Blender 2.81/2.82 - OpenCL - Significan slow render times #71479

New Issue

Grzegorz Wereszko · 2019-11-10T21:03:29+01:00

Grzegorz Wereszko commented

2019-11-10 21:03:29 +01:00

System Information
Operating system: Windows 10 - AMD drivers 19.10.1
Graphics card: RX Vega 64

Blender Version
Broken: 2.81 - eefd806afc & 2.82 - 23e1fb365b
Worked: 2.80 - official build

Short description of error
Cycles OpenCL - Using default BMW pulled from Blenders benchmark section, rendering in latest 2.81/2.82 experimental builds have significant slow render times

Exact steps for others to reproduce the error
Car Demo scene from : https://www.blender.org/download/demo-files/
Blender 256 128 64 32 Build
2.80 1m28s 1m30s 1m39s 1m56s official
2.81 2m.34s 2m34s 2m43s 3m20s eefd806afc
2.82 2m.35s 2m34s 2m43s 3m25s 23e1fb365b

more details and test by others reporting similar issue on blender nation post:
https://blenderartists.org/t/very-slow-gpu-rendering-in-cycles-opencl-blender-2-81/1186239

**System Information** Operating system: Windows 10 - AMD drivers 19.10.1 Graphics card: RX Vega 64 **Blender Version** Broken: 2.81 - eefd806afc15 & 2.82 - 23e1fb365b65 Worked: 2.80 - official build **Short description of error** Cycles OpenCL - Using default BMW pulled from Blenders benchmark section, rendering in latest 2.81/2.82 experimental builds have significant slow render times **Exact steps for others to reproduce the error** Car Demo scene from : https://www.blender.org/download/demo-files/ Blender 256 128 64 32 Build 2.80 1m28s 1m30s 1m39s 1m56s official 2.81 2m.34s 2m34s 2m43s 3m20s eefd806afc15 2.82 2m.35s 2m34s 2m43s 3m25s 23e1fb365b65 more details and test by others reporting similar issue on blender nation post: https://blenderartists.org/t/very-slow-gpu-rendering-in-cycles-opencl-blender-2-81/1186239

Grzegorz Wereszko commented

2019-11-10 21:03:29 +01:00

Added subscriber: @Prodeous

blender-admin commented

2019-11-10 21:03:29 +01:00

#71859 was marked as duplicate of this issue

Jeroen Bakker commented

2019-11-11 10:14:47 +01:00

Added subscriber: @Jeroen-Bakker

Jeroen Bakker commented

2019-11-11 10:14:47 +01:00

Yes, I have noticed this also. The reason is that there are a lot of nodes/complexity (4d noise) added to cycles. The AMD compiler doesn't do an excellent job at reusing registries resulting that less work that can be done in parallel and speeds down the rendering process dramatically. This needs further investigation.

Adam Burke commented

2019-11-11 12:30:53 +01:00

Added subscriber: @Adam-Burke

Brecht Van Lommel commented

2019-11-11 16:01:31 +01:00

Added subscriber: @OmarEmaraDev

Omar Emara commented

2019-11-11 19:00:57 +01:00

After benchmarking the BMW scene on 23564583a4988778b4c43496fd21818b286f6ba1 and 45d4c925799e94c6d442a9a9066af2d3305724e1, which are the commits before and after the new noise functions were added, there isn't any regression.

Since the Noise node is the only node in the BMW scene that we touched during GSoC, then the regression must be due to the increased size of the kernel as Brecht described before, perhaps due to the Voronoi node.

I will bisect the code and try to find the source of the regression.

After benchmarking the BMW scene on `23564583a4988778b4c43496fd21818b286f6ba1` and `45d4c925799e94c6d442a9a9066af2d3305724e1`, which are the commits before and after the new noise functions were added, there isn't any regression. Since the Noise node is the only node in the BMW scene that we touched during GSoC, then the regression must be due to the increased size of the kernel as Brecht described before, perhaps due to the Voronoi node. I will bisect the code and try to find the source of the regression.

