Author Topic: Corona Renderer 4 for 3ds Max - Daily Builds Discussion  (Read 109742 times)

2019-04-29, 16:33:58
Reply #330

rowmanns

  • Corona Team
  • Active Users
  • ****
  • Posts: 1892
  • Corona for 3ds Max QA Team
    • View Profile
Hi!
When using IES, the mirror does not reflect light at all, only a black spot.

Hi,

Are you able to send over your scene? You can use our uploader which is here: https://corona-renderer.com/upload

Please don't forget to let me know the name of your file.

Thanks,

Rowan
Please read this before reporting bugs: How to report issues to us!
Send me your scene!

2019-04-29, 18:36:02
Reply #331

cgcrazy

  • Active Users
  • **
  • Posts: 11
  • CG Artist
    • View Profile
    • cgcrazy.com
Hi!
When using IES, the mirror does not reflect light at all, only a black spot.

Hi,

Are you able to send over your scene? You can use our uploader which is here: https://corona-renderer.com/upload

Please don't forget to let me know the name of your file.

Thanks,

Rowan

Sent to dropbox. The file is called "hallway"
Real art begins with an idea that is improved with its implementation.
cgcrazy.com
info@cgcrazy.com

2019-04-30, 00:16:20
Reply #332

Artisticpixel

  • Active Users
  • **
  • Posts: 74
    • View Profile
    • artisticpixel
Hi guys

Im using the latest daily and I have a current interior scene with corona sun/sky.  I just rendered off a 3000 pixel wide image with the pass limit hittng 200 and the noise level at 4% and its still noisy?
never had this before?  Anyone else experiencing this?

Yes I am noticing heavier noise in my scenes with the latest build. I also had to render an interior scene for 10 hours on a Xeon dual CPU which Is unheard of for this machine, said it had 72 hours left to render for 2.9% noise. I had rendered it on a previous build for 6 hours and the noise was definitely acceptable. Seems to be going backwards in the cleaning noise department on my machines, but I accept it as it’s a daily build and probably work in progress. I just hope they have Xeons they are testing the software on too.

2019-04-30, 01:45:44
Reply #333

PROH

  • Active Users
  • **
  • Posts: 1219
    • View Profile
Yes. I've also had some unexpected prolonged rendertimes, and made a little test with a very simple scene that showed, that just by enabling Corona HQ Denoiser the rendertimes got 6-12 times longer than when rendering with "Remove fireflies only". The "12x" experience was with HQ filtering enabled, and the "6x" without HQ filtering. The calculated noise level was simply changing dramatically for the same scene depending on denoiser being on or off, and the HQ filtering setting.

Biggest problem is, that after going back to v3,1 this still happens (6 times longer with HQ Denoiser).

What's going on?

2019-04-30, 12:51:11
Reply #334

PROH

  • Active Users
  • **
  • Posts: 1219
    • View Profile
Hi. A bit more info regarding the problems with rendertimes and noisecalculation I mentioned in my previous post.

The scene used is lit by an HDRI (exr) and contains 1 tree-object placed on a ShadowCather-plane. I haven't found the culprit, but here's my findings so far (all done with DB April 23 2019, with a 5% noise limit):

1) Setup: HQ filtering, no denoise

Estimated time                          - 1min 3 sec
Noise calculated (after 5 passes) - 8,26%

Actual rendertime                      - 2min 16sec
Final noise-level                        - 4,79%
Passes total                              - 15


2) Setup: HQ filtering, HQ denoise

Estimated time                          - 10min 24sec
Noise calculated (after 5 passes) - 24,70%

Actual rendertime                      - 28min 20sec
Final noise-level                        - 4,90%
Passes total                              - 110


3) Setup: normal filtering, no denoise

Estimated time                          - 46sec
Noise calculated (after 5 passes) - 7,17%

Actual rendertime                      - 1min 12sec
Final noise-level                        - 3,57%
Passes total                              - 10


4) Setup: normal filtering, HQ denoise

Estimated time                          - 8min 32sec
Noise calculated (after 5 passes) - 21,06%

Actual rendertime                      - 6min 03sec
Final noise-level                        - 4,90%
Passes total                              - 30


I never noticed this behavior before, and honestly thinks this is a bug. After theese findings in DB, I rolled back to v3,1 but things didn't change - even after a uninstall/reinstall :(

Does anybody here have similar experiences or explanations?

