What is important and priority for FF render speed ? NOT HARDWARE !

Hello,

I have been testing many different photos and I have seen that the render speed is really very different depending on what image you choose (resolution and complexity), the filter you choose, probably the preview you use ...

Some images and filters combinations takes only a few seconds, and others take many minutes, and some even reach for 1 hour or more.

As i do not know how Filter Forge works I want to make some questions:

Quote
1 - What are the things that are important, priority and relevant for the render speed of Filter Forge ?

2 - FILTER FORGE FILTERS

I have seen that the many different filters have many different render speed depending on what the filter does, and if it depends on a source image or renders the image generated only by FF code without any source image. Photo filters are much faster than Creativity filters, so I want to ask

Is there a way to know really which filters are faster and which are slower ?

3 - STANDALONE VS PHOTOSHOP PLUGIN

Does Filter Forge have exactly the same speed if used as a standalone without any other software running than used as a photoshop plugin inside Photoshop ?

Is faster if run standalone vs plugin ?

I mean that as FF is a very resource hungry application, is it faster if used as standalone vs used as photoshop plugin ?

As standalone, it does not need to share the computer resources with any other software, and as a plugin it must share the resources with another very resource hungry application, photoshop.

Perhaps this depends also if you have much more RAM for both, I mean, if you have just 2 GB or 4 GB for both, or if you have 8GB, 12GB or even 16GB RAM, so as FF can´t use more than 1,5 GB RAM, but Photoshop can use the other ram that FF is using as is a native 64 bit and is programmed to get the most our of RAM.

4 - IMAGE COMPLEXITY

How important is the source image or photo you choose ?


I mean in how simple or complex is the photo, is slower to render if it has lots of complex colors, and faster if the photo is simple with few colors ? Or is the same for FF ?

5 - IMAGE SIZE

Of course, the resolution of the original source image will be a very important thing, because higher and bigger resolution, higher render speed time. BUT I wonder if:

Is the resolution proportional to the render speed or is it exponential ?

I mean if I have a 1000 x 1000 photo and then resize it to 2000 x 2000, will the render speed be just proportionally or be exponential?. I have tested that when you go to higher resolutions the speed is more non proportional and is more exponential as renders are higher resolutions.

Is the speed of a 1000 x 1000 and a 4000 x 4000 exactly same photo, just proportional the render speed ? ( the 1000 is a reduction from the 4000 photo)

6 - IMAGE PREVIEW

Does it Filter Forge render at the same speed if using reduced preview than actual preview size ?

Because it seems for me than when using reduced preview after selecting "save image as" is faster than saving the same image with the actual size preview activated.

Perhaps this is because it must make double render for the actual screen preview and for saving the image. Or perhaps this is an ilusion I have and is not true, as I do not know how FF works.

7 - ANY OTHER IMPORTANT AND RELEVANT THINGS ?

Probably there could be other important things that I have missed here and are revelant for the Filter Forge render speed. So, please, if you know anymore things, just put them here.

Thanks very much for your help

NOTE for Hardware as far as I know the most important is the CPU speed and as many cores as you have and even virtual cores as intel HT technology, as RAM, graphic card and hard drive speed is not important for FF.

GMM Thanks really very much for all the answers

It is very interesting to know the answers.

2 - FF Filters Speed

About number 2 about filter speed, this list is very good and very interesting and useful and great to know as I did not know about it, thanks very much, BUT for this to be used you must enter inside the filter editor and see HOW the filter is built and see all the components used, and do not how the combination of ALL the components toghether are affecting the final real speed. Yes, perhaps if all belongs to the Fast categories it will be fast anyway.

Now I will ask again the same questions in a different way

Please, Is there a way to know really which filters are faster and which are slower without having to open the filter in the filter editor and look for the components that have been used ?

3 - Standalone VS Plugin

Is it ONLY the transferring data to and from the host that will make the plugin a 3-5% slower ?

I mean, that if the FF standalone would be run in a fresh computer start without any other software running it would have ALL the computer resources for himself and so it will not need to share the CPU processing with any other software running at the same time.

