Page MenuHome

Cycles stopping with error: CUDA error: Launch exceeded timeout in cuCtxSynchronize()
Closed, ArchivedPublic

Description

System Information

Mac Pro (Early 2009)
2 x 2.26 GHz Quad-Core Intel Xeon
20 GB 1066 MHz DDR3 ECC

Model Name:	Mac Pro
Model Identifier:	MacPro4,1
Processor Name:	Quad-Core Intel Xeon
Processor Speed:	2.26 GHz
Number of Processors:	2
Total Number of Cores:	8
L2 Cache (per Core):	256 KB
L3 Cache (per Processor):	8 MB
Memory:	20 GB
Processor Interconnect Speed:	5.86 GT/s
Boot ROM Version:	MP41.0081.B07
SMC Version (system):	1.39f5

2 Video Cards: NVIDIA GeForce GTX 580 3071 MB x 2

Operating system and graphics card

Blender Version
Broken: Version 2.77 (2.77 2016-03-18, Blender Foundation)
Worked:

Short description of error

Part way through rendering (Variously 10 minutes to 30 minutes) Rendering stops with error:
CUDA error: Launch exceeded timeout in cuCtxSynchronize()

(bpy.context.space_data.viewport_shade = 'WIREFRAME'
bpy.context.space_data.viewport_shade = 'SOLID'
bpy.ops.view3d.object_as_camera()
bpy.context.space_data.context = 'RENDER'
bpy.context.scene.render.resolution_x = 800
bpy.context.scene.render.resolution_y = 600
bpy.context.scene.render.resolution_percentage = 50
bpy.context.scene.render.pixel_aspect_x = 1
bpy.context.scene.cycles.device = 'CPU'
bpy.context.scene.cycles.aa_samples = 3
bpy.context.scene.cycles.preview_aa_samples = 3
bpy.context.scene.cycles.aa_samples = 2
bpy.context.scene.cycles.preview_aa_samples = 2
bpy.context.scene.layers[12] = False
bpy.context.scene.layers[13] = False
bpy.context.scene.layers[14] = False
bpy.context.scene.layers[15] = False
bpy.context.scene.layers[16] = False
bpy.context.scene.layers[16] = True
bpy.context.scene.layers[17] = False
bpy.context.scene.layers[9] = False
bpy.context.scene.layers[8] = False
bpy.context.scene.layers[8] = True
bpy.context.scene.layers[7] = False
bpy.context.scene.layers[7] = True
bpy.context.scene.layers[6] = False
bpy.context.scene.layers[5] = False
bpy.context.scene.layers[4] = False
bpy.context.scene.layers[3] = False
bpy.context.scene.layers[2] = False
bpy.context.scene.layers[2] = True
bpy.context.scene.layers[1] = False
bpy.context.scene.layers[0] = False
bpy.context.scene.cycles.device = 'GPU'
bpy.context.scene.layers[12] = True
bpy.context.scene.layers[13] = True
bpy.context.scene.layers[13] = False
bpy.context.scene.layers[12] = False
bpy.ops.transform.translate(value=(0, -9.68792, 0), constraint_axis=(False, True, False), constraint_orientation='GLOBAL', mirror=False, proportional='DISABLED', proportional_edit_falloff='SMOOTH', proportional_size=0.122845, release_confirm=True)
bpy.ops.object.select_all(action='TOGGLE')
bpy.context.scene.layers[12] = True
bpy.context.scene.layers[12] = False
bpy.context.scene.layers[13] = True
bpy.context.scene.layers[13] = False
bpy.context.scene.layers[10] = True
bpy.context.scene.layers[10] = False
bpy.context.scene.layers[10] = True
bpy.context.scene.layers[10] = False
bpy.context.scene.layers[11] = True
bpy.context.scene.layers[11] = False
bpy.context.scene.layers[11] = True
bpy.context.scene.layers[10] = True
bpy.context.scene.cycles.max_bounces = 9
bpy.context.scene.cycles.max_bounces = 8
bpy.context.scene.cycles.max_bounces = 7
bpy.context.scene.cycles.max_bounces = 6
bpy.context.scene.cycles.max_bounces = 5
bpy.context.scene.cycles.max_bounces = 4
bpy.context.scene.cycles.max_bounces = 5
bpy.context.scene.cycles.aa_samples = 3
bpy.context.space_data.context = 'DATA'
bpy.context.object.data.type = 'AREA'
bpy.ops.transform.resize(value=(2.29422, 2.29422, 2.29422), constraint_axis=(False, False, False), constraint_orientation='GLOBAL', mirror=False, proportional='DISABLED', proportional_edit_falloff='SMOOTH', proportional_size=0.122845)
bpy.context.space_data.viewport_shade = 'RENDERED'
bpy.context.area.type = 'VIEW_3D'
bpy.context.object.data.type = 'SPOT'
bpy.context.object.data.type = 'AREA'
bpy.data.node_groups["Shader Nodetree"].nodes["Emission"].inputs[1].default_value = 100
bpy.data.node_groups["Shader Nodetree"].nodes["Emission"].inputs[1].default_value = 200
bpy.data.node_groups["Shader Nodetree"].nodes["Emission"].inputs[0].default_value = (0.798751, 1, 0.38944, 1)
bpy.data.node_groups["Shader Nodetree"].nodes["Emission"].inputs[0].default_value = (1, 0.83349, 0.603827, 1)
bpy.data.node_groups["Shader Nodetree"].nodes["Emission"].inputs[0].default_value = (1, 0.881094, 0.772659, 1)
bpy.data.node_groups["Shader Nodetree"].nodes["Emission"].inputs[0].default_value = (0.76304, 0.749842, 1, 1)
bpy.data.node_groups["Shader Nodetree"].nodes["Emission"].inputs[0].default_value = (0.76304, 0.749842, 1, 1)
bpy.data.node_groups["Shader Nodetree"].nodes["Emission"].inputs[0].default_value = (0.499976, 0.576211, 1, 1)
bpy.data.node_groups["Shader Nodetree"].nodes["Emission"].inputs[0].default_value = (0.499976, 0.576211, 1, 1)
bpy.data.node_groups["Shader Nodetree"].nodes["Emission"].inputs[1].default_value = 150
bpy.context.space_data.viewport_shade = 'SOLID'
bpy.context.space_data.context = 'RENDER'
bpy.context.scene.cycles.aa_samples = 4
bpy.context.scene.render.resolution_percentage = 100
bpy.context.scene.render.resolution_x = 1600
bpy.context.scene.render.resolution_y = 1200
bpy.context.scene.render.resolution_percentage = 100
bpy.context.area.type = 'VIEW_3D'
CUDA error: Launch exceeded timeout in cuCtxSynchronize()
CUDA error: Launch exceeded timeout in cuCtxSynchronize()
bpy.context.area.type = 'CONSOLE'
bpy.context.area.type = 'INFO'
)

