Daz 4.11 Beta testing

FlipTopBin

Member
Game Developer
Dec 5, 2017
176
223
Just been doing some testing with the 4.11 beta and although my testing is very limited so far, the results are frankly astonishing. The upgraded version of Iray appears to be roughly four times faster than 4.10!

Using a scene with one figure, a reasonably high poly count background, 1 HDRI and two spots rendering at 1080p using an i7+1080TI:

4.10 - 38 minutes
4.11 - 10 minutes

The denoiser didn't seem to make a scrap of difference either so this is just down to the engine upgrade I believe.

It's not perfect yet though. Something goes horribly wrong at high resolutions and the thing just won't render at all but if you can drop the pixel count you are golden!
 

Loys

Member
Game Developer
Oct 19, 2018
411
4,761
You don't have permission to view the spoiler content. Log in or register now.
That's kinda surprising... On my PC with 1070 i've noticed about ~20% increasing render times of the same number of iterations on 4.11 beta comparing to 4.10. But the denoiser helps daz to complete rendering faster...:(Somehow you have totally different results.
 

FlipTopBin

Member
Game Developer
Dec 5, 2017
176
223
I am working with a cenvergence target rather a specific number of iterations. Perhaps each iteration takes longer but they converge faster?
 

Loys

Member
Game Developer
Oct 19, 2018
411
4,761
I am working with a cenvergence target rather a specific number of iterations. Perhaps each iteration takes longer but they converge faster?
It always confused me - understanding how that thing works. So i decided to turn it off... It can be the reason of this difference, why not. I don't know to be honest.:confused:
 

Aeilion

Member
Jun 14, 2017
125
144
That's kinda surprising... On my PC with 1070 i've noticed about ~20% increasing render times of the same number of iterations on 4.11 beta comparing to 4.10. But the denoiser helps daz to complete rendering faster...:(Somehow you have totally different results.
I have not calculated how much but I also have a slowdown in 4.11 and an increase in memory requirements. And I still have not understood what denoiser was used for ^^ I'm really not good at it.
 

Nolander

Member
Oct 19, 2018
108
111
I have not calculated how much but I also have a slowdown in 4.11 and an increase in memory requirements. And I still have not understood what denoiser was used for ^^ I'm really not good at it.
The denoise filter previously hasn't worked for me and I've always used Photoshop so I'll give it a test run.
 

basium

Member
Game Developer
Feb 22, 2017
186
776
Wow need to give this a try. Let's see if my library gets fucked up beyond repair :p
 
S

SteelRazer

Guest
Guest
There is a new thing where primitives are created with default iray uber shader, iray got updated so new gpus work(optix on new gpus not tho). Denoiser is kinda meh, if you have good gpu it's a very much disabled because it kinda washes out skin detail.
 
Last edited by a moderator:

basium

Member
Game Developer
Feb 22, 2017
186
776
Put my white lab coat on and did some quick non scientific tests :p
Win 10 with 1070 + 1080ti
Max samples is set to 1000
Firefly filter enabled
Lighting used: Sun primitive + Ghost lights + Emissive lights + 1 spot light

Daz Studio 4.10
daz 4.10 - 5 minutes 22 seconds  -- stock.png 5min 22sec - My regular settings
daz 4.10 - 5 minutes 17 seconds  -- render quality on - cov ratio 99%.png 5min 17sec - Added render quality on and set cov ratio to 99%
daz 4.10 - 5 minutes 19 seconds  -- render quality on - cov ratio 99% - noise filter on.png 5 min 19sec - Added noise filter

Daz Studio 4.11
daz 4.11b - 6 min 43 sec  -- stock.png 6min 22sec - My regular settings
daz 4.11b - 6 min 35 sec  -- render quality on - cov ratio 99%.png 6min 35 sec - Added render quality on and set cov ratio to 99%
daz 4.11b - 6 min 34 sec  -- render quality on - cov ratio 99% - post denoiser.png 6min 34sec - Added Post denoiser

It's definitely slower for me. But holy crap what a difference with the post denoiser. It seems to make it more blurry but never seen anything so effective at cleaning up the image. Maybe dark night scenes finally are doable?
 