If FF is run as a plugin it will need to share the resources, and it have less CPU processing or will need to share more things, and I thought that this would be a thing that could make it somewhat a little more slower, but if you say that the only problem is the transfering data and FF does not care if the resources and CPU Power is shared or not, is interesting to know.

4 - Image complexity

Interesting to know that is the geometry and not the color amount what is really important, So looking at your answer FF does not care in terms of speed if it has only 4 colors, 16 colors, 256 colors, 4096 or 16 million colors and how this colors are populated and distributed in the image. I will put later an photo example to make it more clear, as is said an image is better than thousands words.

5 - Exponential speed increase

Perhaps you have not been able to reproduce this, but I have been able to do it with 14 different test renders with Vladimir Creepy filter and also with a few other filters.

I know that a 2000x2000 images is NOT the double of 1000x1000 because if make the math and multiply 1000x1000 = 1.000.000 and 2000x2000 = 4.000.000 and so FF has to process 4 times more pixels, and in this is very little difference in time, but what happens when you try to render a 4000x4000 = 16.000.000, is it really 16 times slower than the 1000x1000, i can say that the answer is no, at least in the test I have made this is not true, although I have not made extensive tests.

Is the resolution source size proportional to the render speed ? Or asking the same question in another way

Will a 4000 x 4000 source photo be 16 times slower than the same photo reduced to 1000 x 1000 pixels using with both the same filter ?

And what would happen with a 5000x5000, will it be 25 times slower ? And with a 6000 x 6000 should be 36 times slower I do not think so looking at the test results I had as when rising the resolution the speed difference is more slower and not proportional.

Quote
SpaceRay wrote: Is the resolution source size proportional to the render speed ? Or asking the same question in another way Will a 4000 x 4000 source photo be 16 times slower than the same photo reduced to 1000 x 1000 pixels using with both the same filter ?

It should be. A 4000x4000 image should render ~16.1 times longer than the same 1000x1000 image.

If you're willing to test this I would appreciate if you do the test in the command-line renderer. First, it eliminates all possible preview-related and other GUI delays; second, it's much easier to describe and reproduce the result. Also, I don't recommend extensive testing of huge resolutions: if you find an issue with a 60000x60000 render I don't think our testers will be happy to spend hours trying to reproduce it

Quote
GMM wrote: If you find an issue with a 60000x60000 render

Quote
SpaceRay wrote: The 6000x6000 resolution does NOT take ALWAYS hours to render on ALL the filter

I really meant 60k x 60k, not 6000x6000

Imagine rendering a 60k x 60k filter with max antialiasing all pixels
I have the perfect filter in mind

I was thinking about the ink filter that you had a thread about making faster
You said it gets very slow if you go above 600x600, so imagine the time it would take if you make it 360 Megapixels

Oh, and there are many uses for a 60k x 60k image. You could cover the moon with it, for example

You know, the actual print quality of huge billboards, or those advertisements on builtings are only 30 dpi. It's just that you see them from far away, so your eyes can't see the low resolution and the image seems fine.
Imagine how far the moon is... Probably a print res of 1 dpi, or even less will be enough.

Here's an example of what I mean:
If you step back from the computer you'll start seeing the image better. The farther away you will go, the less pixely the image would seem.

Quote
Morgantao Also take a look at This website.

I am very sorry that it has taken nearly one year to answer (6 March 2012 - 1 March 2013)

I like the "small" logo that can be seen from space

You are right that it could be done in FF and printed really at very much less resolution

Please, before reading this post, read first the above one I have put above this post about the positive side of this topic of FF render speed, and now is the negative part of it shown in this other thread

List of threads about the FF render speed problem

OH!

Sorry I confused the RAM data is not 1.5 MB , the correct one is 1.5 GB

Thanks Skybase, good that you did see it, of course it was a big mistake

Please if possible for the moderator to change the MB to GB in the above post of mine would be appreciated to avoid confusion for FF users