Replies: 13 comments 49 replies
-
|
Hi @samhodge , What issues do you see with running the example scenes? In addition to running ALab with USD via the hdMoonray hydra delegate, we've also put up a MoonRay-native RDL2 conversion of ALab for testing. It can be found on our Test Scenes page here: |
Beta Was this translation helpful? Give feedback.
-
@xXTR0YXx How did you fix this issue?? I'm facing exact the same error right now. |
Beta Was this translation helpful? Give feedback.
-
|
changed directory inside ALab solved my problem.. This works |
Beta Was this translation helpful? Give feedback.
-
|
Renders okay for me now after upping the ulimit to 8192. Is there a way to save a denoised snapshot out of moonray_gui? Pressing K just saves the RAW render to disc |
Beta Was this translation helpful? Give feedback.
-
|
@xXTR0YXx I found |
Beta Was this translation helpful? Give feedback.
-
I added |
Beta Was this translation helpful? Give feedback.
-
|
Works!
|
Beta Was this translation helpful? Give feedback.
-
|
This is my script for rendering out the sequence on a single box |
Beta Was this translation helpful? Give feedback.
-
|
@samhodge Thank you for sharing this!! I'm trying your script now and I got following warnings. Did you also see those warning messages? Is this ignorable for this rendering?? |
Beta Was this translation helpful? Give feedback.
-
|
My above question solved: #64 (reply in thread) |
Beta Was this translation helpful? Give feedback.
-
|
sorry for the delay, this is expected as you now know. |
Beta Was this translation helpful? Give feedback.
-
|
Thank you for your help! I wrote blog post about MoonRay on Rocky Linux 9, please have a look! |
Beta Was this translation helpful? Give feedback.
-
|
The stoat should have white fur and a red pullover knitted top. |
Beta Was this translation helpful? Give feedback.
-
|
I've got a furry stoat with a red sweater! CPU: 16-core Intel(R) Xeon(R) Gold 5122 CPU @ 3.60GHz Thanks, DW for releasing this great tool! |
Beta Was this translation helpful? Give feedback.
-
|
Tried this and rendered from Arnold and found the following missing textures after running sed to replace .exr with .tx and also hitting the .exr files with txmake and adjusting the extension from .exr to .tx Because Arnold doesn't recognise a .exr as a valid mipmapped file extension and post converts the +mipmap .exr files anyway |
Beta Was this translation helpful? Give feedback.
-
|
Looks like those some of textures are from the high resolution textures only. |
Beta Was this translation helpful? Give feedback.
-
|
Looks like I am going to the high resolution textures now to solve the problem |
Beta Was this translation helpful? Give feedback.
-
|
So close kick-compare_AL_native006.1007.zip Just need to grab these textures from the procedural folder: And convert them to |
Beta Was this translation helpful? Give feedback.
-
|
we got there in the end |
Beta Was this translation helpful? Give feedback.
-
|
Can you tell me what the following were set to with this command hd_usd2rdl -in ./ALab/entry.usda -purpose render -time ${i} -camera root/camera01/GEO/renderCam_hrc/renderCam_buffer/renderCam_srt/renderCam -out stoatrender.${i}.rdlb How many camera anti aliasing sample? What is the total ray depth? What is the transmission ray depth? What is the specular/glossy ray depth? What is the diffuse ray depth? How many diffuse samples? How many specular samples? How many transmission samples? How many light samples? Do any of these questions have no meaning to Moonray? |
Beta Was this translation helpful? Give feedback.
-
|
All of our scene settings can be found in our SceneVariables attribute reference, and a more detailed explanation on our depth settings can be found here. You can set these attributes in two different ways:
Hope that helps! |
Beta Was this translation helpful? Give feedback.
-
|
Also, if you want to see what the settings are with your current command you can use the It tells you what the SceneVariables are set to (along with other logging info). Likely they're just the defaults since you didn't explicitly set them. Another option would be to use rdl2_print: |
Beta Was this translation helpful? Give feedback.
-
|
I guess my question is what are the default values for the samples per closure and the ray depth? I am over the moon with the information I have so far, I can find the answers out for myself and report back |
Beta Was this translation helpful? Give feedback.