Omar Emara commented

2019-11-11 22:13:26 +01:00

Just confirmed that the new Voronoi is the source of the regression. What can we do about this?

Jeroen Bakker commented

2019-11-12 11:03:59 +01:00

@OmarEmaraDev first check if we can speed it up by commenting out the case 4: in svm_node_tex_voronoi and if this speeds the rendering up we could add this case in a special NODE_FEATURES construction.

My experience is that sometimes small changes can have huge impact in OpenCL rendertimes. It is not based on what is used, but what is compiled.

Last week we did several performance tests and it seems that the Vega architecture is really dropping behind. The Polaris architecture seems to hold up better. I don't have any data about NAVI atm.

The next results are from BMW on an AMD RX480

2.80 (3:10)
2.81 (5:48)
2.81 excluding 4d voronoi (4:05)
2.81 excluding 4d voronoi+mus+noise (4:05)
2.81 excluding 4d voronoi+3d smooth (3:52)
2.81 excluding 4d voronoi+3d (3:52)
2.81 excluding 4d voronoi+3d smooth + 2d smooth (3:50)

@OmarEmaraDev first check if we can speed it up by commenting out the `case 4:` in `svm_node_tex_voronoi` and if this speeds the rendering up we could add this case in a special `NODE_FEATURES` construction. My experience is that sometimes small changes can have huge impact in OpenCL rendertimes. It is not based on what is used, but what is compiled. Last week we did several performance tests and it seems that the Vega architecture is really dropping behind. The Polaris architecture seems to hold up better. I don't have any data about NAVI atm. The next results are from BMW on an AMD RX480 * 2.80 (3:10) * 2.81 (5:48) * 2.81 excluding 4d voronoi (4:05) * 2.81 excluding 4d voronoi+mus+noise (4:05) * 2.81 excluding 4d voronoi+3d smooth (3:52) * 2.81 excluding 4d voronoi+3d (3:52) * 2.81 excluding 4d voronoi+3d smooth + 2d smooth (3:50)

Brecht Van Lommel commented

2019-11-12 15:04:37 +01:00

Added subscriber: @brecht

Brecht Van Lommel commented

2019-11-12 15:04:38 +01:00

Tagging with 2.81 since this fix really should go in the release.

Jeroen Bakker self-assigned this 2019-11-13 08:00:40 +01:00

blender-admin commented

2019-11-13 08:03:44 +01:00

This issue was referenced by e527544b76

This issue was referenced by e527544b766bb01fb95bff1d98fe5a246aaf7398

Jeroen Bakker was unassigned by Brecht Van Lommel

2019-11-13 15:43:56 +01:00

Brecht Van Lommel commented

2019-11-13 15:43:56 +01:00

I think this is sufficiently resolved now for the 2.81 release.

During the 2.82 cycle we can look at getting back a bit more of the performance since it's still not fully the same.

I think this is sufficiently resolved now for the 2.81 release. During the 2.82 cycle we can look at getting back a bit more of the performance since it's still not fully the same.

Brecht Van Lommel commented

2019-11-13 15:45:03 +01:00

@Jeroen-Bakker, not sure if you wanted to be assigned this follow up task, up to you.

Grzegorz Wereszko commented

2019-11-14 22:32:42 +01:00

Just wanted to say thanks for the work you are doing. I can confirm that render time is much better on my setup as well.

2.80 - 1m28s
2.81 - 2m34s
Nov14 builds
2.81 - 1m46s

Just wanted to say thanks for the work you are doing. I can confirm that render time is much better on my setup as well. 2.80 - 1m28s 2.81 - 2m34s Nov14 builds 2.81 - 1m46s

Jeroen Bakker commented

2019-11-24 16:16:58 +01:00

Added subscriber: @tm9000

Marcin commented

2019-11-25 21:22:21 +01:00

Added subscriber: @MarcinJ

Marcin commented

2019-11-25 21:22:21 +01:00

I would like to confirm that the issue is still present on official 2.81.

