Omni light samples faster than vraylight?

[q976431]

Yep, that's the best solution, but we don't have it in Vantage.

Greetings,
Vladimir Nedev

I look forward to your follow-up updates. I like this software very much. It has changed my production method. It has become my main production tool. Although it still has some shortcomings, sometimes I choose to return to Vray, but I prefer to use it as much as possible, and come here to discuss with you how to better achieve the desired effect.

[q976431]

Calculations are simpler for omni lights, yes.
However, I doubt you will notice a difference on a complex scene.

Maybe your omni light doesn't have "shadows" turned on ?
When shadows are disabled, the extra shadow rays don't have to be fired into the scene which will increase the FPS.

You can disable shadows for the VRayLight as well bringing you a similar improvement in FPS.

Greetings,
Vladimir Nedev

New news: vraylight will indeed bring more performance consumption than omin. Because I tested a scene with more than 100 vraylights, which significantly reduced FPS. If it was omin, I could use nearly a thousand omins. Maybe we have to wait until the emergence of adaptive sampling to alleviate this problem?

[q976431]

Increasing the number of light tree samples doesn't help ?

Greetings,
Vladimir Nedev

If I turn off the lights tree function, is it necessary for me to add the parameters of samples when I finally output? I tested a scene, no matter samples300 or 100, the final rendering quality does not change significantly, but the rendering time is very different. Can I reduce the parameters of samples after turning off lights tree? Or can I think that samples are parameters designed for the function of light tree?

[vladimir.nedev]

New news: vraylight will indeed bring more performance consumption than omin. Because I tested a scene with more than 100 vraylights, which significantly reduced FPS. If it was omin, I could use nearly a thousand omins. Maybe we have to wait until the emergence of adaptive sampling to alleviate this problem?

I am not sure, what do you mean by "significant" ? Is this with the light tree OFF or ON ?

If I turn off the lights tree function, is it necessary for me to add the parameters of samples when I finally output? I tested a scene, no matter samples300 or 100, the final rendering quality does not change significantly, but the rendering time is very different. Can I reduce the parameters of samples after turning off lights tree? Or can I think that samples are parameters designed for the function of light tree?

The "samples" parameter determines how many image samples Vantage will take.
Yes, when you turn off the light tree, each image sample will be less noisy, so you might be able to use fewer.

However, there might be GI noise, which is not affected by the light tree being off or on, so you might need more image samples to clear the GI noise.

Greetings,
Vladimir Nedev

[q976431]

I am not sure, what do you mean by "significant" ? Is this with the light tree OFF or ON ?



The "samples" parameter determines how many image samples Vantage will take.
Yes, when you turn off the light tree, each image sample will be less noisy, so you might be able to use fewer.

However, there might be GI noise, which is not affected by the light tree being off or on, so you might need more image samples to clear the GI noise.

Greetings,
Vladimir Nedev

Sorry! This may be a mistake caused by the translation. I mean, it obviously lowers the FPS. I usually turn off the light tree now, because I can't get the right result under hundreds of lights after turning on the light tree. I have reported this problem before.

Vraylight affects FPS more than omin. I hope it can optimize this matter, because Vraylight can bring more delicate results, but now I dare not use it because of a large amount of use, For example, 100 are used, then it will reduce FPS.Omin can use more without affecting FPS.


Maybe a document can teach us how to set the rendering parameters of vantage correctly, because I see that there are many parameters about rendering rays in vantage, but when rendering image sequences under the 3dsmax real-time link plug-in, only a few key items are available, and other options are not necessary items when outputting the final sequence? They don't need to adjust, just keep the default values?

[vladimir.nedev]

Vraylight affects FPS more than omin. I hope it can optimize this matter, because Vraylight can bring more delicate results, but now I dare not use it because of a large amount of use, For example, 100 are used, then it will reduce FPS.Omin can use more without affecting FPS.

There shouldn't be a significant difference between VRayLight and Omni lights in terms of speed.
Are you sure you have "shadows" enabled for the omni light ?
Shadows are off by default for Omni lights, while they are on by default for VRayLight.

Sorry! This may be a mistake caused by the translation. I mean, it obviously lowers the FPS. I usually turn off the light tree now, because I can't get the right result under hundreds of lights after turning on the light tree. I have reported this problem before.

Yes, I have your scene, but we haven't gotten to it.

Maybe a document can teach us how to set the rendering parameters of vantage correctly, because I see that there are many parameters about rendering rays in vantage, but when rendering image sequences under the 3dsmax real-time link plug-in, only a few key items are available, and other options are not necessary items when outputting the final sequence? They don't need to adjust, just keep the default values?

