today I enabled NVlink of my two 2080ti for testing NVlink and VfR. But it was not working like expected. So disabled SLI again, but now I can render one image only and if I start the next one, than I get a crash after a short while. Restarting Rhino helps for one rendering.
I removed the NVlink from the cards, updated the driver to the current recommended driver … but nothing helps. This can be found at the progress window:
D:\cgrepo\vraysdk\samples\vray_plugins\rt_private\rt_opencl_pri(1715) : CUDA error 700
D:\cgrepo\vraysdk\samples\vray_plugins\rt_private\rt_opencl_pri(2059) : CUDA error 700
Sidenote: at my computer D:\cgrepo.. doesn’t exist.
Preparing renderer...
Creating bitmap manager
Bitmap manager created.
Scanning scene for light plugins.
9 lights found.
Using full light evaluation
Light tree is enabled with clamp coefficient 0.1 and even split coefficient 0.5.
CUDA renderer requested.
CUDA renderer plugin already loaded.
CUDA renderer plugin instance successfully created.
EXT_RTOPENCL interface obtained successfully from CUDA renderer plugin instance.
Using low thread priority.
Max ray intensity is enabled: rendered result may have incorrect brightness.
Estimated progressive buffer memory usage of 2644 MB is over the limit of 2048 MB. Falling back to per-pixel filtering.
Dome light resolution increased from 512 to 2048 for light "/Dome Light"
Bitmap file "C:\Program Files\Chaos Group\V-Ray\V-Ray for Rhinoceros\RhinoDefault.exr" loaded.
Dome light resolution increased from 512 to 2048 for light "/Dome Light#1"
Using Embree ray tracing.
SDTree statistics:
Total number of faces stored: 747523
Max tree depth: 42
Average tree depth: 24.9892
Number of tree nodes: 333801
Number of tree faces: 2202549
Number of tree leaves: 143607
Average faces/leaf: 15.3373
Memory usage: 92.27 MB
Building static SD trees took 176 milliseconds
Running RTEngine
Initializing CUDA renderer...
Initializing environment kernel
Number of CUDA devices: 2
nVidia driver version: 441.28
Min shading rate set to : 6
Building trees for 6 meshes with all threads per mesh took 0ms.
BVH statistics:
Primitives count: 6 (Mem usage:0.00 MB)
Meshes count: 6 (Mem usage:0.00 MB)
Tree depth: 4
Number of tree nodes: 11
Tree takes 0.58 MB. Build time 0.21 ms.
GPU tex loaded RhinoDefault.exr (idx=0) mip 0, w 1024/1024, h 512/512, channels 3, 3072KB
Texture "/Boden/Base/V-Ray BRDF/Bulge" is not supported and will be baked.
Bitmap file "D:\Projekte\...\tex\CHERRY270x240 03.png" loaded.
Bitmap file "D:\Texturen\Haut Leder\Leder-braungrau-02.png" loaded.
GPU tex loaded Leder-braungrau-02.png (idx=3) mip 0, w 2000/2000, h 2000/2000, channels 3, 11718KB
GPU tex loaded wood-095_mahogany-american-1-D2s.png (idx=2) mip 0, w 2000/2000, h 2000/2000, channels 3, 11718KB
GPU tex loaded Leder-ws.png (idx=4) mip 0, w 2000/2000, h 2000/2000, channels 1, 3906KB
GPU tex loaded CHERRY270x240 03.png (idx=1) mip 0, w 3927/3927, h 3490/3490, channels 3, 40152KB
Scene contains 8 lights
Enabled Light Tree on GPU
Building light cache.
Tracing 1000000 image samples for light cache in 64 passes.
Using 99.26 MB for light cache preview buffer
Merging light cache passes.
Light cache contains 3784 samples.
Light cache base memory usage is 0.6 MB.
Using premultiplied light cache.
Lightmap successfully loaded.
GPU light cache: 5405 samples stored.
Mtls 76.98MB, Geo 70.21MB, Lights 160.00MB, LightCache 0.08MB, Reserved 3.11MB (data index 18327)
Building optimized CUDA kernel (0) for comp cap 7.5 for device 0...
PTX file generated with CUDA Toolkit v10.0 for CUDA compute capability 3.0. CUDA runtime version 7050.
Optimized CUDA kernel assembled successfully
Total memory for device 0 : 11264 MB, free 8199 MB
Building optimized CUDA kernel (0) for comp cap 7.5 for device 1...
PTX file generated with CUDA Toolkit v10.0 for CUDA compute capability 3.0. CUDA runtime version 7050.
Optimized CUDA kernel assembled successfully
Total memory for device 1 : 11264 MB, free 7505 MB
Progressive sampling base buffer memory usage: 434.14 MB
D:\cgrepo\vraysdk\samples\vray_plugins\rt_private\rt_opencl_pri(1715) : CUDA error 700
D:\cgrepo\vraysdk\samples\vray_plugins\rt_private\rt_opencl_pri(2059) : CUDA error 700
Closing RTEngine (cancelled)
Max. Mpaths per second: 0.000000
Sampling level reached: 64 samples per pixel
V-Ray GPU kernel render time: 36.76s
Total GPU samples 526363648
Mtls 76.98MB, Geo 70.21MB, Lights 160.00MB, LightCache 0.08MB, Reserved 2814.80MB
Device 0 statistics:
GPU time: 90.591438 %
GPU ray trace time 96.538289 %
Transfer to GPU time 0.246350 %
Transfer from GPU time 3.215361 %
Device 1 statistics:
GPU time: 90.618499 %
GPU ray trace time 96.511315 %
Transfer to GPU time 0.220322 %
Transfer from GPU time 3.268362 %
Number of raycasts: 7408205 (2.29 per pixel)
Camera rays: 1000365 (0.31 per pixel)
Shadow rays: 5080670 (1.57 per pixel)
GI rays: 1873125 (0.58 per pixel)
Reflection rays: 431175 (0.13 per pixel)
Refraction rays: 55009 (0.02 per pixel)
Unshaded rays: 0 (0.00 per pixel)
Number of light evaluations: 10603728 (3.27 per pixel)
Maximum memory usage for resman: 0.00 MB
Number of intersectable primitives: 747525
SD triangles: 747523
MB triangles: 0
Static primitives: 0
Moving primitives: 0
Infinite primitives: 2
Static hair segments: 0
Moving hair segments: 0
Cleaning up bitmap manager
Maximum memory usage for texman: 0.00 MB
Thank you for reporting this!
Our Dev team has already made an improvement preventing this error from appearing.
This will be most likely implemented in the next update for V-Ray for Rhino.