Add Legacy/PT GI method "toggle switch"

Ideas & Requests
1

I’d like to suggest D5 add a toggle switch which would allow users to quickly toggle between the legacy GI and the path tracer. My suggestion would be to put the toggle switch in the Display menu, just as the path tracer and realtime quality setting as there already.

Perhaps add the switch option to the Even with higher end GPU’s, the new path tracing with video rendering is exponentially slower than the legacy.

2

Hey tim,

Thanks a lot for sharing your suggestion and for explaining your use case.

Just to clarify, the legacy GI is currently kept only as a compatibility option for older projects, which is why it’s located under Preferences.
Our future development will focus on improving the path tracer’s performance and quality, so we don’t plan to expand support for the legacy GI. You can still enable it via Preferences for your existing projects if needed.

Best regards,
Lia

3

Thanks for the reply. The performance impact of the path tracer for video rendering in particular is quite large- which is to be expected. I’m hopeful the team can improve performance in that regard, either by allowing the user to dial down quality settings further (alternate GI method) or via software improvements. I’ll go ahead and state my preference that I would be willing to lower GI quality only on video output in order to save time (currently, hours, with the path tracer) as is what is needing to be done for the time being in my case. Or else I simply cannot offer video rendering for my peojects, as it isn’t cost effective for my time.

4

Hi tim,

Thanks so much for sharing your thoughts with us, we really appreciate the detailed feedback.

In the 2.11.1 hotfix, the path tracer has been significantly optimized, and with default settings, the video rendering speed of PT and legacy GI should not differ much.

To better understand your case, could you let us know the PT parameters you’re using (for example, whether custom settings are enabled and what values are set)? If convenient, it would also help a lot if you could share your D5 project file with us for testing.

We’ll continue to work on improving the rendering efficiency of PT to make it more practical for video workflows.

Best regards,
Lia

5

I reached out to the support email (support@d5techs.com) on Sept. 21st about video performance but haven’t heard back yet. I was particularly asking if anyone there could test a model/project I have on a 5090 to see if the performance of that GPU would justify the cost for my work. I’ve checked the benchmark results posted but the benchmark model and scene isn’t particularly heavy.

In the 2.11.1 hotfix, the path tracer has been significantly optimized, and with default settings, the video rendering speed of PT and legacy GI should not differ much

I haven’t found this to be the case. It might have been improved, but PT on a video is very long. I’m within VRAM limitations and have monitored system performance during renders and can see nothing is being throttled, etc.

To better understand your case, could you let us know the PT parameters you’re using (for example, whether custom settings are enabled and what values are set)?

I typically output 1080p/30 (2k/30 if I have LOTS of down time) video with custom PT setting the same as most of my stills, which is 1/3/64.

Just to give an ideas as to what my projects look like, here’s some general stats:

  • A large project for me (as is the current case):
    30k sq. ft. floorplan (retail/grocery store) with ~450 light objects (1:1 representation of the lighting plan). A 3D model paired with a real-time render software like D5 is an excellent tool in helping to design a lighting plan.
  • Video’s are full-scale walk-through paths of the entire floorplan all in one clip. Typically that’s just shy of 2 minutes of video. I expect it to take some time to render that length, for sure. My stills are ~2 minutes each to render at 1440p as a reference point to speed.

Hardware is an Intel 8700k (OC’d to 4.4GHz) paired with an RTX3080 12GB. I’m intending to update hardware shortly, hence the debate between the 5080 and the 5090. Additional VRAM would be nice but of particular interest is the massive increase in RT cores (84 vs 170) between the two and if that makes a difference for D5.

1 Like
6

@tim1 Sure, you can send me the file, either share the download link with me in a private message, or email it to li.huang@d5render.com.

We’ll run the test and share the results with you afterward.

Please rest assured that your file will be kept strictly confidential and used only for internal testing.

7

I emailed you the project file directly. Thank you for taking the time to do this for me.

1 Like
8

Hi @tim1 ,

Thanks again for sharing your scene with us. We ran a series of tests on our side using your project file, and here are the results:

Test machine

  • CPU: 13th Gen Intel(R) Core™ i7-13700KF
  • RAM: 64 GB
  • GPU: NVIDIA GeForce RTX 4080 SUPER
  • Driver: 32.0.15.7688

Video render tests (2K, 30fps)

  • Legacy GI
    • Clip 1 (70s): 14:57
    • Clip 2 (20s): 06:07
  • PT (default settings)
    • Clip 1 (70s): 20:44
    • Clip 2 (20s): 08:57
  • PT (custom: 1,1,64)
    • Clip 1 (70s): 21:21
    • Clip 2 (20s): 09:10
  • PT (custom: 1,3,64)
    • Clip 1 (70s): 36:19
    • Clip 2 (20s): 15:34

For this scene, the performance looks within expectations, the render times between Legacy GI and PT (default) are not very different. The longer render times mainly come from using higher custom PT parameters.

We’ll continue to optimize PT efficiency in future versions so you can expect even better results going forward.

Thanks again for your detailed feedback.

9

Thank you again for testing. As reference, my machine has an RTX3080 12GB and it rendered the longer Clip 1 (70s) in 1h6m. That gives me a great benchmark example to help decide upgrading to a newer GPU.

1 Like