We try to set the most sensible defaults, so you don't have to change them.
If there is an issue, we most likely have to improve our algorithms.
The perfect setup is scene dependent however.

Greetings,
Vladimir Nedev

[q976431]

There shouldn't be a significant difference between VRayLight and Omni lights in terms of speed.
Are you sure you have "shadows" enabled for the omni light ?
Shadows are off by default for Omni lights, while they are on by default for VRayLight.



Yes, I have your scene, but we haven't gotten to it.



We try to set the most sensible defaults, so you don't have to change them.
If there is an issue, we most likely have to improve our algorithms.
The perfect setup is scene dependent however.

Greetings,
Vladimir Nedev

1. Yes, I also enabled shadows for omin lights. Their performance is so different. At first, I even thought that vraylight lights should have better performance. When I try. I found that using omin has better performance. When I use more than 100 vraylights, it significantly reduces FPS. If it is the same number of omins, this problem will not exist. If necessary, I can provide a case scenario.

Because I have experience in making many large scenes with vantage, sometimes there may be thousands of lights and dozens of forests in the scene, so the problems I encounter may not be encountered by others at present. Maybe the difference occurs when there are hundreds of quantities, and there is no difference between several vraylight and omin,Maybe these feedback can help vantage optimize better.

I even found that in some scenes with a lot of lights: some frames only need 50 samples to get a good effect, and some frames need to increase samples to get a better effect. However, if the same samples are used in the whole time axis, such as 200 samples, it may lead to a long rendering time. Now I will use 50 samples for some frames and 200 samples for some frames, The way to distinguish them is the height of the camera from the ground. The higher it is, the more samples it needs to sample the details of plants. When the camera is closer. 50 samples can be sampled quickly and accurately without any artifacts. This operation method can save 1/3 of my rendering time without affecting the final quality

[vladimir.nedev]

1. Yes, I also enabled shadows for omin lights. Their performance is so different. At first, I even thought that vraylight lights should have better performance. When I try. I found that using omin has better performance. When I use more than 100 vraylights, it significantly reduces FPS. If it is the same number of omins, this problem will not exist. If necessary, I can provide a case scenario.

Yes, please, if you could send me both setups, with VRayLight and Omni light, where the FPS is much different.

The way to distinguish them is the height of the camera from the ground. The higher it is, the more samples it needs to sample the details of plants. When the camera is closer. 50 samples can be sampled quickly and accurately without any artifacts.

Can you provide a scene or screen shots for this too ?

Greetings,
Vladimir Nedev

[q976431]

Yes, please, if you could send me both setups, with VRayLight and Omni light, where the FPS is much different.



Can you provide a scene or screen shots for this too ?

Greetings,
Vladimir Nedev

1：good I will send a test scenario to explain their differences.

2：I will try to send a rendered finished MP4, which may take a few days, because it is still in production (please keep it confidential for me, it is an unpublished project file). There are two kinds of samples in the rendering sequence. At close range, I use 50 samples (16 seconds per frame), and when the lens is pulled up, I use 200 samples (2 minutes per frame). Output them separately in sections. Then combine them into output MP4 in AE.
If 200 samples are used at close range, the time of a single frame is close to 2 minutes. I use this method. Saves rendering time for close shot segments.
I tested in this lens. There is no difference between using 200 samples and 50 samples at close range, but there will be a big gap in rendering time. But after the lens is pulled up, I must increase the number of samples, because if the number of samples is insufficient, the picture noise will appear.

[vladimir.nedev]

1：good I will send a test scenario to explain their differences.

Thank you for the scene.
The huge difference ( the omni light version is 3 times faster) is caused mostly by two factors
- the Far Attenuation option that you have enabled on the Omni light
- the Specular option that you have disabled on the Omni light
The far attenuation option causes a lot less shadow rays to be fired for the omni lights.
The rectangle light doesn't have a similar option.

However, I want to stress again that when you have this many lights and disable the light tree, results might be incorrect.
For this scene you can
- leave the light tree enabled
- increase the max sec. ray clamp (e.g. from 20 to 500)
- increase the light tree samples to 16
You will have correct results and a much better frame rate.
(I still haven't looked at your other scene, where the above setup doesn't seem to work).

Greetings,
Vladimir Nedev

[q976431]

Thank you for the scene.
The huge difference ( the omni light version is 3 times faster) is caused mostly by two factors
- the Far Attenuation option that you have enabled on the Omni light
- the Specular option that you have disabled on the Omni light
The far attenuation option causes a lot less shadow rays to be fired for the omni lights.
The rectangle light doesn't have a similar option.

