hi brecht
is there are now features limitation in optix in comparison with cuda? i mean by similarity with this pic…
this picture is EXAMPLE - which features i mean…
this picture is EXAMPLE - which features i mean…
I believe some Cycles nodes are not supported yet for OptiX: the nodes that are ray-based, like the Ambient Occlusion and Bevel node.
See the release notes:
https://wiki.blender.org/wiki/Reference/Release_Notes/2.81/Cycles#NVIDIA_RTX
2 Likes
Currently still missing are baking, branched path tracing, ambient occlusion and bevel nodes, using CPU memory, and combined CPU + GPU rendering.
all this technically (in future) possible realize in Optix render mode?
Yes, but we don’t know exactly when it will happen.
2 Likes
Hi all,
Just built the latest Cycles branch as of 18/10/2019 and having issues:
I1018 01:57:19.319701 33078 device_optix.cpp:219] All knobs on default.
I1018 01:57:19.333133 33078 device_optix.cpp:219] Opened database: “/var/tmp/OptixCache_alex/cache7.db”
I1018 01:57:19.333184 33078 device_optix.cpp:219] Cache data size: “101.2 MiB”
I1018 01:57:19.520637 33385 device_optix.cpp:219] Cache miss for key: ptx-7258562-key9c6857aa7c271602f2f629cab91aaf24-sm_75-rtc1-drv435.21
I1018 01:58:59.775705 33385 device_optix.cpp:219] Inserted module in cache with key: ptx-7258562-key9c6857aa7c271602f2f629cab91aaf24-sm_75-rtc1-drv435.21
I1018 01:58:59.900295 33385 device_optix.cpp:219] Info: Pipeline parameter “__params” size is 1888 bytes
Info: Module uses 6 payload values. Pipeline configuration: 6.
Info: Module uses 2 attribute values. Pipeline configuration: 2 (default).
Info: Entry function “__raygen__kernel_optix_path_trace” with semantic type RAYGEN has 9 trace call(s), 0 continuation callable call(s), 0 direct callable call(s), 12181 basic block(s), 126775 instruction(s)
Info: Entry function “__raygen__kernel_optix_displace” with semantic type RAYGEN has 0 trace call(s), 0 continuation callable call(s), 0 direct callable call(s), 34 basic block(s), 1059 instruction(s)
Info: Entry function “__raygen__kernel_optix_background” with semantic type RAYGEN has 0 trace call(s), 0 continuation callable call(s), 0 direct callable call(s), 8 basic block(s), 178 instruction(s)
Info: Entry function “__miss__kernel_optix_miss” with semantic type MISS has 0 trace call(s), 0 continuation callable call(s), 0 direct callable call(s), 2 basic block(s), 6 instruction(s)
Info: Entry function “__anyhit__kernel_optix_local_hit” with semantic type ANYHIT has 0 trace call(s), 0 continuation callable call(s), 0 direct callable call(s), 24 basic block(s), 216 instruction(s)
Info: Entry function “__anyhit__kernel_optix_shadow_all_hit” with semantic type ANYHIT has 0 trace call(s), 0 continuation callable call(s), 0 direct callable call(s), 16 basic block(s), 136 instruction(s)
Info: Entry function “__anyhit__kernel_optix_visibility_test” with semantic type ANYHIT has 0 trace call(s), 0 continuation callable call(s), 0 direct callable call(s), 6 basic block(s), 22 instruction(s)
Info: Entry function “__closesthit__kernel_optix_hit” with semantic type CLOSESTHIT has 0 trace call(s), 0 continuation callable call(s), 0 direct callable call(s), 7 basic block(s), 49 instruction(s)
Info: Entry function “__intersection__curve” with semantic type INTERSECTION has 0 trace call(s), 0 continuation callable call(s), 0 direct callable call(s), 126 basic block(s), 1676 instruction(s)
Info: 15 non-entry function(s) have 6004 basic block(s), 97266 instruction(s)
E1018 01:59:00.128399 33385 device_optix.cpp:213] COMPILE ERROR: “__exception__kernel_optix_exception” not found in programDescriptions[5].exception.module
OptiX error OPTIX_ERROR_INVALID_VALUE in optixProgramGroupCreate( context, group_descs, NUM_PROGRAM_GROUPS, &group_options, nullptr, 0, groups), line 457
I am building against Gaffer though, so I am not too sure if it is me who is doing something wrong here.
I’m using Cuda 10.1, OptiX 7.0.0, RTX 2070, Linux/Kubuntu 18.04, Nvidia driver 435.21.
Cheers
All good this is just debug mode which had the issue, release build worked like a charm and my pixels converge noticeably faster now, amazing!
hi Brecht
this is just my rumbling - no need to answer to this - i just want that - this read those who can do something about this… but - Brecht how it is even exist - e-cycles? How other version of cycles can be faster than yours? As if some one know your render better than you. For me this sound like some where be Vray faster than Vray which chaosgroup sell on their page, and wherein Vlado know that some where are Vray faster than his. I hope you “fix this issue” soon…
1 Like
Isn’t that exactly in the spirit of open source? Anyone can pick it up and improve upon it.
As far as I know, many of the improvements are going to be submitted by Mathieu for review around the end of the year. To me, this looks like a win-win situation. He develops improvements and initially provides the to his customers and after some time, the whole Blender community profits.
The way you are describing it sounds more like an ego thing, while in my opinion it should be more community thing.
6 Likes