I’m using Maya 2015 in WIndows 7 with a 980ti GTX, but in CPU mode. Vray 2.40.02.
I had quite a bit of difficulty even getting RT to start as it misleadingly throws that odd “Scene is empty” error that now after reading the forum I understand I can safely ignore.
That said, as you can see by the attached image, on the left is when I just do a render. On the right is when RT starts. It renders progressively as one would expect, but its like the camera is transformed all wrong. Almost like its zoomed ortho or something (it isn’t, just a standard perspective from a fresh scene) When I orbit the camera, things change, but I can’t figure out what I’m looking at, its just all wrong. RT is definitely rendering, but what I don’t know.
I attached the scene as a .ma scene (renamed as zip so the upload would work) but really, its just a simple Maya launch, drop a plane, sphere and dome light - then hit render, and that’s what I get. I haven’t done much with this maya install, not even really any preference tweaks.
I use RT at the office quite often, I’m familiar with how it works having used it for years… even at home, but I hadn’t used it in a while and recently updated to the latest stuff and now I’m seeing this.
Any info on the topic would be appreciated. Thanks! bad_rt.zip (79.4 KB)
I can’t look at the file (it says the .zip is corrupted), but if I had to guess it looks like you are rendering an orthographic view. The RT engine currently only supports perspective views.
Thanks for taking the time to make the video, yeah that’s exactly what I’m doing, but when I hit the IPR, I get the result described in my original post. I’m not surprised your results are correct, but I’m unclear what could cause it to break like this, such a radical discrepancy between what’s obviously configured correctly for the regular render, but then its so broken when RT starts. Obviously something on my end, but I’m not sure what provoked the behavior. I have absolutely no idea whats causing this, but if you have any ideas what could be corrupting RT’s camera, please let me know. Is there any way we can debug this to see what RT is doing inside?
For what it’s worth, I just tried in Maya 2014 as well, using vray 2.40.01 and the behavior is the same. I have no 3rd party plugins loaded. I’m really baffled I figured perhaps it was a videocard related thing, but I’m using CPU mode so I doubt that has anything to do with it.
For the record, I have a 980ti with NVidia driver 10.18.13.5330.
Btw, I tried with a new camera but no difference. It responds to my interaction/movement of the perspective camera, so I know it isn’t grabbing one of the ortho cameras for some odd reason.
So the renders here are interesting. The one that’s clearly a plane with two spheres is when rendered with the standard renderer. The one with the greenish bands is with RT, exact same camera. Notice the blue and green hues in the band… that’s the two spheres, but obviously something’s completely wrong with, what I’d guess is the camera transform as RT seems to be happily rendering away, progressively generating a cleaner and cleaner render.
Do you have vray.exe started by Maya or is it registered as service?
Is there a chance that if it is registered as service the version that is started by the V-Ray spawner is different?
It must have been something like that. I had to reboot (although I had already done this many times) and it came back working. I was disabling them as I read it had to be off for the GPU rendering to work. I’ve killed all of the vray spawner services and manually started the 2015 one and it worked. I’m not entirely sure what I did differently.
One last question, I tried to turn on the GPU, I tried Cuda figuring it would work best with my Nvidia cards, but it seems to just crash, I’m not seeing any errors. On the other hand, I tried the openCL config and set it to use my two NVidia cards then selected the openCL in the RT GPU options and it seems to work. Is that recommended over the Cuda at this point?
Using RT-CUDA is faster on nVidia cards.
Are you really sure that RT-OpenCL is really working?
Sometimes when there is a problem with RT GPU it will fall-back to RT-CPU?
I believe so, the log seems to suggest OpenCL is working. As far as drivers go, they’re very recent as I got the card only a couple weeks ago. I installed the latest at the time. NVidia driver 10.18.13.5330 dated from June 17th 2015.
Fortunately, it I cought an error this time when trying CUDA:
"
Building SDTree for GPU
Building SDTree for GPU
warning: Scene is empty
Running RTEngine
Initializing CUDA renderer (single kernel version)
Querying for CUDA devices…
Successfully parsed environment variable VRAY_OPENCL_PLATFORMS_x64
Device GeForce GTX 980 Ti has compute capability 5.2
Device GeForce GTX 770 has compute capability 3.0
./src/opencl_main.cpp(1780) : Driver API error : 209
Bindless textures not supported ./src/ocl_tracedevice.cpp(2143) : Driver API error : 400
./src/ocl_tracedevice.cpp(2146) : Driver API error : 400
Maximum kernel work group size : 128
Maximum kernel work group size : 1024
warning Lightmap failed to load. Filling arrays with default data
Initializing material kernels
./src/skyhit_kernel.cpp(107 : Driver API error :400"
Here’s a portion of the openCL log:
[2015/Aug/6|23:02:59] OpenCL renderer requested.
[2015/Aug/6|23:02:59] 1 plugin(s) loaded successfully
[2015/Aug/6|23:02:59] OpenCL renderer plugin successfully loaded from “C:\Progra
m Files\Autodesk\Maya2015\vray\bin\rt_opencl.dll”
[2015/Aug/6|23:02:59] OpenCL renderer plugin instance successfully created.