However, I want to stress again that when you have this many lights and disable the light tree, results might be incorrect.
For this scene you can
- leave the light tree enabled
- increase the max sec. ray clamp (e.g. from 20 to 500)
- increase the light tree samples to 16
You will have correct results and a much better frame rate.
(I still haven't looked at your other scene, where the above setup doesn't seem to work).

Greetings,
Vladimir Nedev

Thank you for your solution. I have consulted relevant questions before. You asked me to turn off clamp sec rays, but I still couldn't get good results after the test. The previous test scenario had thousands of lights. If this is the case, (clamp sec rays) how can I determine how many values are appropriate?

At the same time, I want to explain again. If there are thousands of lights, the maximum value of the light tree is only 16 That may be insufficient.
Just like the lighting test scene I sent you earlier, can you render the same effect as turning off the light tree through the value of clamp sec rays? I tried, but I couldn't.

In addition. According to the parameters you just provided, light tree 16. max sec. ray clamp (e.g. from 20 to 500). I rendered a clip, and the sample was given to 200, and the light jittered obviously


Maybe your method works, but the value of lights tree is limited to 16, which is not applicable to scenes with thousands of lights. If you can temporarily release this limitation, send me an internal beta version, let me test a larger value, and then match it with the value of clamp sc. rays, then I can get a more accurate answer, but now I can't. Lights tree can only be 16 at most. Scene with thousands of lights, the light sampled by this value has a very obvious distortion effect.

in my opinion. If the number of scene lights increases gradually, you need to increase the value of the light tree and adjust Max sec The parameter of ray clamp can effectively improve this problem. Does limiting the light tree to 16 also limit the maximum number of lights that can be arranged in my scene, because this value is too low to deal with the situation that there are many lights in the scene. Maybe a hundred lights can, but what if there are thousands of lights?

Because I can't set a value more than 16, I can only guess based on this situation, but I think it's effective to continue to increase it, because I have compared 1, 8 and 16, and there is a clear gap between them.

[vladimir.nedev]

Just like the lighting test scene I sent you earlier, can you render the same effect as turning off the light tree through the value of clamp sec rays? I tried, but I couldn't.

It might be impossible without improvements to the light tree.

In addition. According to the parameters you just provided, light tree 16. max sec. ray clamp (e.g. from 20 to 500). I rendered a clip, and the sample was given to 200, and the light jittered obviously

You probably need more than 200 samples, when the light tree is enabled., as each sample is more noisy.

Maybe your method works, but the value of lights tree is limited to 16, which is not applicable to scenes with thousands of lights. If you can temporarily release this limitation, send me an internal beta version, let me test a larger value, and then match it with the value of clamp sc. rays, then I can get a more accurate answer, but now I can't. Lights tree can only be 16 at most. Scene with thousands of lights, the light sampled by this value has a very obvious distortion effect.

Allowing more light tree samples is not feasible.
Each one requires ~50 MBs of additional GPU memory.
Also, I don't think the additional samples will improve the result by much with the current algorithm.
I will do some tests when I have time.

We have to improve the light tree or implement something like RTX DI.

Does limiting the light tree to 16 also limit the maximum number of lights that can be arranged in my scene, because this value is too low to deal with the situation that there are many lights in the scene. Maybe a hundred lights can, but what if there are thousands of lights?

As long as the light tree is enabled, there isn't a limit on the number of (non-infinite) lights.

Obviously from your other scene, there is an issue that needs to be resolved, but I don't think allowing more light tree samples is the solution.

Greetings,
Vladimir Nedev

[q976431]

Obviously from your other scene, there is an issue that needs to be resolved, but I don't think allowing more light tree samples is the solution.

Yes, the method I use now may not be correct But it can still be used for the time being. Therefore, I upgraded my graphics card to cope with the FPS reduction caused by turning off the optical tree. hope this problem can be solved in the future.




In addition, I want to explain why you told me that the result may not be so accurate after turning off the light tree, but I am still used to turning off the light tree:
Because when I turn off the light tree, I can quickly get the light sampling results.I can adjust it right away.(But FPS is indeed much lower This is a price). When I turn on the light tree, even if I set the value to 16 I still need to wait a long time to get the sampling results. It seems that it is gradually brightening up.It takes a long time to continue adjusting
I have a GIF to illustrate this, but it can't be shown here,So I sent it to your email