appleseed Users Forum

Appleseed vs Cycles


#1

How powerfull is Appleseed compared to Cycles?


#2

Without defining what powerful means for you it is difficult to give an answer.
I can state you some differences between Cycles and appleseed. This is not a full comparison, just some key points.

Cycles is developed by paid developers and lots of other contributors, our developer base is smaller. Cycles is more feature rich (light portals, IES lights, hair shader, curve system etc…) and has matured to a viable alternative to commercial engines.

appleseed is mainly geared towards animation/vfx but also values exact light transport and precision for industrial use cases.

  • The biggest difference is that appleseed is a cpu only renderer (at this point) while Cycles can use cpu, gpu or both together. So Cycles will render faster on average scenes, though not all.
  • appleseed has a fully OSL based material system, allowing OSL based shader execution without the speed penalty found in Cycles. Though we miss still a hair shader and heterogeneous volumes (both are in the works)
  • appleseed is a spectral render engine which can operate in spectral or RGB mode
  • appleseed can render caustics efficiently with Stochastic Progressive Photon Mapping
  • appleseed has an advanced adaptive sampler (Cycles will get one in the future)
  • Motion blur and texture caching is very efficiently implemented in appleseed

From a Blender user point Cycles has obviously the best integration into Blender. External render engine developers don’t have the same comprehensive access to Blender’s internal functions as the Cycles devs have.
appleseed cannot be optimized for only one DCC platform, we need to support also 3ds Max, Maya and Gaffer in equal fashion. While there exists ports of Cycles to Max, C4D and Gaffer, their usability and features cannot compare to the one integrated in Blender.


#3

By powerful i mean which one have best production quality render engines.thankyou for the reply.


#4

Production quality is also something not very well defined without further information.

Here is a comparison of a commercial product viz project with appleseed and Cycles from the artists point of view.
https://forum.appleseedhq.net/t/battle-against-ycles/667


#5

If you mean production quality in terms of feature completeness then Cycles wins hands down.

If you are referring to image quality that is fairly subjective. Both Cycles and appleseed have exceptional image quality. My own opinion is that color reproduction appears more saturated and visually appealing in appleseed, but that’s hardly scientific.

Speed wise, Cycles is faster to render in almost all situations, at least until deformation blur is used. While Cycles’ Embree integration has helped with that, appleseed still seems quicker.


#6

Imo, Appleseed is more user friendly on OSL support in Blender.
Up to now, Cycles script shader node ignores metadata.


#7

Just today, I played with alpha maps.
A test scene reveals, there is a massive performance difference in Cycles and Appleseed if I use alpha maps.
Comparing with Cycles, Appleseed renders test scene nearly 9 times (!) faster.

Cycles_Test Appleseed_Test

Test scene uploaded on google drive.
https://drive.google.com/open?id=133Ct6RUPLXvxbmCInZWpSYPM19QziFQj


#8

Totally agree with Jonathan_Dent about the subjective nature of image quality. I like to add to that with the ever precious thing called time. Getting that subject look in a set amount of time is often key for production jobs on a deadline.


#9

Interesting


#10

It is difficult to measure speed. Cycles needs more Samples, but Samples are over all faster in Cycles.
I try to estimate how many samples are needed to get similar image noise levels.

Compare
600 samples, Cycles

150 (50*3) Samples, Appleseed.

It looks like Appleseed needs less Samples to get a good noise level.

Speed difference in Cycles, 600 samples

transparency | no transparency
–––
25s | 8s

Same test with Appleseed, 150 samples (not sure … its still not as noisy as the other render)

transparency | no transparency
–––
18s | 17s

Maybe I do a misstake …
I try to adjust noise levels by sample count. Cycles loses performance if alpha maps are used.


#11

Thanks for the detailed testing Leonard!

indeed, interesting that transparency has such a speed impact in Cycles. Recently Alpha inputs were added to Cycles Principled BSDF which may lead to better speed when using transparency. https://developer.blender.org/D4848


#12

Hi Herbert, thank you for your note. I did not notice Brechts commit. Need to wait, but it will be interesting to see if something changed.


#13

Now its in the daily build. Alpha input for PrincipledBSDF probably does not make that much difference.

Blender 2.80, Cycles with 600 samples, cpu only
transparency: 26.48s
no transparency: 11.32s

With GPU on.

cpu+gpu (GeForce GTX 770)
transparency: 21.90 s
no transparency: 6.77s


#14

appleseed has very fast alpha cutout rendering thanks to Fetch. I managed to find a presentation Franz gave back in 2015 where he mentioned that was one of the major development goals for that project.

Edit: You should have seen Cycles back when it first came out. It’s alpha map rendering was HORRIBLY slow then.


#15

Mh … I did not notice. That time I probably had an old celeron box :snail:
Anyone lose hair, if he dares to render something on this machine.

Alpha maps help to keep poly count low. I rely on Cycles to put particle trees on terrain.