Windows 10
GPU: Radeon VII (19.11.3).
CPU: Ryzen 3800X

I've tested rendering using Class Room scene (256*256)
2.80 (official) 2m40s
2.81 (official) 3m20s

I would like to confirm that the issue is still present on official 2.81. Windows 10 GPU: Radeon VII (19.11.3). CPU: Ryzen 3800X I've tested rendering using Class Room scene (256*256) 2.80 (official) 2m40s 2.81 (official) 3m20s

Joan commented

2019-11-27 11:42:21 +01:00

Added subscriber: @esciron

Krystian Zajdel commented

2019-12-04 18:35:20 +01:00

Added subscriber: @enenra

Yusuf Umar commented

2019-12-08 10:08:26 +01:00

Added subscriber: @ucupumar

Matthew Wiebe commented

2019-12-16 19:07:03 +01:00

Added subscriber: @cerebral

Matthew Wiebe commented

2019-12-16 19:07:03 +01:00

Thank you so much for this task and making this a priority! Maybe my new GPU won't have been a mistake after all <3

Benjamin Xiao commented

2019-12-31 00:11:01 +01:00

Added subscriber: @UrbenLegend

Jeroen Bakker self-assigned this 2020-01-21 08:59:16 +01:00

Jeroen Bakker commented

2020-01-21 08:59:16 +01:00

Due to the changes done we are not in line with 2.80. There is still one task open to be able to create OpenCL2.0 contextes which will improve memory intensive renders (Barbershop). Will mark this issue with TODO and close it when the patch is finished

Jeroen Bakker removed their assignment 2020-02-21 09:46:07 +01:00

Brecht Van Lommel commented

2020-02-21 13:50:12 +01:00

Marking as high priority since this is something we want to address for the 2.83 release. @Jeroen-Bakker has submitted a patch to use OpenCL 2.0 which will helps, but currently requires some manual configuration which I'll try to find an automatic solution for.

For commented

2020-02-24 23:19:34 +01:00

Added subscriber: @Forshu-2

Bartosz Moniewski commented

2020-03-11 17:31:21 +01:00

Added subscriber: @BartekMoniewski

Aaron Lalor commented

2020-03-14 10:18:03 +01:00

Added subscriber: @CoderWolf

blender-admin commented

2020-03-24 00:09:06 +01:00

This issue was referenced by blender/cycles@d3dada9dfe

This issue was referenced by blender/cycles@d3dada9dfe44b70a796118c6d855fa2da6e9978d

blender-admin commented

2020-03-24 00:09:06 +01:00

This issue was referenced by 5801a016d4

This issue was referenced by 5801a016d4cf2384274b0f813e29d61662dc8ee2

blender-admin commented

2020-03-24 01:27:38 +01:00

This issue was referenced by blender/cycles@8817bed777

This issue was referenced by blender/cycles@8817bed777242ffa6526dc260dfbbc240bf367cb

blender-admin commented

2020-03-24 01:27:38 +01:00

This issue was referenced by 2bec6f1f06

This issue was referenced by 2bec6f1f06dff400d3b5abd08064ef05312a29bf

blender-admin commented

2020-03-24 19:42:23 +01:00

This issue was referenced by blender/cycles@c26a983fe4

This issue was referenced by blender/cycles@c26a983fe457b56bc09350fbcc5d98be6c02727c

blender-admin commented

2020-03-24 19:42:23 +01:00

This issue was referenced by 394a1373a0

This issue was referenced by 394a1373a0cd20b7d0660df4bf80e1231e33cba9

Brecht Van Lommel commented

2020-03-24 20:17:51 +01:00

Changed status from 'Confirmed' to: 'Resolved'

Brecht Van Lommel closed this issue

2020-03-24 20:17:51 +01:00

Brecht Van Lommel self-assigned this 2020-03-24 20:17:51 +01:00

Brecht Van Lommel commented

2020-03-24 20:17:51 +01:00

From my testing this is resolved now. If others still find significant performance regressions please let me know.

