I need a little guidence on setting up for GPU renders

[rimcrazy]

So I just upgraded my system from an "old" K5000 card to 2x GTX1080's. Using Richard Yot's example files for rendering in Modo I saw about a 5x improvement in render speed using GPU's vs CPU on my system. I then went over to a VRay Maya scene that I'm working on for a client. The client wants my entire project, about 6000 frames, rendered in 4K, which is the primary reason I upgraded to the new GTX1080 cards. Well, on first pass things are not so stellar. CPU renders on a 4K frame in Maya 2016.5 render in 1 minute. The GPU render took just under 3mins. Clearly I'm doing something wrong with my scene. Let me preface that while I've used VRay off/on for a while I'm by no means a wizard with it. If someone could offer a few words of wisdom as to what I should be looking at to optimize my renders it would be greatly appreciated. All things being equal, I'm assuming I should be looking at getting around 12 seconds if all things can be equal in Maya as I see in Modo.

Tnx

[savage309]

It is a bit hard to guess what is going on.

Can you do the following for a test - use RT as IPR, set time limit to 20 seconds, set rays per pixel to 128 and ray bundle size to 256 and the noise level to 0.
The render should stop after 20 seconds ... If the quality of the result seems nice, than we are good and you should set similar settings to the production mode too.
If not, we will have to take a look to see what is going on.

Best,
Blago.

[rimcrazy]

Problems with GPU/CPU renders in VRay Maya.

Hi Blago,

Thanks for your help. I followed your instructions but I don't see any difference. I did two batch runs, one with CPU and one with GPU and created log files but they are too large to include and you don't allow attachments. I uploaded the zip file to my website and you can download the zip from the following link:

http://philfx.com/wp-content/uploads...7/LogFiles.rar

I also did a screenshot of what I believe you asked me to do. The first screenshot is my render settings, the second is a screenshot doing a GPU render in the VFB window and then a CPU render in the VFB.
I shrunk the size down from 4K to 960x540 just to make things run faster. As you can see from the log files, GPU render/frame (frame1) = 42.1 seconds, CPU render/frame (frame1) = 11.5 seconds

Looking at the log files it appears that VRay does see both of my GTX1080 cards. For the batch render, the render settings (second tab on the Render Settings Menu) are identical. All I did was change the drop down select from CPU to GPU and redid the batch.



UPDATE: I thought it may be scene related so I loaded in a much simpler scene that I had and tried again. This time CPU render was just about equal to GPU with GPU being about 5% slower. Still not near what I would have expected and quite far from the Modo VRay results I see.

[Svetlozar Draganov]

What was the GPU utilization on both scenes?
Did V-Ray utilize them up to 100% or not? If not can you show is a graph that shows the GPU usage during the whole frame rendering?

[savage309]

If you can share the scenes, this would help to debug what is going on.

Best,
Blago.

[rimcrazy]

Here is the header from the log and the render of 1 frame. As far as I can tell from the log they were both utilized, all be it appears R/W time far exceeds compute time. Still that being the case why such a disparity between Modo and Maya? In a sample of 1 scene (Richard Yot's example file from your help file) I'm seeing a 5x improvement of GPU vs CPU using Modo. In 2 different Maya files I'm seeing CPU faster than GPU in both cases.