Regards

2019-04-30, 16:44:16
Reply #335

rowmanns

  • Corona Team
  • Active Users
  • ****
  • Posts: 1892
  • Corona for 3ds Max QA Team
    • View Profile
Yes. I've also had some unexpected prolonged rendertimes, and made a little test with a very simple scene that showed, that just by enabling Corona HQ Denoiser the rendertimes got 6-12 times longer than when rendering with "Remove fireflies only". The "12x" experience was with HQ filtering enabled, and the "6x" without HQ filtering. The calculated noise level was simply changing dramatically for the same scene depending on denoiser being on or off, and the HQ filtering setting.

Biggest problem is, that after going back to v3,1 this still happens (6 times longer with HQ Denoiser).

What's going on?

Hi,

We have done some tests today and we weren't able to reproduce the slowdowns which you are experiencing.

Are you able to send over your scene? Can you also provide the specification of you computer too?

Thanks,

Rowan
Please read this before reporting bugs: How to report issues to us!
Send me your scene!

2019-04-30, 17:14:00
Reply #336

PROH

  • Active Users
  • **
  • Posts: 1219
    • View Profile
Hi Rowan

The file is uploaded via your privat uploader. The filename is: 1556636633_HQ-Plants1-birch-04-11-5m-max2014-C4-daily23-04-2019-v1.zip

My PC specs are: i7-3930K, 32Gb RAM, Windows 10 PRO, 64 bit

I've tested this with both 3dsMaxDesign 2014 & 3dsMax 2016. Both giving same result.

I've tried to use Sun & Sky instead of HDRI, tries without translucency, tried without bump, and tried without ShadowCatcherMtl. Nothing helps. The only thing that changed things dramatically was when I tried without both bump & ShadowCather-Mtl. But unfortunately this only worked the very first time. After that it returned to be problematic. As already mentioned, I'm now using v3,1 without anything being better.

Hope you'll find out whats causing this.

Best regards

2019-05-01, 23:46:32
Reply #337

Artisticpixel

  • Active Users
  • **
  • Posts: 74
    • View Profile
    • artisticpixel
I never noticed this behavior before, and honestly thinks this is a bug. After theese findings in DB, I rolled back to v3,1 but things didn't change - even after a uninstall/reinstall :(

Does anybody here have similar experiences or explanations?

Regards

Yes I rolled back 2,3 and 4 versions back and the long render time remains ( 64 hours to do a 3k render) and the same scene took 6hrs for a 4.5k render a few months ago. Unfortunately this scene I cant share due to restrictions on an NDA I have signed but I usually share when I have an issue, i just double checked the scene again after reading the HQ Filter test you performed. Looks like my scene went from 64hr render to 14hrs once I deactivated HQ filtering in the latest daily build. Working on a Dual Xeon E5-2698v3 with 128 Gb RAM.
« Last Edit: 2019-05-02, 00:37:17 by Artisticpixel »

2019-05-02, 04:32:59
Reply #338

ihabkal

  • Active Users
  • **
  • Posts: 253
    • View Profile
You have a bit of an older system, what is your graphics card?


I never noticed this behavior before, and honestly thinks this is a bug. After theese findings in DB, I rolled back to v3,1 but things didn't change - even after a uninstall/reinstall :(

Does anybody here have similar experiences or explanations?

