created IRAY has not directly addressed the issue you keep pressing. Neither Has Microsoft. Does this not say something to you?
For me, as I said, a little common sense says it is not a big issue. You keep trying to beat a dead horse.
The data in the Task Manager is gathered directly from VidSch and VidMm. As such, performance data for the GPU is available no matter what API is being used, whether it be Microsoft DirectX API, OpenGL, OpenCL, Vulkan or even proprietary API such as AMD's Mantle or Nvidia's CUDA. Further, because VidMm and VidSch are the actual agents making decisions about using GPU resources, the data in the Task Manager will be more accurate than many other utilities, which often do their best to make intelligent guesses since they do not have access to the actual data.
Well, If the Windows resource manager ( Task Manager ) GPU Memory Useage indicates my GTX 1080 is using 7.86 GB of VRam
Well, If GPU z is indicates my GPU Memory Useage indicates my GTX 1080 is using 7.86 GB of VRam
Yea, I am going to decide FOR MYSELF that I am satisfied that it is in fact useing nearly all of my 8GB.
It is all explained Here:
https://blogs.msdn.microsoft.com/directx/2017/07/21/gpus-in-the-task-manager/
I also Use NVidia IRay within 3DS MAX ( Autodesk 2017 ) and again I can load a entire scene up to 8GB in VRAM and it will show that IRay and my VRam are utilizing nearly all of the allocated 8GB on my GTX1080.
Edited to combine quote and reply
In any case, read that article. There's some shared memory stuff going on. Also this:
With Daz Studio closed I get 0.7 of 11.0 gb vram used. I run 3 monitors btw.
davegv, I apologize for the "dude", but if you're 62 years old I'm sure you recall that "dude" is a good thing, right? It's for your buddies. Y'know, like "Dude, wazzup?".
Anyway, I really don't care what you believe, and I'm not trying to make anything into a big issue. I'm just telling you the facts, so it's up to you to decide what you believe. If you don't want to believe the Microsoft forums or other forums then that's fine with me.
Just don't be surprised if others push back when you claim it's not an issue. Because for many 3D folks out there it's a BIG issue.
Over 34,000 links about this subject
Incorrect; both Iray and Octane use CUDA on nVidia GPUs, so exactly the same issue exists for both since both are constrained by the memory allocation system used.
Microsoft has stated "working as intended", since in their view, any use of a GPU with a video display output should be prepared to have a monitor connected at any time. If you want all the VRAM, use a device without video display support, is their view a
There is another thread (somewhere) where it was clearly determined that the Texture Buffer numbers in the DS log file are pre-compression (i.e., raw sizes). People were seeing values reported that were larger than their VRAM, but still rendered in Iray on GPU.
If you are determining your VRAM usage from that, you are using incorrect numbers.
I just bought a new PC with Win10 Pro and a 1080ti (Yay!
) that is also driving my monitor. Up to this point I've rendered in CPU mode, so I'm new to GPU rendering. From the log file in DS, with no scene loaded:
"2017-11-10 10:04:26.986 Iray INFO - module:category(IRAY:RENDER): 1.0 IRAY rend info : CUDA device 0 (GeForce GTX 1080 Ti): compute capability 6.1, 11 GiB total, 9.14933 GiB available, display attached"
Even when I remove the eye candy in Visual Effects by choosing "Best Performance" instead of "Let Windows Decide" and close/open DS, the above numbers in the log file do not change. So 1.85 GiB is not available?!
At this point, I'm not even sure how large a scene I would need to make to exceed the available 9.14 GiB, but I just wanted to jump in and join this discussion, and to see if there is any solution to regain 1.85 GiB without switching OS, adding another card, etc.
This would be a valid question for the NVidia IRay Forum as perhaps it is Iray itself that is re-allocating the memory or perhaps even Daz3D?
I am convinced it is NOT windows 10 that is useing excessive VRam. It certainly is not on my system.
The data in the Task Manager is gathered directly from VidSch and VidMm. As such, performance data for the GPU is available no matter what API is being used, whether it be Microsoft DirectX API, OpenGL, OpenCL, Vulkan or even proprietary API such as AMD's Mantle or Nvidia's CUDA. Further, because VidMm and VidSch are the actual agents making decisions about using GPU resources, the data in the Task Manager will be more accurate than many other utilities, which often do their best to make intelligent guesses since they do not have access to the actual data.