[2015/Aug/6|23:02:59] EXT_RTOPENCL interface obtained successfully from OpenCL r
enderer plugin instance.
[2015/Aug/6|23:02:59] ========================
[2015/Aug/6|23:02:59] Starting frame 1.
[2015/Aug/6|23:02:59] warning: Camera FOV is too small, possible round-off error
s.
[2015/Aug/6|23:02:59] Preparing scene for frame…: done [ 0h 0m 0.0s]
[2015/Aug/6|23:02:59] Building SDTree for GPU
[2015/Aug/6|23:02:59] Building SDTree for GPU
[2015/Aug/6|23:02:59] warning: Scene is empty.
[2015/Aug/6|23:02:59] Running RTEngine
[2015/Aug/6|23:02:59] Initializing OpenCL renderer (single kernel version)…
[2015/Aug/6|23:02:59] Querying for OpenCL devices…
[2015/Aug/6|23:02:59] Successfully parsed environment variable VRAY_OPENCL_PLATF
ORMS_x64
[2015/Aug/6|23:02:59] Found 2 OpenCL platforms
[2015/Aug/6|23:02:59] Using the following OpenCL devices:
[2015/Aug/6|23:02:59] AMD Accelerated Parallel Processing Intel(R) Core™ i7 C
PU 920 @ 2.67GHz
[2015/Aug/6|23:02:59] NVIDIA CUDA GeForce GTX 980 Ti
[2015/Aug/6|23:02:59] NVIDIA CUDA GeForce GTX 770
CPU TUPLE SIZE = 6
[2015/Aug/6|23:02:59] Using memory buffers for textures
[2015/Aug/6|23:02:59] Using global ray states
[2015/Aug/6|23:02:59] cl_nv_compiler_options supported!
[2015/Aug/6|23:02:59] Building OpenCL trace program for device NVIDIA CUDA_GeFor
ce GTX 980 Ti…
[2015/Aug/6|23:02:59] OpenCL program built in 0.203 s
[2015/Aug/6|23:02:59] cl_nv_compiler_options supported!
[2015/Aug/6|23:02:59] Building OpenCL trace program for device NVIDIA CUDA_GeFor
ce GTX 770…
[2015/Aug/6|23:03:00] OpenCL program built in 0.203 s
[2015/Aug/6|23:03:00] cl_nv_compiler_options extension not found.
[2015/Aug/6|23:03:00] Building OpenCL trace program for device AMD Accelerated P
arallel Processing_Intel(R) Core™ i7 CPU 920 @ 2.67GHz…
[2015/Aug/6|23:03:00] OpenCL program built in 0.203 s
[2015/Aug/6|23:03:00] Kernel work group size: 256
[2015/Aug/6|23:03:00] Maximum device work group size: 1024
[2015/Aug/6|23:03:00] Maximum device memory allocation size (MB): 1536
[2015/Aug/6|23:03:00] Kernel work group size: 1024
[2015/Aug/6|23:03:00] Maximum device work group size: 1024
[2015/Aug/6|23:03:00] Maximum device memory allocation size (MB): 512
[2015/Aug/6|23:03:00] Kernel work group size: 1024
[2015/Aug/6|23:03:00] Maximum device work group size: 1024
[2015/Aug/6|23:03:00] Maximum device memory allocation size (MB): 3069
[2015/Aug/6|23:03:00] warning: Lightmap failed to load. Filling arrays with defa
ult data
[2015/Aug/6|23:03:00] Initializing material kernels
[2015/Aug/6|23:03:00] Number of OpenCL devices: 3
[2015/Aug/6|23:03:00] Transferring SD tree to OpenCL devices…
[2015/Aug/6|23:03:00] SD tree transferred to device 0
[2015/Aug/6|23:03:00] SD tree transferred to device 1
[2015/Aug/6|23:03:00] SD tree transferred to device 2
[2015/Aug/6|23:03:00] OpenCL renderer requested.
[2015/Aug/6|23:03:00] OpenCL renderer plugin already loaded.
[2015/Aug/6|23:03:00] OpenCL renderer plugin instance successfully created.
[2015/Aug/6|23:03:00] EXT_RTOPENCL interface obtained successfully from OpenCL r
enderer plugin instance.
