Difference between 300 and 2500 Samples?

Support Lighting and Rendering
1

I followed this tutorial to create a Teddy Bear:

I did 2 different saves. 1 was using 300 Render Samples, which took around 20 minutes on my PC. I then did another render using 2500 Render Sample which took a few hours.

However when I look at both renders on my PC, I cannot tell the difference at all in quality, how come?

I would have thought the 2500 Render Sample one would have looked a lot better in detail and quality than the one I did at 300 Render Sample?

Thanks for your replies :slight_smile:

2

It depends on the scene how many samples you need to get rid of the noise. If the one with 300 samples already has no remaining noise then you wont see any difference if you further increase the samples.

3

So how do I determine what is best to use for projects?

4

Per-scene basis. Some scenes are okay with only a few samples, some need more, and there’s a myriad of settings to reduce fireflies.
If you want a really nitpicky look at the difference between the two, put one over the over in Photoshop or Gimp and set the top layer to Subtract. It’ll be mostly black, but if you then run a white balance you’ll see every pixel that’s ever so subtly different. I’ve done that a lot in LuxRender to monitor progress because after a few days you can’t tell just by looking.

5

I find that for fine detail like hair, grass or fir, a bigger quality improvement can be made by rendering at a higher resolution than your final output. Try reducing the samples back to 30 and do a render at 400%.

6

300 samples should be enough to reproduce even very fine details. Beyond that, more samples are only going to further reduce variance (noise), which may not be apparent in your scene.

Doing that is the equivalent of rendering 30*(4*4)=480 samples, to some degree. The difference is in how your image reduction algorithm averages the samples, compared to the Cycles reconstruction filter (which you can tweak for a smoother/sharper image, btw). If you do this with a tone-mapped, low-dynamic range image (e.g. a PNG), it has the benefit of lowering the impact of bright outliers (fireflies). Also, it helps with certain aliasing artifacts you might get during compositing using e.g. ID or depth masking.