Slow render with GTX 780 ti (VRay for Rhino 2.0)

[Jenzie]

Hi,

I just upgraded my workstation with a GTX 780 ti from a Quadro 600 to speed up RT renderings but I suspect something is wrong because I can't really tell if there's any difference.
Just to see if the problem are among the drivers or hardware I downloaded Octane but there everything is blazing fast, I can render a whole train car body within seconds (2-3) while Vray needs about 5+ min (I didnt even wait for it to finish, took too long time) for the same.

I can see in the render node it's using CUDA and the 780 ti.

When I installed I removed the old drivers and did a restart and then installed the new (and my windows is disabled from installing drivers by itself).

I have tried different settings, CUDA and OpenCL, lowering the mesh density. Nothing helps.

Any tips appriciated.

Edit: I'm using Win7 64, Rhino 5 and Vray 2.0. Everything is updated today.

Kind regards,
Jens

[vlado]

Are you sure it's using the GPU? By the looks of it, it might be falling back to the CPU for some reason.

Best regards,
Vlado

[joe_b]

We could tell if it was falling back on CPU by checking the output of the spawner. I am assuming you are using your local machine for the RT renders. Would you mind doing the following for me:

1. Close all instances of Rhino
2. Manually open our spawner ( Start->All Programs->Chaos Group->VRay for Rhino 5 x64->Distributed Rendering )
3. Open Rhino and start an RT Cuda render
4. Copy and paste the output from XMLDRSpawner so we can see what is happening ( You can copy all the output by clicking on the icon at the top left of the spawner window, then selecting "Edit->Select all" and then pressing "Enter" when all the text is highlighted) .

If it is failing to initialize cuda it should log that in the spawner's output.

[Jenzie]

Hi, thank you for helping.

It lost the top of the text, but here it goes.

CUDA renderer requested.
CUDA renderer plugin already loaded.
CUDA renderer plugin instance successfully created.
EXT_RTOPENCL interface obtained successfully from CUDA renderer plugin i
nstance.
[RenderView] startCameraTime=0.000000, endCameraTime=1.033333
[RenderView] numCameraTMs=62, numFrames=31, frameSamples=2
Preparing camera sampler.
Preparing scene for frame...
Mesh file "M:/Projects/X/CAD/stol-low.vrmesh" link
ed successfully.
Mesh file "M:/Projects/X/CAD/stol-high.vrmesh" lin
ked successfully.
Setting up 8 thread(s)
Bitmap file "C:/ProgramData/ASGVIS/V-Ray Express/Rhino/HDRI/Stud
io_16.hdr" loaded.
Threads completed
Compiling geometry...
Preparing ray server.
Building SDTree for GPU
Creating and initializing 8 thread(s).
Allocating memory for build data of 839937 faces (6719496 bytes,
6.4 MB).
Initializing face build data.
Creating 'done' event.
Starting first thread.
Waiting for thread completion.
Releasing thread resources.
SDTree statistics:
Total number of faces stored: 839937
Max tree depth: 47
Average tree depth: 24.3693
Number of tree nodes: 518799
Number of tree faces: 3252212
Number of tree leafs: 208265
Average faces/leaf: 15.6157
Memory usage: 90.43 MB
Building SDTree for GPU
Scene bounding box is [-9861,-3197,507]-[10263,1458.68,4055]
Preparing direct light manager.
Preparing global light manager.
Transferring SD tree to CUDA devices...
SD tree transferred to device 0
Transferring MB tree to CUDA devices...
numNodeFaces=0
numNodes=1
numFaces=0
ZERO LEFT child for node 0
MASK = 1
Successfuly transferred MB tree to CUDA devices...
Transferring geometric data
Transferring vertex normals
669257 vertex normals transferred to device 0
Transferring MOVING vertex normals
0 MOVING vertex normals transferred to device 0
Transferring nodes and meshes
Nodes and meshes transferred to device 0
Transferring MOVING nodes, meshes and perface node indices
Successfuly transferred MOVING nodes and meshes to device 0
Transferring map channels
Map channels transferred
Transferring moving map channels
Moving Map channels transferred
Transferring moving triangles
Moving triangles count = 0
Building dynamic SD tree...
Creating and initializing 8 thread(s).
Allocating memory for build data of 18790 faces (150320 bytes, 0.1 MB).

