Did Cycles overcome Blender Internal?

No, you just make the red cube a tiny bit emissive and turn on approximate indirect lighting. There’s your color bleed effect, and it renders in a fraction of the time it would take to render with Cycles.

How is Cycles without GI looking worse than BI? It’s identical.
What are “Terminator issues”?

Here’s Suzanne rendered in BI, with one point lamp above and a mesh plane below. Plain direct lighting, no fancy stuff.


And here is the exact same thing rendered in Cycles:


These, my friend, are terminator issues. And look at the time it took to render this crap.

Remember the cave scene I posted earlier? Here it is again, this time without any GI. Pay attention to the shadow boundaries. Do you see the polygons?


Again, note the render time. Plain direct lighting with no bounces actually takes longer to render than the Half-Lambert fake GI version I showed you earlier. And this is just in Cycles. If you could see how fast it renders in BI, your jaw would drop to the floor.

The idea that Blender could do without BI any time soon is completely delusional. Frankly, I don’t have the slightest idea why threads like this keep popping up again and again. Even if they totally freeze the code and refuse to add any new features submitted by third parties, BI will remain indispensable for a long, long time.

But I am arguing for a scenario where BI and Cycles are as fast for the same result.

There’ll be no such scenario. It ain’t gonna happen, period.

I could as well argue for a scenario where BI has raytraced indirect lighting, because last year someone submitted a patch for that. But the patch has been mothballed, like so many others. Rather than asking for a completely new renderer, people should take a look at what’s lying around in the patch tracker for BI. That would also bust the myth that no one wants to work on BI any more.

eye208, +1

I think this thread has run its course…

Thanks eye208, that is a very good reference for later

I think the main reason against BI is not that the technology or idea is bad. Of course not. Cycles is great but not the best fit for everything.

But fact is that the BI code is a mess and thus a disaster to maintain as far as I know.
Before Cycles I mainly used Yafaray which has terrific shader and render settings for both GI and non-GI based images.

But in todays time I really would like BI work like a normal engine but use Cycles material node approach. The BI node concept of mixing materials was again full of hacks and bad workflows.

its almost impossible to find a bi tutorial these days, everything is in cycles. I find bi necessary so for me all this focus on cycles gives the impression that the days for bi are numbered.

Oh yes! + 1 to this one.

Turn Blender Internal into Nodesville. Christ on a crutch!

Well,now that cycles development is delayed a bit so, Bi isn’t dead yet.

Yes, BI is old, and a mess IMHO, it had its glory… HOWEVER saying that, what i REALLY miss is what the Renderman people did, INTEGRATING the “old” scanline algorithims into the new modern PRman 18. I think it would have been possible to get cycles to render to scanline or even REYES as an option for animation while keeping Cycles clean.

However, being said that (again) we should be REALLY happy to have Cycles the way it is… especially looking forward to more and more powerful CPUs/GPUs… but i see the point, a scanline option would have been great!

Many love scanline, like myself, 90s old school (enilnacs<>scanline ;-)), maybe future development would make a scanline/integration possible.
Imagine rendering some Geometry in scanline CPU and other layers in Raytracing GPU, AT THE SAME time, by pressing the render button only. Per object selectable :wink: Initiating 2 render engines at the same time is not a problem nowadays with at least 4 cores on any actual CPU.

Also i don’t believe that we are at the height of a scanline<>raytracing combination in the actual raytracing hype. Maybe after some Siggraphs years, some will see how much waste of cpu/gpu power is wasted on raytracing only… maybe then somebody will try a hybrid approach that is completely new and doesn’t use the old concepts of integrating raytracing into scanline, but rather complement each other in one go. There are still much ideas out there also with other renderers like adding voxels, etc… :wink:

would be nice and tidy up the UI issues/inconsitencies…
doubt you’d find a willing dev.

mostly end up using material nodes in BI anyway to get good results… and you quickly start hitting performance issues when mixing matreials that are outweighing advantages…

I gave YafaRay a try the past two weeks, because I need visible light rays coming through windows in an animation. Cycles volumetrics is too slow for the task right now and BI can only do this with approximated shadows and the results are not really convincing.
I was surprised, how well YafaRay implementation in blender is designed, how easily you can switch between direct light and different levels of pathtracing/raytracing and how good render times are, so now I use it for my interior scenes.

What I’m saying is that when you’re trying to achieve a certain result as an artist, you should look what technology can do the job best and not which one has the cleanest underlying code or the most state of the art technology.

I can understand that developers don’t like to work on legacy code and drop it at a certain time, but not why artists look down on other artists for not using the same tools as they are.

Saying things like “noone uses BI anymore” on BlenderArtists without having the numbers to back it is just plain arrogant.

I agree with Eye208 and hope BI wont be sacrificed any time soon, but some development time would be spent to add some of the useful patches.

I miss the freestyle system in cycles:no:

It’s all very well to say that cycles is better programmed and neatly coded, but if you remove BI from Blender you will immediately cripple it. People have gone way overboard with the soft shadows thing - sometimes all you want is a clean, hardedged, smooth surfaced illustration and you want it at 15 sec a frame. I use BI all the time.

The biggest stumbling block is the lamps not translating. If every lamp were to have separate BI and cycles settings - that would make things as integrated as one might need.

Personally, I have no problem with seeing patches like the half-lambert shading making their way into BI because they don’t actually use any sort of raytracing to produce their effects.

I think continuing development of BI is fine as long it’s not in a direction where it tries to become a full-fledged pathtracer (because the raytracing part isn’t even remotely optimal for that). The monkey comparison isn’t really a good one to use either because the BI image is using pure scanline rendering and Cycles is using raytracing (the former having a natural bent toward lower polygon counts).

If you want a serious comparison, render out a complex glass object with internal reflections and absorption and watch BI choke while Cycles zips ahead (remembering to set the diffuse bounces to 0 and checking ‘no caustics’ to simulate feature-parity). You will then see why it’s logical to have future BI development concentrate more on becoming a truly comprehensive scanline based solution (perhaps even having real GI by way of a new radiosity implementation that doesn’t suck).

Rendering isn’t about replicating real world. Anything that comes out of Cycles (what’s been released so far) looks like it’s either a realistic rendering or it’s trying hard to be.

Not to mention that Freestyles isn’t going to be working with Cycles.

Not to mention Cycles is a mess to setup for basic scenes.

I gave YafaRay a try the past two weeks, because I need visible light rays coming through windows in an animation. Cycles volumetrics is too slow for the task right now and BI can only do this with approximated shadows and the results are not really convincing. I was surprised, how well YafaRay implementation in blender is designed, how easily you can switch between direct light and different levels of pathtracing/raytracing and how good render times are, so now I use it for my interior scenes.

Hey 2ndClemens glad you discovered YafaRay for animating. I discovered it much like you did with a completely enclosed scene to be animated on a *mid tier machine. And, I find it fast and intuitive to set up. I’m using Direct Lighting and find it handles reflective material way faster then BI. There is one thing it can’t do right now that BI can and that is import and play a video clip. As in a TV screen playing. If you happen across a work around in the future please post it. But, the YafaRay developers are of course aware of that and who knows. *($700 in 2012)

Hi theoldgost, I didn’t know image sequences and movies are not supported, but I just checked and you are right. You could probably convert the movie to an image sequence and write a simple python script that generates a plane with a new material and texture for each image and sets renderability and visibility to true for only one frame each. If you do this in an extra blend file, and link a group with all the images that might be a workaround that is not too messy…

I find it unfortunate that out of the box, Cycles can’t render a decent scene without all the noise, fireflies etc. I don’t like to have to jump through hoops, set up complicated node setups and adjust multiple settings, all which are bandaid solutions, just to get a somewhat decent render. I wish all the time spent on Cycles had been put into BI and then we’d have a good renderer that works.

Ever try to render an animation in cycles? Impossible without a super computer.

a little over dramatic… you could just let it ‘cook’ longer…

doing lots of animation in cycles… takes about 5-20 times longer a frame than when I’m using BI… but that’s the tradeoff for the relatively easy setup and iteration time… many more workarounds in BI to get good looking results, you really pay upfront for reduced render times in teh end…

cycles baking is of course an option for a lot of cases… bake once then render in BI… or even in cycles as emmisive but that is a bandaid…

I wish people would read history before jumping onto conclusions. The render25 branch during Sintel, tried to extend BI with advanced features and GI, but was not pursued afterwards because the code base was simply too difficult to extend. :wink:

@motorsep
Freestyle will work with Cycles soon: http://lists.blender.org/pipermail/bf-blender-npr/2014-June/000041.html