Arion: a Fryrender RT engine + GPU support. Coming soon

[percydaman]

Oh, yes, but do you remember?

First: Radiosity, next Photon map, next Irradiance map, next LC, next PPT and now RT engine. And every time, more CPU or GPU power, more photo realistic image and less scene setup time.
Do you remember the good old procedure for the photon mapping (thanks Goncalo ehh) . Each light will be setup. In the future, for arch-viz, catalog etc... the unbiased, with very low setting procedure + good VGA or WS... it's the future.
So, cheat or not cheat, time is money. And, if we have more power in the HW, we can lower the time with this renderer.

And Chaos, with VRay-RT, VRay 1.5 (or 2.0) with GPU support + 32CPU core etc... + PPT follows this road.


PS: I think Photon mapping GI in VRay could be deleted ehh...

occlusion maps, normal maps. etc etc. Newish technologies that are cheats and fakes. I guess my point is, regardless of how my render is done, unbiased or not Ill still be using methods that enhance it.

[3DMK]

Look at the world of high end 'real - life' movies to see where its all at.
Show me one movie crew that doesn't use 'Fill' lights to artificially brighten dark areas. This is not physically correct but it's all about getting the results you want/need at the end of the day.

Cheers

[simmsimaging]

Always the same trade-off with 'unbiased' vs 'biased'. It really comes down to simple/easy to setup versus complex setup with commensurate flexibility. It really comes down to what you need and how you like to spend your time. Anyway - Vlado makes an excellent point about the cleverness required for the 'tricks approach: Hats off to you and Chaos for being top of the game.

b

[Alan Iglesias]

It has always amused me to see so much grandstanding regarding making your renderings so "photo-real", and it's still going on. Most of the time, the last thing I would want is to have everything in a rendering system behave as a camera would! The millions of hours spent over the years retouching photographs and trying to get cameras and film to do more than they can physically do is proof enough of that. And since most 3D scenes do not contain everything the real world contains, there will always resaons for having ways to simulate what isn't there!

3D rendering systems are the "cameras" of the mind, and should be just as unlimited as the imagination. It will always be good to have the fundamental classic abilities of traditional imaging systems at your fingertips. But when an artist needs to create the imagery he "sees" in his head he needs any and all the functional tools that can be designed and used to realize the vision. This includes all the excellent "fakes", or whatever you want to call them. I would call them tools and the more you have, the better and faster you can create what is in your mind's eye, whether it's for yourself, or a impatient client.

-Alan

[yyk]

this is a very interesting point, never thought about it that way. everything unbiased it's sold telling you the exact opposite: it's all groundbreakingly revolutionary, technologically speaking. I guess one tends to believe the hype, after a while.

To be fair once again, Chema has pointed this out on the Fryrender forum.. I don't remember when but something like few years ago he said it's easier to write an unbiased renderer.

Still I prefer Vray for all it can do in less time Most of the time the easy to set up photoreal quality equals a bad photograph.. but who cares because it's unbiased, right? go tell that to the client :P

[pixela]

I think "cheat and fakes" are essentials and what makes cg.
Because it is impossible to create a physically correct worlld in our viewports.
For example, if I am making a house, I can try to create the environment around it as good as I can to match real-world but is it really possible ? NO ! In order to bounce light correctly as possible, if the house on a street, then I must model and texture all the other houses on the street, trees, sidewalks, all elements no matter if they are visible to camera or not. We all know that when there is a white painted building near to your house, the light coming through your window is different than a grey painted house. For example, even white-washed buildings have much different behavior than the white painted buildings as the light kind of diffuses inside white-washed layer and looks like a SSS effect. I can make this list longer for exteriors, interiors, characters etc.
So in short, it is impossible to create and model the physical world 100% into our scene because we are limited with our hardware and moreover it will be a great time-loss.
So what makes a good and realistic cg work is fakes and cheats.
Closest we can get to physically correct is studio setups. Because in real life, in studio setups there is a really fake and controlled environment So with cg, it is possible to mimic this 100% accurate. But it is always impossible to mimic all the environement accurately.