For reference, here are my performance measurements:
{F8427295, size=full}

I have this set up now to test CPU, CUDA, Optix and OpenCL performance, so we should be able to catch issues like this early in the future.

From my testing this is resolved now. If others still find significant performance regressions please let me know. For reference, here are my performance measurements: {[F8427295](https://archive.blender.org/developer/F8427295/AMD_radeon_pro_WX9100.png), size=full} I have this set up now to test CPU, CUDA, Optix and OpenCL performance, so we should be able to catch issues like this early in the future.

Grzegorz Wereszko commented

2020-03-24 21:46:24 +01:00

Unfortunately I can NOT confirm this issue is solved.

I'm using Vega 64GPu (also tested on 2 other vega 56 GPU's i have in my build. times blow are for Vega64.

WIndows 10, fresh install (last weekend)
Radeon drivers 20.3.1

BMW27_gpu.blend file pulled from blender.org site. no changes to any settings.

Times are average of 3 runs, excluding first as that includes also kernel compile.

blender 2.80 - 1m27s
Blender 2.82a - 1m54s
Blender 2.83 - 1m58s (todays build - 94b8166a8b)

Unfortunately I can NOT confirm this issue is solved. I'm using Vega 64GPu (also tested on 2 other vega 56 GPU's i have in my build. times blow are for Vega64. 1. WIndows 10, fresh install (last weekend) 2. Radeon drivers 20.3.1 BMW27_gpu.blend file pulled from blender.org site. no changes to any settings. Times are average of 3 runs, excluding first as that includes also kernel compile. blender 2.80 - 1m27s Blender 2.82a - 1m54s Blender 2.83 - 1m58s (todays build - 94b8166a8b05)

Brecht Van Lommel commented

2020-03-24 22:12:52 +01:00

94b8166a8b does not include the fix yet, it will be in the next build.

94b8166a8b does not include the fix yet, it will be in the next build.

Grzegorz Wereszko commented

2020-03-24 22:37:01 +01:00

Classroom scene - file dated 2019-06-13 (pulled from blender https://www.blender.org/download/demo-files/

changes : set to GPU and 256x256 tiles

2.80 - 3m12s
2.82a - 5m34s
2.83 - 7m15s... (build - 94b8166a8b)

Classroom scene - file dated 2019-06-13 (pulled from blender https://www.blender.org/download/demo-files/ changes : set to GPU and 256x256 tiles 2.80 - 3m12s 2.82a - 5m34s 2.83 - 7m15s... (build - 94b8166a8b05)

Grzegorz Wereszko commented

2020-03-24 22:37:54 +01:00

fingers crossed, and eager to test. will test again tomorrow and report back.

Brecht Van Lommel commented

2020-03-24 22:47:21 +01:00

There's a new Windows build now with the fix.
https://builder.blender.org/download/

There's a new Windows build now with the fix. https://builder.blender.org/download/

Grzegorz Wereszko commented

2020-03-24 23:17:50 +01:00

Thanks, here are results, confirming the fix.

BMW
Blender 2.80 - 1m27s
Blender 2.83 - 1m23s (fixed version)

Classroom
Blender 2.80 - 3m12s
Blender 2.83 - 2m57s (fixed version)

nice to get even faster renders. don't see any visual differences between the version.

Again thanks for the work.

Thanks, here are results, confirming the fix. BMW Blender 2.80 - 1m27s Blender 2.83 - 1m23s (fixed version) Classroom Blender 2.80 - 3m12s Blender 2.83 - 2m57s (fixed version) nice to get even faster renders. don't see any visual differences between the version. Again thanks for the work.

Sign in to join this conversation.

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Blender 2.81/2.82 - OpenCL - Significan slow render times #71479