[Ernesto]

More Speed.
I know this is not a new wish...
Maxwell is agreat renderer, but I could use it only in simple lighting conditions, due to the long rendering times for other cases.

I have just finished a night image of a pedestrian street with light coming from all the shop windows as well as from back light adds and urban lightfixtures. There are some lights coming from cars crossign the pedestrian way. The resolution was 1500 x 1000 and took 6 days, but heavy noise is still there. Balancing the lights was very difficult because it is not possible to use multilights feature due to the number of lightsources that cannot be joined into several channels. So I spent several days trying to balance light levels.
Of course this image was not for a client who would complain for the schedules.


If we could reduce rendering time, it will be great.

Ernesto

[Bubbaloo]

Compositing, scene optimization, using a render farm, these are all great options for scenes that are very large and complex. But, of course I want the core engine further optimized also! Who wouldn't?

[Ernesto]

What looks not practical, for the architectural use, is that it reaches the time limit with simple jobs, under night lights conditions, which is not an exceptional thing in the Architectural market..

Just as a sample of what I mean, you can look at most rendered jobs in this forum, using maxwell, and you will find the following:

1) lots of small objects, or small spaces under all lighting conditions.
2) some medium spaces under not so complex lighting.
3) buildings or urban spaces under very simple lighting conditions.

Analizing this, It is obvious that Maxwell is being a limit for architectural use.

The bad new is that Maxwell works ok, in situation where other renderers do it great at a faster speed, but is situation where Maxwell shows the quality diference, it is being limited by the extremely long rendering times.

I found that a simple raytracer is a great tool, when you show a huge building with lots of windows under daylight, because you do not need detail in the shadows as they are too little.
An image made with maxwell of such a subject, will not be recognized as a Maxwell image, because the diference which is in the shadows is almost undetectable.

The great diference would be noticeable in a night scene with lots of lights, as it is usual to see in a pedestrian commercial street. But unfortunately this takes forever!

If I were a jeweler, the rendering times of maxwell would be perfect for me!

Ernesto

[subspark]

You know what?
10 years from this day Maxwell could very well be a real-time application. If you look into quantum computing and its expanding field of research, computers of this kind will solve math like rendering complex super-resolution images on the fly.
If by some breakthrough we soon find a way to keep the quantum information stable and permanent, then quantum computers will become widespread and you can simply tell your machine to solve equations for rendering graphics rather than solving particle simulations which to be honest requires far less resources.

It's all math. And quantum computers love math. While I hate to take your brains off the present, you best keep this food for thought because capacity like this is inevitably going to occur.

Anyway back to the present, I am inclined to agree with Ernesto. Maxwell needs a few speed tweaks.
Paul.

[JTB]

For what I understand from Ernesto's post, Maxwell is not the right renderer for architecture...
Well... if you think of the deadlines and how low paid job it is, he's probably right up to a point...
I use Maxwell for exteriors all the time, I am very happy but I must admit that the quality of my work is not the best around here, so probably if I spend the time you guys do, I will probably have the same problems.

I use a quad core and I am happy to see a nice looking picture at 2000X1500 after some hours... No night shots unless 2-3 projects and it really was difficult.

PS. You will find more than 5 posts where I wish for more speed... This is still my top wish but I must say that Maxwell is working well for me now.
Since we are waiting for displacement and other nice features, speed will be more important.

[Matthias Wirth]

Hi NL Team

the Maxwellrenderstudio is very good tool

i wish me for the future more renderspeed

i found this is the important point in the 3D production

the problem is the rendertime when my clients make changes on the

renderscene and then i have big problems

Please give Maxwell more Speed

greets matthias

[eezydo]

Speed is allways usefull. Thats banal.
But these recurring speedrequests are a bit " have my cake and eat it too" -like.
There are plenty of fast engines around if thats what you need.
We use Cinema for most of our daily work and it does very nice Pic's in reasonable time
but every time I do a Maxwell overnight rendering and come back in the morning it drives a smile of pleasure on my face.

Yes I'm sure There are things it cannot be used for yet. - So what!
It is the currently most realistic renderer and I hope Next Limit never
compromises on that for the sake of speed.

Before Maxwell there was nothing to complain about. - Now that a new
level of quality has been made possible, people miss the speed.
Thats Human nature.

We could also look at it this way : Nothing has changed. The Speed is still there. Just like yesterday. But for some cases we have gained incredible quality.

I hope Next Limit keeps the priority on realism above all.

[Bubbaloo]

The speed is coming with faster processors...

[Maximus3D]

No you can't put all faith in hardware because not everyone can constantly upgrade to the latest and greatest most expensive hardware, the engine itself needs to get constantly optimized for speed and with quality maintained. Although i'm sure they are performing these constant optimizations to the code with each release i think there's more that can be done. You only have to look at what the competition managed to squeeze out, they have pushed it a bit further when we're talking about speed yet they have in some ways kept the quality they wanted and also improved further on it.

I'm not saying that their way is the way to go, but it's one way and it shows it can be done. Remember that there are many roads that leads to Rome, Next Limit just have to find their road to Rome and follow it.

/ Max

[Bubbaloo]

I agree, it may be possible to optimize the algorithms, but it is and will always be a processor intensive process of heavy calculations to produce photoreal results.
Also, if you are making money from the renderings that you produce with Maxwell, it is a business expense to keep your hardware to the level that you need.

[eezydo]

only have to look at what the competition managed to squeeze out,
/ Max

Thats the point - If quality and speed is there, be happy and use it.
There's much more to "squeeze" with a biased approach,because it's all about tricking things into place.
I love Cinema for this. It lets you do all those biased artistic things.
But I also love that other approach where I ask myself how something
occurs in the real world and then know what I have to do.
I just don't think the latter kann be "squeezed" in the same way.

[JTB]

I want to be able to choose and I will never understand why NL doesn't want to make a biased Maxwell... Imagine the simplicity of Maxwell and the power of Maxwell mats and the rendering times of ....(You choose).
NL always declares that they don't want to make a biased version of the software.
On the other hand, I think that if it was possible to have a quality near Maxwell's using biased algorithms, then it would have been already accomplished.

Anyway, as I've said before, Once a Maxwell user, Always a Maxwell user.

[def4d]

A lot of 5 bounces renders made the Maxwell success!!
For the speed we all dream of, i think we have to buy a farm, or wait that quantic computer
There was a thread asking about GPU using, but the answer was no i think

[Fernando Tella]

They could add something like fry's aeronoise; wait! they prefer to develop something new instead of copying things (I like that way of thinking), ok: make it automatic: Mxcl detects where there's more noise and focus the calculation there. ---> Huge improves in rendering time without biasing the render.

[-Adrian]

Mxcl detects where there's more noise and focus the calculation there. ---> Huge improves in rendering time without biasing the render.

Man, coming to think of it, that would be amazing!