-
|
https://docs.openmoonray.org/user-reference/scene-objects/scene-variables/SceneVariables/ bsdf_sampler_strategy Permalink to bsdf_sampler_strategy bsdf_samples Permalink to bsdf_samples No documentation available bssrdf_samples Permalink to bssrdf_samples No documentation available disable_optimized_hair_sampling Permalink to disable_optimized_hair_sampling Forces all hair materials to sample each hair BSDF lobe independently. This will enable the LPE label syntax for 'hair R', 'hair TT', 'hair TRT' and 'hair TRRT ' but will result in slower rendering light_samples Permalink to light_samples No documentation available lock_frame_noise Permalink to lock_frame_noise By default, the random number generators are seeded by considering the frame number. However, if lock_frame_noise is true, the same seed values are used for each frame, which is typically undesirable. max_depth Permalink to max_depth No documentation available max_diffuse_depth Permalink to max_diffuse_depth No documentation available max_glossy_depth Permalink to max_glossy_depth No documentation available max_hair_depth Permalink to max_hair_depth No documentation available max_mirror_depth Permalink to max_mirror_depth No documentation available max_presence_depth Permalink to max_presence_depth No documentation available max_subsurface_per_path Permalink to max_subsurface_per_path No documentation available pixel_samples Permalink to pixel_samples The square root of the number of primary samples taken for each pixel in uniform sampling mode. For example, a value of 4 will result in 4*4 = 16 uniform pixel samples. presence_threshold Permalink to presence_threshold The presence threshold defines the point at which the accumulated presence can be considered opaque, skipping the generation of presence continuation rays. russian_roulette_threshold Permalink to russian_roulette_threshold The Russian roulette threshold specifies the point at which point Russian roulette is evaluated for direct light sampling and BSDF continuation. The unit is luminance of the radiance. transparency_threshold Permalink to transparency_threshold The transparency threshold defines the point at which the accumulated opacity can be considered opaque, skipping the generation of new transparency rays. |
Beta Was this translation helpful? Give feedback.
-
|
off we go again with a big texture cache |
Beta Was this translation helpful? Give feedback.
-
|
this is the ray depth info |
Beta Was this translation helpful? Give feedback.
-
|
I have uploaded my log but with the larger texture cache we got a frame in 29 minutes 54 seconds down from 1 hr 31 min with the default max texture cache size not sure if the file cache was warmed or not |
Beta Was this translation helpful? Give feedback.
-
|
I have a question: What is the mirror_depth relate to? Where can I find the effective tranmission depth and samples? I am hoping to do a 1:1 comparison using USD Arnold on the same hardware with the same scene. |
Beta Was this translation helpful? Give feedback.
-
|
The max mirror depth is the number of allowed mirror (perfectly specular) interactions a ray is allowed to have. It also applies to perfect transmission/refraction. https://docs.openmoonray.org/user-reference/how-to-guides/ray-depth/#max-mirror-depth |
Beta Was this translation helpful? Give feedback.
-
|
Just to boil this down a bit more with what Keith said: We use the term mirror to differentiate from glossy reflection/transmission lobes. The term lobes in this context is really just an implementation detail about how we describe specific types of light interactions to the integrator. Mirror lobes (with roughness = 0) produce perfectly smooth transmission/reflection, while glossy lobes (with roughness > 0) produce from slightly to heavily blurred transmissions/reflections. These two terms exist as a common distinction between these potentially only slightly different light-surface interactions in many modern renderers, including MoonRay. Practically speaking this distinction may not be all that important in most scenes. Sometimes mirror-like materials are perfectly smooth, or sometimes the material has a very slight roughness. Often production scenes contain materials that produce both 'mirror' and 'glossy' specular lobes. What depth limits should you set? Often again the person doing the final rendering and tuning these final render depth settings (aka the 'lighter') is not the same person who set up the materials (aka the 'surfacer') and doesn't know which materials are considered by the render as 'mirror' or 'glossy', nor do they honestly care. In practice I recommend matching the mirror depth and glossy depth, unless you are really diving deep into optimizing a particular scene where you can leverage this distiction to gain a performance boost. |
Beta Was this translation helpful? Give feedback.
-
|
So sorry for not adequately reading the reference material But I think the distinction between transmission and reflection is not handled in the same way as Solid Angle's Arnold But basically I can make sure the termination of ray depth looks similar enough and make sure the sample count of the BSDF path is similar and the antialiasing samples are similar The glossy vs mirror is simple make them the same and make them the same for specular/transmission in Arnold Now I am not so sure about presence vs total ray depth? Any insight into presence other than the source code? |
Beta Was this translation helpful? Give feedback.
-
|
Trying now as follows |
Beta Was this translation helpful? Give feedback.
-
|
So the results are in
LOL
`` |
Beta Was this translation helpful? Give feedback.
-
|
These are the steps that I followed for the original "stock" ALab before I
saw the link to the DW version:
- Downloaded and unzipped the three files for ALab. (Used wget).
- Used "rsync" to merge the downloaded "fragment" directory into
"ALab/fragment"
- Moved ALab/baked_procedurals to ALab/baked_procedurals.orig
- Moved the downloaded baked_procedurals into ALab.
- Used hd_usd2rdl to convert entry.usda into 'stoat.1004.rdlb' (at frame
1004, with the suggested camera).
- Rendered rdlb with moonray (this is the one that took 6hrs since it
has stock textures.)
…--S
On Sun, Apr 9, 2023 at 5:32 PM Sam Hodge ***@***.***> wrote:
Here's more complete stats using the original (red sweater) ALab USD file
and the DW modified (grey sweater) rdlb version.
This confirms that the out-of-box performance is greatly improved with a
bigger texture cache.