I believe fakes and cheats are essential in cg and they can be seen as mastering skills.
All very photo-real works mimics the ral world to a smart extent. And I believe mastering the technique is to be able to decide what to include and what to exclude from our scenes.

[rivoli]

To be fair once again, Chema has pointed this out on the Fryrender forum.. I don't remember when but something like few years ago he said it's easier to write an unbiased renderer.

didn't mean to sound unfair to him, actually I didn't have fryrende in mind wrinting that.

[cecofuli]

Released

http://www.randomcontrol.com/purchase

Now I'm waiting our GPU VRay-RT

[yyk]

I don't like how they restrict access to Arion forum by labeling it "customers only". As always with new stuff there are bound to be problems with this technology and if I can't even access the software specific forum to check out what users say about this product, forget it.

Released

http://www.randomcontrol.com/purchase

Now I'm waiting our GPU VRay-RT

Second that, would love to hear some news on RT. Haven't had a chance to use it yet, waiting for the CUDA-stuff.. :P

[beestee]

There has been some news on the OpenCL front this past week:

http://www.opengl.org/news/khronos-u...ation-with-op/

I for one would rather not be restricted to Nvidia cards, which CUDA requires.

[vlado]

If you take a closer look at the statistics for the GPU/CPU performance in the Arion movies you'll notice that the GPU contributes more or less as much as all CPU cores together. Which means that you can get the same performance by simply using two machines to render, or one dual-processor machine. Of course, GPUs will get faster... but so will CPUs... the race is on

The nice thing about V-Ray RT is that it already supports DR, so even now you can get pretty good performance. From that point of view, there is no need to "wait" for a GPU version like it will solve all problems... on the outside, it's exactly the same, but somewhat faster.

Best regards,
Vlado

[^Lele^]

I think there are at least two very different perspectives at play here, along with the usual shades.
Mentioning the two extremes, you have a product/arc visualizer with little to no computer knowledge on one side, and on the other end you have people which do RnD for estabilished VFX/Photoreal CG studios.
The two workflows will look EXTREMELY different, involve a very different number of people, and a very different amount of specific knowledge.
The first will look at his render pretty much as a finished piece, with little to no retouching after the fact, the other will recreate the render in post from a myriad of different channels, with a post professional (and BOY do i know a few of these...) taking care of the nifty stuff to polish it up.

From my perspective, and little experience, complete flexibility is very much preferred in a place where there are many an individual taking care of a single shot.
Whether the cost to achieve it is a bigger renderfarm, or one equipped with Quadros, it's little interesting.
All you want is WAYS to treat and change your render after the fact.

In the first case, you do not "eyeball" it, you achieve, or fail to achieve the render.
Then you need VFB post controls, because you can't use a post software afterwards, with good enough results, and a gazillion embedded shortcuts to change object materials' colors and reflections through material IDs and integrated reflection passes.

Two very different philosophies.

Both would benefit of advancements in speed and quality, but the one depends on those, where the other is relying on its own ability to cook up a solution to change the final result to his liking.

Bring on the debate

[bardo]

a little contribution for the "next gen" renderers.

http://www.luxrender.net/wiki/index...._OpenCL#Status

For the luxrender the GPU contribution is good.... but I think that it's not a rule for every software.

Anyway in that webpage the final words are " CPU Vs GPU ? No, thanks, CPU + GPU + Network rendering is better" ....
So, again, the flexibility also for the hardware support could be a feature to stay one foot ahead.

[Morbid Angel]

whatever you guys do....just make the rendering faster:
the faster we can render, the sooner we can go home to our families!

[instinct]

whatever you guys do....just make the rendering faster:
the faster we can render, the sooner we can go home to our families!

To be honest i stopped hoping for that day...we'll just end up stuffing more things in there hehe...there was that diffraction (slit experiment) thread just recently...so after RGB then Spectral we will prolly move towards
less biased methods (that are way more computational expensive) step by step (wich brings me back to the idea that current renderers might be less biased or more biased but surely not unbiased :P )

Regards,
Thorsten