V-Ray: Starting render
[2016/Jul/24|16:38:17] V-Ray: Exporting scene to V-Ray.
[2016/Jul/24|16:38:17] V-Ray: Parsing light links time 0h 0m 0.0s (0.0 s)
[2016/Jul/24|16:38:17] V-Ray: Translating scene geometry for V-Ray
[2016/Jul/24|16:38:17] V-Ray: Total time translating scene for V-Ray 0h 0m 0.0s (0.0 s)
[2016/Jul/24|16:38:17] V-Ray: Rendering.
[2016/Jul/24|16:38:17] V-Ray: Clearing bitmap cache memory...
[2016/Jul/24|16:38:17] V-Ray: Rendering frames.
[2016/Jul/24|16:38:17] V-Ray: Updating frame at time 1.00
[2016/Jul/24|16:38:17] V-Ray: Total time updating frame 0h 0m 0.0s (0.0 s)
[2016/Jul/24|16:38:17] V-Ray: Preparing renderer...
[2016/Jul/24|16:38:17] V-Ray: Preparing scene for rendering...
[2016/Jul/24|16:38:17] V-Ray: Creating texture cache with size 1000 MB
[2016/Jul/24|16:38:17] V-Ray: Creating bitmap manager
[2016/Jul/24|16:38:17] V-Ray: 5 lights found.
[2016/Jul/24|16:38:17] V-Ray: CUDA renderer requested.
[2016/Jul/24|16:38:17] V-Ray: 1 plugin(s) loaded successfully
[2016/Jul/24|16:38:17] V-Ray: CUDA renderer plugin successfully loaded from "C:\Program Files\Autodesk\Maya2016.5\vray/bin/rt_cuda.dll"
[2016/Jul/24|16:38:17] V-Ray: CUDA renderer plugin instance successfully created.
[2016/Jul/24|16:38:17] V-Ray: EXT_RTOPENCL interface obtained successfully from CUDA renderer plugin instance.
[2016/Jul/24|16:38:17] V-Ray: Max ray intensity is enabled: rendered result may have incorrect brightness.
[2016/Jul/24|16:38:17] V-Ray: Preparing scene for frame...
[2016/Jul/24|16:38:17] V-Ray: Compiling geometry...
[2016/Jul/24|16:38:17] V-Ray: (Estimated time remaining: 0h 0m 0.0s): 64 % completed
[2016/Jul/24|16:38:19] V-Ray: (Estimated time remaining: 0h 0m 1.1s): 71 % completed
[2016/Jul/24|16:38:20] V-Ray: (Estimated time remaining: 0h 0m 0.6s): 83 % completed
[2016/Jul/24|16:38:20] V-Ray: (Estimated time remaining: 0h 0m 0.3s): 90 % completed
[2016/Jul/24|16:38:20] V-Ray: Using embree ray tracing.
[2016/Jul/24|16:38:20] V-Ray: Building static raycast accelerator...
[2016/Jul/24|16:38:20] V-Ray: SDTree statistics:
[2016/Jul/24|16:38:20] V-Ray: Total number of faces stored: 14354
[2016/Jul/24|16:38:20] V-Ray: Max tree depth: 36
[2016/Jul/24|16:38:20] V-Ray: Average tree depth: 16.4066
[2016/Jul/24|16:38:20] V-Ray: Number of tree nodes: 12701
[2016/Jul/24|16:38:20] V-Ray: Number of tree faces: 53316
[2016/Jul/24|16:38:20] V-Ray: Number of tree leafs: 5101
[2016/Jul/24|16:38:20] V-Ray: Average faces/leaf: 10.4521
[2016/Jul/24|16:38:20] V-Ray: Memory usage: 2.01 MB
[2016/Jul/24|16:38:20] V-Ray: Building static SD trees took 3 milliseconds
[2016/Jul/24|16:38:20] V-Ray: Running RTEngine
[2016/Jul/24|16:38:20] V-Ray: Initializing CUDA renderer (single kernel version)...
[2016/Jul/24|16:38:20] V-Ray: Querying for CUDA devices...
[2016/Jul/24|16:38:20] V-Ray warning: Devices query yielded no results; using all available devices!
[2016/Jul/24|16:38:20] V-Ray: Device GeForce GTX 1080 has compute capability 6.1
[2016/Jul/24|16:38:20] V-Ray: Setting CUDA stack size to 14000
[2016/Jul/24|16:38:20] V-Ray: Device GeForce GTX 1080 has compute capability 6.1
[2016/Jul/24|16:38:20] V-Ray: Setting CUDA stack size to 14000