Exact steps for others to reproduce the error
just use the .blend file attached with the settings as saved using all populated layers and render with GPU.

Event Timeline

below messages were output in a console window.
Is the cause not this?

Cycles shader graph connect: can only connect closure to closure (bsdf.BSDF to convert.Color).
Cycles shader graph connect: can only connect closure to closure (bsdf.BSDF to convert.Color).
Cycles shader graph connect: can only connect closure to closure (bsdf.BSDF to mix_closure.Fac).
Cycles shader graph connect: can only connect closure to closure (bsdf.BSDF to convert.Color).
Cycles shader graph connect: can only connect closure to closure (bsdf.BSDF to convert.Color).
Cycles shader graph connect: can only connect closure to closure (mix_closure.Closure to bsdf.Color).

I don't think that would cause this error. This is I believe because of a tdr time out. However, I do not know how to fix this on OSX

Probably they may not inspect it with the attached blend file.
Data are missing.
And a file is too big.
I think that you should show a concise sample file to inspect a problem.

Sergey Sharybin (sergey) changed the task status from Unknown Status to Unknown Status.Mar 22 2016, 1:01 PM
Sergey Sharybin (sergey) claimed this task.

This is an intrinsic limitation of CUDA which puts a timelimit on the kernels when using display device for compute. Workaround could be to use smaller tiles, but with increased scene complexity the issue will happen again, Real fix would be to use dedicated card for compute.

Thanks for the report, but it's how CUDA works and nothing we can solve from Blender side.

I just got this same timeout error in OSX using a dedicated Titan X GPU (no monitors or devices connected to the card). Im using 2.77a and OSX 10.10.4

@Mike (LMProductions), please attach system-info.txt generated by Help -> Save System Info.