Initializing face build data.
Creating 'done' event.
Starting first thread.
Waiting for thread completion.
Releasing thread resources.
SDTree statistics:
Total number of faces stored: 18790
Max tree depth: 42
Average tree depth: 18.3484
Number of tree nodes: 13605
Number of tree faces: 63312
Number of tree leafs: 4593
Average faces/leaf: 13.7845
Memory usage: 2.16 MB
Creating and initializing 8 thread(s).
Allocating memory for build data of 10432 faces (83456 bytes, 0.1 MB).
Initializing face build data.
Creating 'done' event.
Starting first thread.
Waiting for thread completion.
Releasing thread resources.
SDTree statistics:
Total number of faces stored: 10432
Max tree depth: 42
Average tree depth: 17.3371
Number of tree nodes: 9483
Number of tree faces: 39445
Number of tree leafs: 3228
Average faces/leaf: 12.2196
Memory usage: 1.98 MB
Creating and initializing 8 thread(s).
Allocating memory for build data of 10528 faces (84224 bytes, 0.1 MB).
Initializing face build data.
Creating 'done' event.
Starting first thread.
Waiting for thread completion.
Releasing thread resources.
SDTree statistics:
Total number of faces stored: 10528
Max tree depth: 43
Average tree depth: 17.2631
Number of tree nodes: 9485
Number of tree faces: 39603
Number of tree leafs: 3231
Average faces/leaf: 12.2572
Memory usage: 1.98 MB
Creating and initializing 8 thread(s).
Allocating memory for build data of 10345 faces (82760 bytes, 0.1 MB).
Initializing face build data.
Creating 'done' event.
Starting first thread.
Waiting for thread completion.
Releasing thread resources.
SDTree statistics:
Total number of faces stored: 10345
Max tree depth: 39
Average tree depth: 16.0925
Number of tree nodes: 8523
Number of tree faces: 34658
Number of tree leafs: 3028
Average faces/leaf: 11.4458
Memory usage: 1.95 MB
Creating and initializing 8 thread(s).
Allocating memory for build data of 8586 faces (68688 bytes, 0.1 MB).
Initializing face build data.
Creating 'done' event.
Starting first thread.
Waiting for thread completion.
Releasing thread resources.
SDTree statistics:
Total number of faces stored: 8586
Max tree depth: 36
Average tree depth: 15.5217
Number of tree nodes: 7241
Number of tree faces: 28006
Number of tree leafs: 2553
Average faces/leaf: 10.9698
Memory usage: 1.89 MB

continues in next post-->

[Jenzie]

Creating and initializing 8 thread(s).
Allocating memory for build data of 10636 faces (85088 bytes, 0.1 MB).
Initializing face build data.
Creating 'done' event.
Starting first thread.
Waiting for thread completion.
Releasing thread resources.
SDTree statistics:
Total number of faces stored: 10636
Max tree depth: 43
Average tree depth: 17.4993
Number of tree nodes: 9377
Number of tree faces: 44853
Number of tree leafs: 3046
Average faces/leaf: 14.7252
Memory usage: 2.02 MB
Creating and initializing 8 thread(s).
Allocating memory for build data of 8620 faces (68960 bytes, 0.1 MB).
Initializing face build data.
Creating 'done' event.
Starting first thread.
Waiting for thread completion.
Releasing thread resources.
SDTree statistics:
Total number of faces stored: 8620
Max tree depth: 33
Average tree depth: 16.662
Number of tree nodes: 9091
Number of tree faces: 31833
Number of tree leafs: 3148
Average faces/leaf: 10.1121
Memory usage: 1.92 MB
Creating and initializing 8 thread(s).
Allocating memory for build data of 16763 faces (134104 bytes, 0.1 MB).

Initializing face build data.
Creating 'done' event.
Starting first thread.
Waiting for thread completion.
Releasing thread resources.
SDTree statistics:
Total number of faces stored: 16763
Max tree depth: 43
Average tree depth: 18.426
Number of tree nodes: 10635
Number of tree faces: 53523
Number of tree leafs: 3831
Average faces/leaf: 13.971
Memory usage: 2.09 MB
Creating and initializing 8 thread(s).
Allocating memory for build data of 20171 faces (161368 bytes, 0.2 MB).