[2016/Jul/24|16:38:20] V-Ray: Building optimized CUDA kernel (0) for comp cap 6.1 for device 0...
[2016/Jul/24|16:38:20] V-Ray: PTX file generated with CUDA Toolkit v7.5 for CUDA compute capability 2.0
[2016/Jul/24|16:38:21] V-Ray: Optimized CUDA kernel assembled successfully
[2016/Jul/24|16:38:21] V-Ray: Total memory for device 0 : 8192 MB, free 6275 MB
[2016/Jul/24|16:38:21] V-Ray: Building optimized CUDA kernel (0) for comp cap 6.1 for device 1...
[2016/Jul/24|16:38:21] V-Ray: PTX file generated with CUDA Toolkit v7.5 for CUDA compute capability 2.0
[2016/Jul/24|16:38:22] V-Ray: Optimized CUDA kernel assembled successfully
[2016/Jul/24|16:38:22] V-Ray: Total memory for device 1 : 8192 MB, free 6279 MB
[2016/Jul/24|16:38:22] V-Ray: Bindless textures not supported.
[2016/Jul/24|16:38:22] V-Ray: Maximum kernel work group size: 1024
[2016/Jul/24|16:38:22] V-Ray: Maximum kernel work group size: 1024
[2016/Jul/24|16:38:22] V-Ray: Initializing environment kernel
[2016/Jul/24|16:38:22] V-Ray: Number of CUDA devices: 2
[2016/Jul/24|16:38:22] V-Ray: nVidia driver version: 368.81
[2016/Jul/24|16:38:22] V-Ray: Building dynamic mesh trees
[2016/Jul/24|16:38:23] V-Ray: Building trees for 17 meshes with all threads per mesh took 842ms.
[2016/Jul/24|16:38:23] V-Ray: Building BVH tree
[2016/Jul/24|16:38:23] V-Ray: BVH statistics:
[2016/Jul/24|16:38:23] V-Ray: Primitives count: 17 (Mem usage:0.00 MB)
[2016/Jul/24|16:38:23] V-Ray: Meshes count: 17 (Mem usage:0.00 MB)
[2016/Jul/24|16:38:23] V-Ray: Tree depth: 6
[2016/Jul/24|16:38:23] V-Ray: Number of tree nodes: 33
[2016/Jul/24|16:38:23] V-Ray: Tree takes 0.58 MB. Build time 0.13 ms.
[2016/Jul/24|16:38:23] V-Ray: Texture size set to 510
[2016/Jul/24|16:38:23] V-Ray: GPU loaded bitmap "D:/projects/sandbox/Barrows/Rad_MIS_Surgery/Maya/Fluoroscope/sourceimages/flouroscope_screen.jpg"
[2016/Jul/24|16:38:23] V-Ray: Tracing 250000 image samples for light cache in 64 passes.
[2016/Jul/24|16:38:23] V-Ray: Building light cache...
[2016/Jul/24|16:38:24] V-Ray: Merging light cache passes...
[2016/Jul/24|16:38:24] V-Ray: (Estimated time remaining: 0h 0m 0.1s): 11 % completed
[2016/Jul/24|16:38:24] V-Ray: (Estimated time remaining: 0h 0m 0.1s): 20 % completed
[2016/Jul/24|16:38:24] V-Ray: (Estimated time remaining: 0h 0m 0.0s): 30 % completed
[2016/Jul/24|16:38:24] V-Ray: (Estimated time remaining: 0h 0m 0.0s): 41 % completed
[2016/Jul/24|16:38:24] V-Ray: (Estimated time remaining: 0h 0m 0.0s): 50 % completed
[2016/Jul/24|16:38:24] V-Ray: (Estimated time remaining: 0h 0m 0.0s): 60 % completed
[2016/Jul/24|16:38:24] V-Ray: (Estimated time remaining: 0h 0m 0.0s): 71 % completed
[2016/Jul/24|16:38:24] V-Ray: (Estimated time remaining: 0h 0m 0.0s): 80 % completed
[2016/Jul/24|16:38:24] V-Ray: (Estimated time remaining: 0h 0m 0.0s): 90 % completed
[2016/Jul/24|16:38:24] V-Ray: (Estimated time remaining: 0h 0m 0.0s): 100 % completed
[2016/Jul/24|16:38:24] V-Ray: Light cache contains 4752 samples.
[2016/Jul/24|16:38:24] V-Ray: Light cache takes 6.6 MB.
[2016/Jul/24|16:38:24] V-Ray: Prefiltering light cache...
[2016/Jul/24|16:38:24] V-Ray: Average rays per light cache sample: 14.14 (min 1, max 223)
[2016/Jul/24|16:38:24] V-Ray: Using premultiplied light cache.
