Can 1.5 replicate 1.47.03 ?

[Adam H. Stewart]

I've been pulling my hair out - and there ain't that much of it to begin with - trying to get 1.5 to work the same as 1.47.03. We often have to re-render scene's or parts of scene's many months after the fact. So my question to myself and to anybody who is interested - with all the anti-alias setting and color mapping changes can 1.5 flawless replicate a scene built and previously rendered in 1.47.03 - my experiments seem to suggest that it can't. There may be some magical setting out there that make 1.5 act like 1.47.03 - if those settings exist I'd give my left bollock for them.

So far for me 1.5 seems to make alot of things a problem that I never had a problem with before - it seems to have difficulty dealing with 3dsmax procedurals - falloff and smoke in particular ( see example above ) If I can't get this figured out I won't be moving the render farm to 1.5 which is a damn shame

[DaForce]

I believe relfection and refraction have changed a fair bit, as well as the rQMC noise threshold, not to mention SSS as well.

There is no magic list of settings that you can dial in to get 1.5 to act like 1.47.03.
You could also look at it as 1.47.03 rendered wrong and 1.5 is now right.

[Adam H. Stewart]

Theory

That's great in theory - and if there was an easy way to switch between versions of Vray ( maybe there is ) but I have been running on 1.47.03 for 2-3 years, it pisses me off that I might have to go back and re-engineer scene's that I was happy with, and God only knows what problems are going to come up scene to scene, the art directors and QC staff I work with are anything but forgiving.
I don't think I am the only one having problems with the new AA settings - I'd be happy with default settings in 1.5 that work the same way that the default settings in 1.47.03 did, all this going into the rQMC and changing .001 to .005 seems so unnessarily tweaky and complicated

[DaForce]

I dont think there is much that can be done. Vray has evolved in a sense.

I guess the default settings could be changed abit, but even that wont help with things like reflect/refract/sss as these have change by design in some ways.
So sure it would be handy to have the SSS reflection as it was in previous builds but in short they were wrong and that needed to be fixed.

By the looks of it, in your situation is may have affect you more than some others, I know for most of the people that I have spoken to it has been no problem at all really.
All depends on your workflow I guess.

[vlado]

If you need the exact same result as the 1.47.03 version, it will save a lot of headaches if you just use that version.

It was the same case with transition from 1.09 to 1.47, and some people still think that 1.09 works for them better than 1.47, even though 1.09 is nowhere near the capabilities of newer versions... so go figure.

Best regards,
Vlado

[metroberlin]

i totally understand ...but nevertheless we have at the moment a big problem here because of that. on the one hand we need to finish a hudge animation project that we started with 1.48.xx. this will go until mid november. we tried to switch to 1.5 and before re.install the renderfarm we did some tests. the biggest issue we have is the AA. the same values that worked in 1.48.xx looks terrible. the new ones 1.5 take much much more time. for such a hudge project not possible to have this extra time. so far we didn´t find out which settings are at least appr. like the old ones in terms of time/quality. but then on the other hand the new features in 1.5 would help us a lot in our every day work. is there any chance to get an idea what to change how in 1.5 to simulate 1.48 ? (although vlados reply answered it actually) Or is there a way to run 1.48 AND 1.5 at the same time on the machines ?

cheers

marc

[vlado]

is there any chance to get an idea what to change how in 1.5 to simulate 1.48?

That depends on what settings you used originally, and what you have in your scene, so if you can provide more information, I can be more specific.

In general, in certain situations the 1.47 version could take too many shortcuts producing worse image quality that required additional tinkering to get right. The 1.5 version is more strict in this regard and cheats less.

On a bright note, the next RC is compiled with a newer compiler version which makes it somewhat faster.

Best regards,
Vlado

[DaForce]

faster on all CPU's or mainly AMD's?

As in that stupid Intel compile issue that made things slower on AMD machines.

[metroberlin]

thanks for the answer..... i have no issue with that a reflection might look a bit different to the old version etc. just the AA kills us. the settings we use in the 1.48.03 version are:



a bit unusual, but for that scene they work perfectly in terms of quality / time. if we just open and render the scene in 1.5 the rendertime increase from 2min to 7min and creates even a bit of a flickering. increasing the treshhold values (like suggested) makes it even worse...of course.

thanks

marc

[percydaman]

adaptive qmc uses your qmc settings to dictate quality, so without showing your qmc settings, your screengrab is pretty useless.

[metroberlin]

yes...sure...forgot...

adaptiv amount: 0,85
noise treshold: 0,005
min. samples: 15
global: 1

[Da_elf]

faster on all CPU's or mainly AMD's?

As in that stupid Intel compile issue that made things slower on AMD machines.

i agree

[vlado]

faster on all CPU's or mainly AMD's?

As in that stupid Intel compile issue that made things slower on AMD machines.

We found it to be faster on both Intel and AMD processors.

Best regards,
Vlado

[DaForce]

faster on all CPU's or mainly AMD's?

As in that stupid Intel compile issue that made things slower on AMD machines.

We found it to be faster on both Intel and AMD processors.

Best regards,
Vlado

Ah ok cool.

What sort of ball park are we looking at... a few percent or something in the range of 10-20

[vlado]

Well, in one particular case, a scene that rendered for 7m43s with the old build, rendered for 7m15s with the new one, which is somewhere about 9% faster. In another case, a scene that rendered for 44s before, rendered for 36s, which is about 18% faster, I think. Of course, it varies from scene to scene. I just hope that the speed increase is not on account of something else

Best regards,
Vlado