Initializing face build data.
Creating 'done' event.
Starting first thread.
Waiting for thread completion.
Releasing thread resources.
SDTree statistics:
Total number of faces stored: 20171
Max tree depth: 41
Average tree depth: 18.3144
Number of tree nodes: 14085
Number of tree faces: 66263
Number of tree leafs: 4739
Average faces/leaf: 13.9825
Memory usage: 3.56 MB
Creating and initializing 8 thread(s).
Allocating memory for build data of 6426 faces (51408 bytes, 0.0 MB).
Initializing face build data.
Creating 'done' event.
Starting first thread.
Waiting for thread completion.
Releasing thread resources.
SDTree statistics:
Total number of faces stored: 6426
Max tree depth: 34
Average tree depth: 15.8952
Number of tree nodes: 6783
Number of tree faces: 24503
Number of tree leafs: 2243
Average faces/leaf: 10.9242
Memory usage: 1.87 MB
Creating and initializing 8 thread(s).
Allocating memory for build data of 13776 faces (110208 bytes, 0.1 MB).

Initializing face build data.
Creating 'done' event.
Starting first thread.
Waiting for thread completion.
Releasing thread resources.
SDTree statistics:
Total number of faces stored: 13776
Max tree depth: 38
Average tree depth: 16.5471
Number of tree nodes: 11639
Number of tree faces: 44863
Number of tree leafs: 3985
Average faces/leaf: 11.258
Memory usage: 2.02 MB
Creating and initializing 8 thread(s).
Allocating memory for build data of 6150 faces (49200 bytes, 0.0 MB).
Initializing face build data.
Creating 'done' event.
Starting first thread.
Waiting for thread completion.
Releasing thread resources.
SDTree statistics:
Total number of faces stored: 6150
Max tree depth: 28
Average tree depth: 15.2803
Number of tree nodes: 5425
Number of tree faces: 20055
Number of tree leafs: 1898
Average faces/leaf: 10.5664
Memory usage: 1.83 MB
Building tree for dynamic objects...
..done building tree for dynamic objects...
...done building dynamic SD tree
Number of lights: 11
Number of area lights: 9
Number of moving area lights: 0
Area light found at PTLindex = 1
Area light found at PTLindex = 2
Area light found at PTLindex = 3
Area light found at PTLindex = 4
Area light found at PTLindex = 5
Area light found at PTLindex = 6
Area light found at PTLindex = 7
Area light found at PTLindex = 8
Area light found at PTLindex = 9
10 point lights transferred to device 0
Total number of lights added by updateLights(): 10
[RenderView] startCameraTime=0.000000, endCameraTime=1.033333
[RenderView] numCameraTMs=62, numFrames=31, frameSamples=2
[RenderView] startCameraTime=0.000000, endCameraTime=1.033333
[RenderView] numCameraTMs=62, numFrames=31, frameSamples=2
[RenderView] startCameraTime=0.000000, endCameraTime=1.033333
[RenderView] numCameraTMs=62, numFrames=31, frameSamples=2
[RenderView] startCameraTime=0.000000, endCameraTime=1.033333
[RenderView] numCameraTMs=62, numFrames=31, frameSamples=2
[RenderView] startCameraTime=0.000000, endCameraTime=1.033333
[RenderView] numCameraTMs=62, numFrames=31, frameSamples=2
[RenderView] startCameraTime=0.000000, endCameraTime=1.033333
[RenderView] numCameraTMs=62, numFrames=31, frameSamples=2
[RenderView] startCameraTime=0.000000, endCameraTime=1.033333
[RenderView] numCameraTMs=62, numFrames=31, frameSamples=2
[RenderView] startCameraTime=0.000000, endCameraTime=1.033333
[RenderView] numCameraTMs=62, numFrames=31, frameSamples=2
[RenderView] startCameraTime=0.000000, endCameraTime=1.033333
[RenderView] numCameraTMs=62, numFrames=31, frameSamples=2
[RenderView] startCameraTime=0.000000, endCameraTime=1.033333
[RenderView] numCameraTMs=62, numFrames=31, frameSamples=2

[joe_b]

It seems cuda is being initialized and everything is getting sent to the gpu ok. Would you mind sending me this scene? In Rhino if you go to the V-Ray menu you should see a "pack scene" option - that will bundle up your 3dm and all the various assets into a zip file. You could share it with me via dropbox or some other method, I won't disclose it to anyone else. You can PM me through the forum to let me know if this works for you.

[Jenzie]

I really can't share this, I would be hanged. But I can do it with some other scene.

[joe_b]

That's fine, if you have another scene that has the same issue we can take a look and perhaps see what the issue is.