[2016/Jul/24|16:38:24] V-Ray: Lightmap successfuly loaded.
[2016/Jul/24|16:38:24] V-Ray: GPU light cache: 4752 samples stored.
[2016/Jul/24|16:38:24] V-Ray: GPU light cache: 850 nodes stored.
[2016/Jul/24|16:38:24] V-Ray: GPU light cache: 851 leaves stored.
[2016/Jul/24|16:38:24] V-Ray: Mtls 8.00MB, Geo 291.25MB, Lights 0.00MB, LightCache 0.34MB, Other 64.31MB (data index 0)
[2016/Jul/24|16:38:24] V-Ray: (Estimated time remaining: 0h 0m 32.5s): 99 % completed
[2016/Jul/24|16:38:24] V-Ray: Building optimized CUDA kernel (0) for comp cap 6.1 for device 0...
[2016/Jul/24|16:38:24] V-Ray: PTX file generated with CUDA Toolkit v7.5 for CUDA compute capability 2.0
[2016/Jul/24|16:38:24] V-Ray: Optimized CUDA kernel assembled successfully
[2016/Jul/24|16:38:24] V-Ray: Total memory for device 0 : 8192 MB, free 5898 MB
[2016/Jul/24|16:38:24] V-Ray: Building optimized CUDA kernel (0) for comp cap 6.1 for device 1...
[2016/Jul/24|16:38:24] V-Ray: PTX file generated with CUDA Toolkit v7.5 for CUDA compute capability 2.0
[2016/Jul/24|16:38:25] V-Ray: Optimized CUDA kernel assembled successfully
[2016/Jul/24|16:38:25] V-Ray: Total memory for device 1 : 8192 MB, free 5901 MB
[2016/Jul/24|16:38:25] V-Ray: Progressive sampling base buffer memory usage: 71.43 MB
[2016/Jul/24|16:38:26] V-Ray: Noise threshold lowered to 0.166667 with 24.3 percent active pixels.
[2016/Jul/24|16:38:30] V-Ray: Noise threshold lowered to 0.125000 with 11.7 percent active pixels.
[2016/Jul/24|16:38:31] V-Ray: Noise threshold lowered to 0.083333 with 1.4 percent active pixels.
[2016/Jul/24|16:38:32] V-Ray: Noise threshold lowered to 0.050000 with 1.6 percent active pixels.
[2016/Jul/24|16:38:33] V-Ray: Noise threshold lowered to 0.027778 with 5.4 percent active pixels.
[2016/Jul/24|16:38:37] V-Ray: Noise threshold lowered to 0.014706 with 13.7 percent active pixels.
[2016/Jul/24|16:38:43] V-Ray: Final noise threshold set: 0.010000
[2016/Jul/24|16:38:58] V-Ray: Device 0 statistics:
[2016/Jul/24|16:38:58] V-Ray: GPU time: 89.599148 %
[2016/Jul/24|16:38:58] V-Ray: GPU ray trace time 0.639756 %
[2016/Jul/24|16:38:58] V-Ray: Transfer to GPU time 0.017800 %
[2016/Jul/24|16:38:58] V-Ray: Transfer from GPU time 99.342443 %
[2016/Jul/24|16:38:58] V-Ray: Device 1 statistics:
[2016/Jul/24|16:38:58] V-Ray: GPU time: 81.287826 %
[2016/Jul/24|16:38:58] V-Ray: GPU ray trace time 1.124920 %
[2016/Jul/24|16:38:58] V-Ray: Transfer to GPU time 0.032292 %
[2016/Jul/24|16:38:58] V-Ray: Transfer from GPU time 98.842788 %
[2016/Jul/24|16:38:58] V-Ray: Closing RTEngine
[2016/Jul/24|16:38:58] V-Ray: Max. Mpaths per second: 394.319489
[2016/Jul/24|16:38:58] V-Ray: Sampling level reached: 10879 samples per pixel
[2016/Jul/24|16:38:58] V-Ray: RT Frame took: 33.03s
[2016/Jul/24|16:38:59] V-Ray: Successfully written image file "D:/projects/sandbox/Barrows/Rad_MIS_Surgery/Maya/Fluoroscope/images/masterLayer/MobileFluoroscope_VRay_Animation.01.exr"

[savage309]

It will be more of a guess to give you more advices only from the logs. Is it possible to send the scene to support@chaosgroup.com with link to this threads so we can check here what is going on ?

Thanks,
Best,
Blago.