GPU rendering?

Can FF use the GPU for rendering yet?
If so... how...?
If not... when...?
I'm in the market for a new card.
I'll likely go with a GTX275.
Would be really great if FF could utilize that power.

Thanks
So it looks like the answer is no...
Well that's too bad.

Check out Nvidia's CUDA Zone for some interesting stuff...

http://www.nvidia.com/object/cuda_home.html#

... kinda bumped a 2009 thread here heh.... although I can still add on a bit.

One of the main issues that I've personally observed with running stuff on the GPU is simply that not everybody has the same, uniform specs for it. Meaning depending on your set up, you can either have a smooth running, amazing program or a clunky annoying, and almost unusable program and there's a great disparity of reported performances. I think it's safe to say that most people probably don't have the best specs for GPUs more also on the end that they can't afford one so easily.

So we typically get dazzled by buzzwords and demo videos but the reality is the technology is still in its infancy and not everybody would have full-on access like the demo videos show.

Quote
Skybase One of the main issues that I've personally observed with running stuff on the GPU is simply that not everybody has the same, uniform specs for it.

Of course that this happens with the GPU, and it will obviusly depend on what graphic card you may have, BUT this happens exactly the same with the CPU and in the same way, not everybody have the same uniform render speed on their computer, and will depend on the CPU speed and perfomance they have.

I agree much with Betis too

Quote
Betis Yeah but... any program can run that way, CPU or not, depending on your setup.

Quote
Skybase Meaning depending on your set up, you can either have a smooth running, amazing program or a clunky annoying, and almost unusable program and there's a great disparity of reported performances

Well, I have to maybe disagree with this, because for GPU that is NOT going to be used for 3D Rendering or for really powerful hungry big and complex softwares, you will NOT need a very expensive and new graphic card, and even a very old graphic and mostly cheap (well at least from the last 5 years) will work well and right with a GPU optimized software.

I have a very old Nvidia 9600 GT graphic card that is very low on the Graphics cards benchmarks and really much slower than most of the modern and new graphics cards and all the GPU optimized software I have tried (not 3D rendering software) work very well, fluid and fast, and do not have any problem to run them in REALTIME and work very smooth and fast without any glitch.

For example I have Photoshop with GPU, a few of Photoshop Plugins GPU optimized and enabled hardware accelration, and also Adobe´s Pixel Bender that works very well with my slow GPU and makes everything in realtime, and there is also some other software that are also working with GPU acceleration and I do not have any problem.

Also have tried the DAZ 3D Studio software with GPU preview and works very well and smoothly, so from my experience and point of view you do not need to have a very high quality Graphics card for using GPU optimized software.

FF GPU SETTINGS AND PREVIEW

As it happens in DAZ 3D Studio, you can have the GPU optimized ONLY for the settings and preview and NOT have it for rendering that is done without any GPU rendering acceleration.

What I really would love and like would be that Filter Forge could use the GPU for the adjusting and modifing the settings of a filter in realtime with GPU acceleration, and so it would be much smooth, faster and better to really customize the result you may want, and then I do not care much if the final result render would not be GPU enabled.

Although thinking HOW FF works this would NOT be possible and only a full GPU acceleration would be possible as any preview is already done with the real render engine, so it is always like rendering the final result.

BUT a possible alternative could be to have only a GPU enabled acceleration only available for a 600x600 preview that could be done in realtime and that modifying each setting would be done much faster than is now that you have to wait until the result is shown until testing a new setting, and this again and again.

IF Filter Forge could have GPU acceleration, it would really be much more popular and would sell surely much much more.

Conclusion: Probably rewrite the program from scratch to incorporate it. That's the best.

Boy what are we going to do about it now? haha... guess more discussion? Please please threads? lol you name it we've extinguished it all. Your turn FilterForge lol! I don't mind waiting 3 years before you come up with something.

*yawn*

I've just returned from a short vacation and can't resist a little teasing:

Quote
wish that Vladimir or any of other FF team members could find a way to make FF faster

why don't you wish that filter authors create faster filters instead?

lol Spaceray, don't you have a pretty spoiled multi-core computer? Or was that a figment of my imagination? I thought I read in some other post about your computer specs.

And regarding filter speeds I think 60% of those filters can use some decent level of optimization. It takes some deep level of skill to really optimize something. If you know algorithms you win this game. But how many people are expert level programers? Hint: I'm not one and you obviously aren't. But I'm learning, isn't that better than not?

Also, at the point that you're making lists and threads (for the 5th or 6th time) and justifying yourself over this same issue, you should be 1. Doing art. 2. Making filters. 3. Improving your own methodology. 4. Doing things manually and finding faster ways of doing something. 5. Learn programming and make yourself some C++, dedicated photoshop plugins. 6. Or learn QuartzComposer, VVVV, MAX and other visual programming languages to make real-time dedicated filters. Is 60fps enough?

Should I compile you a "Book of Arguments" of all the posts and threads you've made regarding speed? I'm sure it'll be the best selling novel of the year. Probably after chapter 2 or 3 you start repeating yourself... oh right the same argument from 2010! Next page... oh the same thing.

Look, these forum replies I make to you are becoming a farce. I can write a comic with what we're saying. I swear I can, it'd be funny.

One of the reasons why I pick on you is because you're this "one big voice of all users" and the guy who parrots statements by others at a totally different level. You're this one dominating guy who just has everything to say, but your statements are basically the same thing, vague and sometimes unruly. And to me, and I'm sorry I have to say this so blatantly, you just aren't the expert if you were to make a debate that convinces people. I don't mean to tell you to stop saying stuff, by all means don't stop, I'm saying your mannerisms aren't helping the actual argument for something.

Look, I'm sorry I have to be so flat out blatant especially on a friendly forum like this. But I'm the type of guy who cares about stuff. I like things to be better and if something isn't happening, there's something wrong. We have to change the way we approach the issue!

Aye, I'm done. Now get back to doing art. Make some filters. Learn something fresh. Don't waste your life writing petty arguments. You have better things to do!

Yeah... methods in 1.0 were heavily restricted by what nodes we had, which forced us to do interesting tricks constructing elements of a filter. And a lot of the best 1.0 filters do follow optimal techniques that's usable in many other node-based compositing programs!

So in my view, regarding "speeding up" I always thought of it as "using what's already available in the best possible way."

Here's a little story:
If anybody's interested in 3D, maybe you heard of mental-ray. It's one of the top-notch render engines out there for photo-realistic imagery. The catch is, the render engine often gets called out being "slow" when it's not. If settings are tinkered properly it can run at a ridiculous speed. In fact, if you do even more prep work to the scene you're making, you can get a render done in 5 seconds with all the bounce lighting, shading, and anti-aliasing. Nowadays mental images does have iray which is their updated alternative render engine. It must have taken at least 10 years of in depth research and development before even coming to a point where they were able to say "hey we have something."

To me, having the sort of in-depth technical knowledge is pretty difficult. I personally think this is something people have to find for themselves and see for themselves as much as I can help along the way. I've written some tutorials and video tutorials but they can only go to certain levels. The rest is some careful study and observation.

Of course, don't get me wrong, I don't think FilterForge's speed development should end where it is. I just think everything in the world can improve one way or another. My thoughts people!