I would show you the Best Redshift Sampling Setup and this is kind of the way that I approach my scenes to get the best render output. So this method you just sort of focusing your sampling areas where you need them.
That’s definitely not an official way of doing it but it’s kind of the way I’ve kind of discovered the worked well for me. I think this will just be an interesting overview of the way that I approach Redshift Sampling Setup.
Redshift Global Illumination (GI)
The first thing I do is enable the Global illumination (GI), set to:-
Primary GI Engine – Brute Force
Secondary GI Engine – Irradiance Point Cloud
Redshift Adaptive Sampling
When I start setting up the sampling, first want to think about the anti-aliasing I think it’s gonna need.
So it seems to have no motion blur no depth of field and not to contrasty texture. Probably the minimum I would ever use i.e, Samples min- 4, Samples max- 16 is the default I think it’s still pretty low.
So normally I would times two for example (16*2) it automatically takes the value i.e 32. In this way, we increase the value by not putting random numbers, so keep that in mind again.
And we use samples min- 4, Samples max- 64 is just my opinion, but I think 4 and 64 was kind of the minimum amount you would probably use for production same for final render. Even this is a little bit low.
This is the 1st test that we can see how much noise present and how the scene looks like. You can see here in this scene that there is lots of noise present.
By supporting these values are a little bit more we’re actually giving redshift more of an opportunity to be adaptive in its sampling. Now if the scene has motion blur and a little bit of depth of field, I will start pushing these values up further.
In this scene, there is a very small amount of motion blur and depth of field so we push samples max 2 times again. (samples min- 4, Samples max- 128). Increase the sample max as depend on your scene, In this scene, I’m satisfied with the (Samples max- 128) and I’m sticking with it.
If the (Samples max- 256) I probably start increasing the (samples min) as well that I did in the above example (Sample min- 8). It just kind of meaning that we are keeping spread between these two samples (Max and Min), we’re not letting the samples min get too far behind.
If you’re using too high amounts of samples Max and the too-small amount the samples min, then you run the risk of redshift miscalculation, and it’s not putting enough samples in the scene before starts figuring out with needs to add any more by the Adaptive threshold.
Now If the scenes get stronger with a depth of field and motion blur increasing the sample max further level up to 512 or 1024 or 2048 as a requirement. And also with the same time increase the sample max maybe 16 or 32. And we keep going, to be honest.
Now you notice with this current setting there is still grain in the output. So we will approach it other by adjusting Global Illumination and Sample overrides.
Keep in Mind
In the Redshift IPR Window, always check the render output in bucket mode to show the actual result.
Redshift Sampling Overides
We are gonna look at all the lights, so we will enable the clay render mode.
Also, we are gonna turn off the GI for the moment, just disable the primary G.I engine to None.
Enable the Bucket Rendering for the actual result
Already this is looking pretty good but I think the noise here is still a little bit much in the shadows.
So I’m gonna come to Sampling overrides and enable the Light overrides, this will replace all light samples and giving the detail in samples as a result you get noiseless render.
Now there’s no point having anything less than 128 in Light Samples. You remember the reason for this is because 128 divided by 128 mean you’re getting 1 secondary for every primary ray.
There’s no point having anything less than this value here then that would be pointless and Redshift will not calculate
Straight away I’m gonna double it’s not because we’re probably gonna need more than that. So in my scene, I am satisfied with the 512 Light samples. You can double this up until the light not clean up.
So essentially 512 divided by 128 getting 4 kinds of light rays for every primary ray. so a sampling this light 4 times to every kind of time we simply pixel.
Adjusting Redshift Global illumination
Now we have a look at the Global illumination (GI), so I switched back on to a primary G.I engine, set to Brute force, that we turn off for calculating the light samples.
Come down to the Brute force Number of rays and again I’m gonna take the unified sample max 128 as the default number put in the Number of rays here in the global illumination parameter. And again this is because anything less than this would be less than the sample so redshift will automatically round up.
So this is 1 Second sampling right per primary ray. So you have to double that up until the primary rays i.e the main light rays clean up and there is no noise. I am pretty much 4 G.I rays for every primary camera ray.
In my scene, I am satisfied with the 512 number of rays
Now you disable the clay mode in the Redshift IPR Window, this mode is used only when you adjust the Light samples.
Now the reflection effects already look pretty good here in the scene I think it will help just to have a few more samples there. So I’m gonna go for maybe 2 reflections ray for every primary camera ray.
So I copy the value from the unified sample max i.e, sample max 128, and paste the value in the reflection sample. I’m gonna multiply that value by 2 i.e 256. (256/128=2) that is 2. We’ve got 2 reflection rays for every primary camera ray.
This is gonna help the anti-aliasing a bit, so the anti-aliasing doesn’t have to do all the work for the reflection. Because this isn’t very strong effects not really blur I think that’s going to be enough.
I actually feel the same way about the refraction I think it needs a little bit but probably not too much. I’m gonna use the same value they have 256. Which will give us 2 refraction rays for every primary ray.
The reason it’s faster because I’m now actually focusing on where my samples to be. So we can see the ratios between all samples Max and other secondary (samples i.e the sampling override) you can see
For every primary ray, it comes in 2 reflection rays 2 refraction rays. We’re getting 4 light rays and 4 Global illumination rays.(for example- 256 reflection is sample divided by sample max 128 = 2 reflection rays).
So kind of balance of the shadows from the lights on the GI and not so worried about the reflection refraction.
This scene didn’t really need that many samples but because we had that many available to us. Redshift was going a lot higher than it needed to and using a lot more of those samples.
So I think this is a good I rule sort of set up here I’ll be happy sending this off to render.
Now You can clearly see that with this method you can get the noiseless and fast render output very easily!
How do I speed up redshift renders?
The advice I could give you is that:-
- Always try to keep some separation between the samples Max samples min, if you set these to be the same number you don’t really giving redshift any chance to be adaptive so you always want to have a nice spread between these two.
- As you rise Sample max up you can start rising Sample min up too, make sure there’s always a difference between these 2 values.
- I would always generally just trying to think of these in multiplies of this value in Unified sampling. The number of multiplies you put in sampling overrides is based on the strength of the effect.
I hope you clearly understand the concept of the Redshift Sampling Setup. The value I tried is not the actual number, it varies from different scene setups. So play with the redshift samples values and implement my method, I hope it definitely worked out!