Last edited:
  • Like
Reactions: Domiek

xht_002

Member
Sep 25, 2018
342
353
Put my white lab coat on and did some quick non scientific tests :p
Win 10 with 1070 + 1080ti
Max samples is set to 1000
Firefly filter enabled
Lighting used: Sun primitive + Ghost lights + Emissive lights + 1 spot light

Daz Studio 4.10
View attachment 179982 5min 22sec - My regular settings
View attachment 179985 5min 17sec - Added render quality on and set cov ratio to 99%
View attachment 179986 5 min 19sec - Added noise filter

Daz Studio 4.11
View attachment 179987 6min 22sec - My regular settings
View attachment 179994 6min 35 sec - Added render quality on and set cov ratio to 99%
View attachment 179995 6min 34sec - Added Post denoiser

It's definitely slower for me. But holy crap what a difference with the post denoiser. It seems to make it more blurry but never seen anything so effective at cleaning up the image. Maybe dark night scenes finally are doable?
:coffee: they all look exactly the same, except the last image, if you look real close at the leather gradient on the chairs
 
Last edited:

basium

Member
Game Developer
Feb 22, 2017
186
776
yup. Nothing changed. Just the cleaning up (last image) and it looks like 4.11 added more overall lighting
 

Aeilion

Member
Jun 14, 2017
125
144
I did some personal tests too. Because it bothered me a bit this story.

Finally there is a big change in daz4.11. Which ? No idea !

I use a i5 7400, 16gb ram, gtx 1060 6g
use of cpu and gpu for render and activation of optiX

It is 1920 by 1080, filter mitchel 1
Quality at 10, iteration max (In short ... all the max except time)

render 4.10 of 30min:

iteration 1458
convergence 24.09

render in 4.11 (same starting file and setting):

iteration 945
convergence 4.20

identical images almost exept ... Daz4.11 manages the lighting differently so it would seem the image is brighter (over-illuminated if you start from a well-lit stuff ...)
I tested with a render of one hour and the differences at the iteration and convergence level remain the same.
So I would say in the end almost no differences except in the numbers. And that damn light ...

-----------------------------------------------------------------------------------------------

Then I tested the denoiser ...

Two things :

It is super effective to smooth ... a difficult rende that takes me an hour can be finished in ... 5 min.
Except that you screw up some hair (not good quality) but especially the skin details.

And it moves things ... literally!
Strangely for one of my tests it was the eyes and it was better actually ahahah
For another it was a bikini cord, strange ...

But ... it can remain very interesting for some specific scenes and especially for the devs that starts (with small budjet). The saving of time and "cleanliness of the image" is quite impressive.
For people who want to make beautiful pictures ... It does not do it except for some very specifics situations. (or maybe with a good computer ^^)
 

basium

Member
Game Developer
Feb 22, 2017
186
776
Did full renders without any sample limits.
I put rendering quality on and cov ratio to 99%

Daz 4.10
Daz 4.10 42m 27s 10 409samples.png 42 min 27 s - It ran to around 10 409 samples. Noise filter was off

Daz 4.11
Daz 4.11b - 56m 18s 11100samples -- render quality on - cov ratio 99% - post denoiser.png 56 min 18 s - It ran to around 11 100 samples. Post denoiser was on
 
Last edited:

basium

Member
Game Developer
Feb 22, 2017
186
776
Found this testing scene from daz forums by DAZ_Rawb:


daz 4.10
4 min 27 sec

daz 4.11
4 min 47 sec

In advanced I ticket in CPU too and it used both my gpus + cpu to render and it took 4 min 54 sec...
Me no understand. Me go fuck daz 4.11. Me go drunk now.
 
Last edited: