Brecht's easter egg surprise: Modernizing shading and rendering

Yeah, I know. I’ll never trust half read CG-society articles anymore. Reliable information on arnold is really difficult to get…

[EDIT]

Thanks, those links are very helpful!

Here are two videos about Arnold

http://vimeo.com/15878348
http://vimeo.com/16155555

And here a nice article about the industry’s switching to physically based materials

http://www.cgchannel.com/2011/05/fmx-2011-tuesdays-highlights/

Hi ! Maybe it’s not the right topic, but I can’t render a 2500 x 1800 px size image on my GTX 460 1Go Ram ( win7 64, official blender 2.62). Can someone tell me if it’s a bug or something else please ?! Thx in advance…

That might be true, but I would think it still wouldn’t by any means reduce the necessity of doing major optimization work for a render engine to make it a lot faster then it might’ve been previously.
That’s not how it works. If it’s cheaper to use a “slower” renderer that is faster to use (e.g. artist time vs. render time), then that’s a reason to use the slower renderer. Of course most blender users aren’t really used to the economical considerations of it, and they are quick to demand one focus or another…

…there are many people on here who don’t have the luxury to spend thousands of dollars on a new rig or a new CPU whenever they need one. Heck a lot of people have very little disposable income for something like a new computer.
This argument could really be made ad infinitum (et absurdum). Right now there are children in Africa who don’t even have access to electricity, how are they going to make the animated feature film of their dreams?

I would think the target would be…
See, that’s the problem: You just think that Cycles should be this or that, but it isn’t. I’m sorry you’re an unsatisfied customer, but you’ll just have to live with the reality that is Cycles for now. If indeed the demand is so high for the renderer you would like to have, maybe someone else will come and write it.

You might be running out of GPU memory, especially if you’re using passes

The truth is though, the point I made applies to a lot of people on this forum, they already have adequate hardware, they just don’t have the ability to pay a lot more money to build a renderfarm or purchase a highend machine.

To me, these arguments sound like Brecht shouldn’t do any optimization work at all and just let moore’s law do that, one could possibly do it that way, but the only people who would really be able to use it for very complex work would be those with a lot of money and simply asking those who don’t to get a better job or give up the ambition to be a 3D artist.

There is plenty of reasons why it would be beneficial to see some major optimization done in Cycles which includes some of the rather obvious ways like finishing the incomplete QBVH implementation. I would also like to know of your opinion of companies and dev. teams that have spent a lot of effort in optimizing their engines like the ones behind Vray, Indigo, Luxrender, Thea, ect…

Also, I’m not unsatisfied in the way that you put me as, I’ve gotten some good results out of the Cycles engine so it’s not like I’m not able to use it. The CPU I currently have is a Q6600 quad core chip which will soon be among the low-end CPU’s, but my plan is to wait until I see some hardware that will be a good boost of speed similar to what I got when I last upgraded (which was 4-5x)

Can we get a bit of reality here? By ‘The Industry’ we are talking about large groups like ILM and Sony, who can afford to chuck a small warehouse full of Xeons or Opertons at their physically correct render engines.

Last I looked, most Blender users did not fall into this category.

I find it illuminating that the physically correct crowd tend to want a physically correct renderer because it is physically correct, not due to any real or perceived time saving or visual reasons.

As to ‘the waste of time factor’ with biased engines? Bill has a biased render engine, he has to spend a day arsing around with lights and settings to get his model to look right. Then he renders it in a weekend of dedicated crunching. Fred has an unbiased engine, he can set up his scene in a few hours. He hits the render button and waits 2 months for his render to finish. Who won?

Having read all the posts here I think this quote is quite interesting to consider:

Why Moore’s Law won’t save us
But technology will not solve all our woes. Asked if he thought that Moore’s Law – the tendency of hardware to double in processing power every two years – would eventually remove all time pressures from production, SPI’s Larry Gritz responded:

“It would be thrilling if all our frames could take six hours, but we’re not even close. I’ve asked at all the places [I deal with] and average render times are getting longer. If anyone thinks real-time [production rendering] is around the corner, they aren’t looking at the trends.”

Tim Alexander agreed, observing wryly: “If you can render everything in real time, maybe you aren’t putting enough on screen.”

Ultimately, it seems that artists are their own worst enemies, with modellers and texture painters taking advantage of modern hardware to cram beautiful but unnecessary detail into their work – with disastrous consequences for those further down the pipeline.

As Blue Sky Studios’ Andrew Beddini, the head of lighting on the last Ice Age movie, put it: “We’ve found that the thing to do is not to give artists the best machines, especially at the front end of the pipeline. Give a modeller a machine five or six years old, and [I guarantee] it’s going to have a positive impact when you get to the back end.”

Blender since recently became more and more useless compared to modern render system. BI was fast with basic stuff, and painfully slow when using raytracing and not even able to produce very realistic results when needed.
Cycles itself is pushing the render time dramatically up, but I save a lot of time because I do not need to make pre-render shots.
If I want to render faster I could simply get a faster PC. But a faster PC does not speed up my scene set-up. Cycles approach does.

Thus I high agree with the idea about more power and capabilities with software level out any advanced in CPU speed.
It is like with cars, the engines get better and the cars get heavier, thus gas milage remains the same.

I can’t speak for Tea Monster: But I would personally be bummed if I had to wait nearly 2 weeks for an image to converge only to find out I could’ve changed the settings to make it a bit faster or I did something wrong as in the following thread (and that is with the decision to sacrifice some crisp edges with noise-reduction)
http://www.luxrender.net/forum/viewtopic.php?f=15&t=7806

Now it is true that it is a difficult situation for a pathtracer, but there are cases where you might spend weeks tweaking the final look due to the fact that some areas require quite a bit of sampling to fully assess.

I also remember when I started with 3D when I was 14 my computer did not even have a FPU I was only able to do simple things. I think it is unreasonable to expect to be able to do with your home PC what companies do with render farms when you use their work as the measuring tool.

3D is just time intensive - specifically when the software get more and more powerful.

However with all this forward and backward I am not sure if it might be possible to include a render mode like BI into Cycles for fast rendering. AO and mirror reflections quickly bring BI down onto their knees so it is not that super fast as many state to be quite honest. VRay gives you the option like Thea to turn on/off render elements such as direct light GI etc.

This way you can select what is appropriate for your task and also time wise feasible.

I am not sure if the one bounce could somewhat function as plain direct light.

The idea that raytracing is slow I think with the code we have to day is proven wrong as it can be very fast.

Hatchi-matchi, I just wasted 10 minutes of my life reading through a big pile of nothing.

So, any cycles news today?

1 year and it lokos fake. not trying to bash it. but between the render and realism is the ARTIST. the artist make the realistic shaders and the artist paitns realistic textures . textures that are connected to the right shaders.

i dont care if he didnt want realism. rendering an animation that short and that simple for over one months is not normal. its not normal. Mango and in 12months other blender users will show with their animations what biased pathtracers WITH good artist can do. how fast and how real renders can be.

I’m not sure I fully understand this debate of yours but I don’t think anyone expects to be able to render out a frame from James Cameron’s AVATAR on their Dell PC using Cycles.

However is it only me or does it actually make sense to optimize your render engine as much as possible? :wink:

I think part of the problem is people think that is how it actually works, that you render a finished frame out of the 3D application. Avatar is a perfect example of a movie where the scenes are so complex, it’s like 50% faked with 2D in 3D space during compositing anyway, just to be able to manage the scenes at all. We’re talking 8k steroscopic, it doesn’t matter if you run dual xeons and quadro cards, your hardware will just grind to a halt. There will never be enough power to manage stüff like this - and when there is, we just come up with new techniques to eat our performance, hehe… ;D

And what do we get ? complexity and epicness so that our children would feel inspired and inspire some more :smiley:

so darn true …

Patch for Anistropic tangents – http://lists.blender.org/pipermail/bf-cycles/2012-March/000593.html

Yay for anisotropic shader! Very cool. Thanks Mike!

Are these the same tangents that were missing for normal map support? Or is that something else entirely?

Yah thats terrific.

Can this patch also support radial / circular brush patterns like those:
http://xsi.jankin.com/anisotropic_patterns/circular_pattern.jpg

http://xsi.jankin.com/anisotropic_patterns/

About GPUvsCPU - as far as I’m concerned so far some features are really hard to implement for GPU (that’s why they usually support unbiased calculations) and that’s why GPU support is a kind of a development bottleneck I guess.

Does someone have a nice answer for what the Tiles settings do in Cycles under the performance settings? With both CPU and GPU, I get 10-20% faster renders by setting tile size to 2048 and min size to 512. I was just wondering why.