Regards

Yes I rolled back 2,3 and 4 versions back and the long render time remains ( 64 hours to do a 3k render) and the same scene took 6hrs for a 4.5k render a few months ago. Unfortunately this scene I cant share due to restrictions on an NDA I have signed but I usually share when I have an issue, i just double checked the scene again after reading the HQ Filter test you performed. Looks like my scene went from 64hr render to 14hrs once I deactivated HQ filtering in the latest daily build. Working on a Dual Xeon E5-2698v3 with 128 Gb RAM.



2019-05-02, 08:44:26
Reply #339

Artisticpixel

  • Active Users
  • **
  • Posts: 74
    • View Profile
    • artisticpixel
You have a bit of an older system, what is your graphics card?


I never noticed this behavior before, and honestly thinks this is a bug. After theese findings in DB, I rolled back to v3,1 but things didn't change - even after a uninstall/reinstall :(

Does anybody here have similar experiences or explanations?

Regards

Yes I rolled back 2,3 and 4 versions back and the long render time remains ( 64 hours to do a 3k render) and the same scene took 6hrs for a 4.5k render a few months ago. Unfortunately this scene I cant share due to restrictions on an NDA I have signed but I usually share when I have an issue, i just double checked the scene again after reading the HQ Filter test you performed. Looks like my scene went from 64hr render to 14hrs once I deactivated HQ filtering in the latest daily build. Working on a Dual Xeon E5-2698v3 with 128 Gb RAM.

I have a GTX 1060 in one and GTX 1080 in the other

2019-05-02, 10:03:49
Reply #340

Artisticpixel

  • Active Users
  • **
  • Posts: 74
    • View Profile
    • artisticpixel
Reverting back to previous release hopefully speeds the render time up again
« Last Edit: 2019-05-02, 10:07:22 by Artisticpixel »

2019-05-02, 11:36:59
Reply #341

PROH

  • Active Users
  • **
  • Posts: 1219
    • View Profile
Hi ihabkal. My Quadro card died recently, so at the moment I'm using an old Geforce 880 GT card. Since I'm not using Nvidias AI Denoiser, this shouldn't affect Coronas rendertimes.

ATM the only way to use Corona, is to render without HQ denoiser. So I'm hoping there'll be a fix for this very soon!

2019-05-02, 13:38:53
Reply #342

rowmanns

  • Corona Team
  • Active Users
  • ****
  • Posts: 1892
  • Corona for 3ds Max QA Team
    • View Profile
Hi,

Okay so it looks like there are two seperate issues here:

Issue 1: Slowdown with the HQ filtering. - We are aware of this issue and are currently working on a fix for it.

Issue 2: Denoising vastly increasing the render time when enabled while using a noise limit. - I did some testing on this today and I can see that while using a noise limit with the HQ denoiser my render times jump up quite a lot.

However if I use a pass limit instead of a noise limit my render times are roughly the same with/without the denoising selected. @PROH can you try to render using a pass limit instead of a noise limit for the time being and see if that helps for you?

We will continue to investigate this too.

Thanks,

Rowan

Please read this before reporting bugs: How to report issues to us!
Send me your scene!

2019-05-02, 14:32:53
Reply #343

romullus

  • Global Moderator
  • Active Users
  • ****
  • Posts: 8778
  • Let's move this topic, shall we?
    • View Profile
    • My Models
@rowmanns, i think the point of the report is, that simply enabling HQ denoising, reported noise level is vastly higher than without denoising. Testing with pass limit would give no usefull information in this case, since passes takes about the same time to complete in both cases.
I'm not Corona Team member. Everything i say, is my personal opinion only.
My Models | My Videos | My Pictures

2019-05-02, 15:33:13
Reply #344

PROH

  • Active Users
  • **
  • Posts: 1219
    • View Profile
Yes, romullus is right. Rendertimes pr. pass is about the same, but the reported noise level differs a lot, making the calculated noise level useless/unreliable.

And yes, I agree there are the two issues you mention.

Another thing that bothers me is, that if adaptivity is disabled, then there's no noise level calculated.

Best regards