[2015/Aug/6|23:03:00] Preparing scene for frame…: done [ 0h 0m 0.0s]
[2015/Aug/6|23:03:00] Building SDTree for GPU
[2015/Aug/6|23:03:00] Preparing ray server…: done [ 0h 0m 0.0s]
[2015/Aug/6|23:03:00] SDTree statistics:
[2015/Aug/6|23:03:00] Total number of faces stored: 788
[2015/Aug/6|23:03:00] Max tree depth: 25
[2015/Aug/6|23:03:00] Average tree depth: 12.322
[2015/Aug/6|23:03:00] Number of tree nodes: 733
[2015/Aug/6|23:03:00] Number of tree faces: 2467
[2015/Aug/6|23:03:00] Number of tree leafs: 264
[2015/Aug/6|23:03:00] Average faces/leaf: 9.3447
[2015/Aug/6|23:03:00] Memory usage: 1.70 MB
[2015/Aug/6|23:03:00] Building SDTree for GPU
[2015/Aug/6|23:03:00] Transferring SD tree to OpenCL devices…
[2015/Aug/6|23:03:00] SD tree transferred to device 0
[2015/Aug/6|23:03:00] SD tree transferred to device 1
[2015/Aug/6|23:03:00] SD tree transferred to device 2
[2015/Aug/6|23:03:00] Transferring MB tree to OpenCL devices…
[2015/Aug/6|23:03:00] numNodeFaces=0
[2015/Aug/6|23:03:00] numNodes=1
[2015/Aug/6|23:03:00] numFaces=0
[2015/Aug/6|23:03:00] ZERO LEFT child for node 0
[2015/Aug/6|23:03:00] MASK = 1
[2015/Aug/6|23:03:00] Successfuly transferred MB tree to OpenCL devices…
[2015/Aug/6|23:03:00] Transferring geometric data
[2015/Aug/6|23:03:00] Transferring vertex normals
[2015/Aug/6|23:03:00] 435 vertex normals transferred to device 0
[2015/Aug/6|23:03:00] Transferring MOVING vertex normals
[2015/Aug/6|23:03:00] 0 MOVING vertex normals transferred to device 0
[2015/Aug/6|23:03:00] Transferring nodes and meshes
[2015/Aug/6|23:03:00] Nodes and meshes transferred to device 0
[2015/Aug/6|23:03:00] Transferring MOVING nodes, meshes and perface node indices
[2015/Aug/6|23:03:00] Successfuly transferred MOVING nodes and meshes to device
0
[2015/Aug/6|23:03:00] Transferring vertex normals
[2015/Aug/6|23:03:00] 435 vertex normals transferred to device 1
[2015/Aug/6|23:03:00] Transferring MOVING vertex normals
[2015/Aug/6|23:03:00] 0 MOVING vertex normals transferred to device 1
[2015/Aug/6|23:03:00] Transferring nodes and meshes
[2015/Aug/6|23:03:00] Nodes and meshes transferred to device 1
[2015/Aug/6|23:03:00] Transferring MOVING nodes, meshes and perface node indices
[2015/Aug/6|23:03:00] Successfuly transferred MOVING nodes and meshes to device
1
[2015/Aug/6|23:03:00] Transferring vertex normals
[2015/Aug/6|23:03:00] 435 vertex normals transferred to device 2
[2015/Aug/6|23:03:00] Transferring MOVING vertex normals
[2015/Aug/6|23:03:00] 0 MOVING vertex normals transferred to device 2
[2015/Aug/6|23:03:00] Transferring nodes and meshes
[2015/Aug/6|23:03:00] Nodes and meshes transferred to device 2
[2015/Aug/6|23:03:00] Transferring MOVING nodes, meshes and perface node indices
[2015/Aug/6|23:03:00] Successfuly transferred MOVING nodes and meshes to device
2
[2015/Aug/6|23:03:00] Transferring map channels
[2015/Aug/6|23:03:00] Map channels transferred
[2015/Aug/6|23:03:00] Transferring moving map channels
[2015/Aug/6|23:03:00] Moving Map channels transferred
[2015/Aug/6|23:03:00] Transferring moving triangles
[2015/Aug/6|23:03:00] Moving triangles count = 0
[2015/Aug/6|23:03:00] Building dynamic SD tree…
[2015/Aug/6|23:03:00] Building tree for dynamic objects…
[2015/Aug/6|23:03:00] ..done building tree for dynamic objects…
[2015/Aug/6|23:03:00] …done building dynamic SD tree
[2015/Aug/6|23:03:00] Loading bitmap “fractal1”
[2015/Aug/6|23:03:00] Loaded bitmap “fractal1”
[2015/Aug/6|23:03:00] Number of lights: 2
[2015/Aug/6|23:03:00] Number of area lights: 1
[2015/Aug/6|23:03:00] Number of moving area lights: 0
[2015/Aug/6|23:03:00] Area light found at PTLindex = 1
[2015/Aug/6|23:03:00] 2 point lights transferred to device 0
[2015/Aug/6|23:03:00] 2 point lights transferred to device 1
[2015/Aug/6|23:03:00] 2 point lights transferred to device 2
[2015/Aug/6|23:03:00] Total number of lights added by updateLights(): 2
"
One of our GPU devs told me that 2.40 doesn’t support Maxwell cards.
Can you please send an e-mail to support@chaosgroup.com and they’ll give you a build that should work.
Thank you, yes I tried the ocl device select and set it to one of the cards at a time and it seems to work. Having both selected seems to be a problem. Should this work?
Multi GPU setups are supported. However, Maxwell architecture is not supported in 2.4, and 980ti is Maxwell GPU. The good thing is that this is fixed in the stable nightly builds and everything you have to do to get them is to write mail to support@chaosgroup.com with link to this topic. We try to keep information regarding those here Chaos Docs.