Well, If the Windows resource manager GPU Memory Useage indicates my GTX 1080 is using 7.86 GB of VRam
Well, If GPU z is indicates my GPU Memory Useage indicates my GTX 1080 is using 7.86 GB of VRam
Yea, I am going to decide FOR MYSELF that I am satisfied that it is in fact useing nearly all of my 8GB.
Edited to combine quote and reply
Lets start directing out comments to the topic at hand instead of each other per the TOS
What exactly does this mean?
OK, I have not seen or read anything that would imply arguments or "Fighting" Just normal forum banter
Edited to combine quote and reply
To talk about the topic instead of fighting or arguing with each other. Civilly talking with each other is fine.
Consolidated split posts and removed duplicates.
Yes, The GPU does not "Flush The VRam" and only re-starting daz3d will clear the VRam Buffer.
(Edited to combine quote and reply)
Ok so I added a ton of stuff to a scene to see what happens in the new Task Manager. Dedicated GPU memory of Daz Studio alone at 9,142,052 kb, dedicated total like 9.4gb. So this should render on the 1080 Ti with 11gb vram.
Aaaaaaaand... drum roll. It didn't. Clearly that means that the new Task Manager is ommitting the fact that a certain portion of the VRAM is actually not usable. Even if it's technically free, it doesn't matter much when you can't use it.
Given what I just wrote, this confuses me though.
...there is full NVLink connectivity between both CPU-GPU/GPU-GPU but it is only available for Data Centre Servers and Supercomputers using either the Tesla P100 or V100. Depending on the model, each NVLink node can support either 4 or 8 Tesla compute GPUs, thats 64 - 128 GB of HBM 2 memory per node. High end workstation NVLink boards are probably not far behind especially if Nvidia offers an NVlink version of the Quadro P100 to complement it's PCIe version (as they did with both the Tesla P100 & V100). Again, most likely these boards will be expensive compared to even the best PCIe 3,0 ones. The current NVLink modules available for 4 Tesla P/V100s start about 3,500$ (sans cards) and go up from there (around 50,000 - 70,000$ for OCT's HGX-1 and Big Basin chassis without the GPUs).
A set of 2 NVLink connectors (you need 2 for memory pooling) for the current PCIe Quadro P100 are priced at around 1,000$.
Thanks and this is very usefull to an issue a had allong time ago.
I use to render with Luxrender using an AMD R295X2
The 295X2 was a Dual GPU card that had 4 GB of Vram Per GPU.
In windows 8 it ran 4 gb per GPU, then after updating to Windows 10, it went to something like 3.3 GB per GPU.
Then after trying manny driver updates/reverts etc to get that 4gb per gpu back. AMD updated something specificly for OpenCL that allowed both GPU's to share memory, so for Luxrender I was able to use 6.6GB of ram, at that point i stoped messing with it. but that would explain why my current setup with a GTX980 and GTX 1070, the 980 drops out of the rendering a little over 3GB.
...wait, there's and OpenCL driver update that lets you pool memory on a dual GPU AMD card?
not sure if it was intentional, but I was able to load up a 6.6gb scene inside the 295x2 and render fine, wasn't rendering slow or anything, rendered like normal, but was able to use allot more textures.
kind of why i really miss rendering with luxrender, that card was a beast for it. but something happened either inside lux, or luxus, or opencl, where it just all of a sudden sucked up all my system ram, started paging, then crashed, and thats not a small task with 32gb of system ram.
...yeah that's another reason why I dropped Reality/Lux. I have only about 10.5 GB of available physical memory (after Windows and utilities). While it never crashed, rendering was on a geologic time scale as it went to virtual memory. Using the speed boost sacrifices quality.
I certainly hope you're right. Granted I haven't followed this issue in the industry, so maybe there's some new stuff that came out that said it's not an issue any longer. Do you have any industry references (NVIDIA, Microsoft, etc.) that explain why it's not an issue? I'm a C# developer and have done a little graphics work, but this kind of GPU/memory allocation stuff gets awful complicated and I tend to rely on the experts for answers on these types of things. One of my goals is to start to learn GPU programming, but that's a whole new world of stuff to figure out.
It would be great if you could lay everyone's fears to rest. I have a brand new 1080ti, and I hate seeing only 9GB of my 11GB VRAM available in the Iray log.
I can live with the fact that my displays are going to need some VRAM. It's only logical. At this point I really don't understand one thing, and that is why cards with more VRAM have to sacrifice more for the displays. That just doesn't make sense, it's a bit like that stupid virtual memory myth where people told you to have as much virtual as physical memory. Uh yeah ok, if I have 128gb physical memory, sure I'll need to add 128gb virtual.