Machine:
CPU: 16-core Intel(R) Xeon(R) Gold 5122 CPU @ 3.60GHz, 128GB
GPU: Nvidia RTX 2080Ti, 12GB
| Alab version | Container | exec_mode | texture_cache_size | total time |
| ------ | ----- | --- | ----- | ---- |
| dw (grey sweater) | apptainer | xpu | 40960 | 01:06:57 |
| dw (grey sweater) | apptainer | vectorized | 40960 | 01:33:42 |
| stock (red sweater) | apptainer | xpu | 40960 | 01:53:52 |
| stock (red sweater) | apptainer | vectorized | 40960 | 02:04:50 |
| dw (grey sweater) | apptainer | xpu | 4000 | 02:29:48 |
| dw (grey sweater) | docker | xpu | 4000 | 02:36:07 |
| dw (grey sweater) | apptainer | vectorized | 4000 |02:56:09 |
| dw (grey sweater) | docker | vectorized | 4000 | 03:03:57 |
| stock (red sweater) | docker | xpu | 4000 | 06:04:50 |
| stock (red sweater) | docker | vectorized | 4000 | 06:25:09 |
I'm curious to know what optimizations DW performed on the ALab set, as
the improvement is substantial and the GPU benefit is much stronger. It
seems like more than just converting textures to .tx format. (The basic
settings appear to be the same.)
Did you take the original file ALab2 files and convert the .exr files to
.tx files and then link the USD files to the updated file extensions
yourself to get the red jumper and white stoat to work correctly?
I have yet to see a render from Moonray from the original files using the
texture files from the stock ALab2 scene expect maybe for a microscope in
isolation
At the point where the folk from DWA uploaded the new .zip with the mip
mapped .tx file I have been working from there
But if I can render from the original with the stoat looking correct , I
would be interested in knowing the reproduction steps to make that happen
—
Reply to this email directly, view it on GitHub
<#61 (reply in thread)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ADNVAYWT62RUT6E3R44VTQTXANII5ANCNFSM6AAAAAAWDFOBKU>
.
You are receiving this because you were mentioned.Message ID:
<dreamworksanimation/openmoonray/repo-discussions/61/comments/5566652@
github.com>
|
Beta Was this translation helpful? Give feedback.
-
|
It's using OIIO, so I think it's fair to say it doesn't need .tx files,
though it clearly is faster with them. (Thanks Larry Gritz and the OIIO
team!)
…On Sun, Apr 9, 2023 at 6:39 PM Sam Hodge ***@***.***> wrote:
I have yet to see a render from Moonray from the original files using the
texture files from the stock ALab2 scene expect maybe for a microscope in
isolation
MoonRay requires that textures be converted to tiled and mipmapped tx
files prior to rendering.
Let's make that a "maybe not" 🤠
—
Reply to this email directly, view it on GitHub
<#61 (reply in thread)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ADNVAYRNLOG3UO3JNBRPC5TXANQDVANCNFSM6AAAAAAWDFOBKU>
.
You are receiving this because you were mentioned.Message ID:
<dreamworksanimation/openmoonray/repo-discussions/61/comments/5566936@
github.com>
|
Beta Was this translation helpful? Give feedback.
-
OIIO is capable of converting textures, yes, but MoonRay makes tiled a mipmapped tx files a requirement (aside from the cookie light filter). |
Beta Was this translation helpful? Give feedback.
-
|
I just want to propose the following method for getting a clean scene to render with |
Beta Was this translation helpful? Give feedback.
-
|
This is great! I'm running something similar, but rather than modify the usda files I hd_usd2rdl'd into an ascii, filtered one big file, and then converted back to rdlb. I don't know if it will help, but I'm using the --oiio option to maketx to see if that gives better performance. |
Beta Was this translation helpful? Give feedback.
-
|
Got some errors in the (huge) rdla file, so I’m switching to your usda method. |
Beta Was this translation helpful? Give feedback.
-
Using High resolution texture means that it will even fill 70Gb of texture cache I need to add more RAM to my VM to see 100% CPU occupancy I am currently running with a 60Gb Texture Cache and will let it run out a few frames overnight on around 60-80% CPU occupancy Sam |
Beta Was this translation helpful? Give feedback.
-
|
NO MISSING TEXTURES! FULL RESOLUTION! LOOKS GREAT!
|
Beta Was this translation helpful? Give feedback.
-
|
Trying again with a texture cache from 60Gb to 75Gb and see if we get a better efficiency. above render is That is a lotta texture calls. |
Beta Was this translation helpful? Give feedback.
-
stoat_smoothed.mp4stoatrender_log.1010.txt Texture cache at 75Gb, brings down the miss ratio but probably need more like 150Gb for this scene to get full CPU occupancy |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Hey the example scenes are not working out for me.
But what would be really hot would be a tutorial about how to output the ALab2 movie from Animal Logic
https://dpel.aswf.io/alab/
Possibly using AWS compute resources and slam out some pixels as a real world "Hello World"
I have been a render TD for more years than I can count, and if I wanted to make an argument to swapping rendering engine.
This would add weight to this arguement
Sam
Beta Was this translation helpful? Give feedback.
All reactions