Moderator: juanjgon
BorisGoreta wrote:The scene is working properly on all GPUs, TITAN, TITAN Black, TITAN X, GTX 780 and GTX 980 Ti, locally and over network. Memory usage is 4539 MB. 30720000 triangles.
Weird, can you export your 2.x scene render target to a file and send it to me? Perhaps there is a settings corrupted or not translated correctly.Mastoy wrote:It seems there is a problem with the render passes node, when the scene has been created with octane v2.5.
When I try to render a scene I've created with an old version, the performance is awful when rendering passes (10 min where it should be 10 sec).
If I delete the render passes node and recreate it (with Octane V3), problem seems to be gone. Could someone confirm ?
Donejuanjgon wrote:Weird, can you export your 2.x scene render target to a file and send it to me? Perhaps there is a settings corrupted or not translated correctly.Mastoy wrote:It seems there is a problem with the render passes node, when the scene has been created with octane v2.5.
When I try to render a scene I've created with an old version, the performance is awful when rendering passes (10 min where it should be 10 sec).
If I delete the render passes node and recreate it (with Octane V3), problem seems to be gone. Could someone confirm ?
-Juanjo
Sorry, I can't reproduce this problem here only with your render target data. Do you know if you have this problem in all your scenes? What render phase is slow, the info passes, the beauty ...?Mastoy wrote:It seems there is a problem with the render passes node, when the scene has been created with octane v2.5.
When I try to render a scene I've created with an old version, the performance is awful when rendering passes (10 min where it should be 10 sec).
If I delete the render passes node and recreate it (with Octane V3), problem seems to be gone. Could someone confirm ?
----------------------------------------------------------------------juanjgon wrote:BorisGoreta wrote:The scene is working properly on all GPUs, TITAN, TITAN Black, TITAN X, GTX 780 and GTX 980 Ti, locally and over network. Memory usage is 4539 MB. 30720000 triangles.
Thanks Boris. I will try also to reproduce this problem next week in the office, but perhaps the jang2235 Titan X GPU could have a hardware or drivers problem.
jang2235, do you have your drivers updated to the last ones available?
-Juanjo
---------------------------------------------------------------BorisGoreta wrote:This scene does not compute on TITAN X:
00:00:34 (0034.66) | Scene loaded
00:00:34 (0034.66) | Render resolution: 1920 x 1080
00:00:34 (0034.66) |
00:00:34 (0034.66) | IPR: scene loaded
00:00:34 (0034.66) | IPR: CalculateAllNormals 0
00:00:34 (0034.82) | IPR: Enable IPR events processing
00:00:34 (0034.82) | IPR: init rendering
00:00:34 (0034.82) | IPR: set IPR camera options
00:00:34 (0034.82) | IPR: set image callback
00:00:34 (0034.82) |
00:00:34 (0034.82) | Render scene, set renderTarget and update
00:00:34 (0034.82) | ... first update
00:00:34 (0034.82) | ... setRenderPriority
00:00:34 (0034.82) | ... RenderPriority set to 0
00:00:34 (0034.82) | ... set Motion Blur parameters
00:00:34 (0034.82) | ... setRenderTargetNode
00:02:44 (0164.48) | ... second update
00:02:44 (0164.48) | ... isRenderingPaused
00:02:44 (0164.48) | Render setup Ok.
00:02:44 (0164.48) |
00:02:44 (0164.48) | [profile] Function "LWOctane_renderScene" over "" execution time: 129.661 seconds
00:02:44 (0164.48) | IPR: enter into the rendering loop
00:02:47 (0167.21) * OCTANE API MSG: CUDA error 719 on device 5: An exception occurred on the device while executing a kernel. The context cannot be used anymore and must be destroyed. All existing device memory allocations from this context are invalid and must be reconstructed.
00:02:47 (0167.21) * OCTANE API MSG: -> failed to launch kernel (ptLight1)
00:02:47 (0167.21) * OCTANE API MSG: device 5: direct light kernel failed
00:02:47 (0167.21) |
00:02:47 (0167.21) | ++++++++++++++++++++++++++
00:02:47 (0167.21) | +++ IPR RENDER FAILURE +++ processing the failure callback
00:02:47 (0167.21) | ++++++++++++++++++++++++++
00:02:47 (0167.21) |
00:02:47 (0167.22) | IPR: reset image callback
00:02:47 (0167.22) | IPR: wait for the getPreviewImage function
00:02:47 (0167.32) | IPR: free the OpenGL buffers
00:02:47 (0167.32) | IPR: rendering done
To fix this problem I've added a new parameter to the Octane custom object plugin to disable the NGons tessellation functions. With this option enabled you should get the same behavior with the OpenSubD surfaces that you had before this release.FrankPooleFloating wrote: NGons support is really nice, and this will definitely speed up workflow.. however, just for the hell of it, I made a cylinder and turned on Octane OpenSubDiv, and the results were strange. Not that I would normally subd an ngons mesh, but it appears to subdivide the bottom half of the cylinder and not the top.. try this. Weird.
