V-Ray 5 very slow in comparison to 3.7

[PdZ]

Hi all,

I dont know if I do something wrong (I know Irradiance Map and Photonmap are nonexistent) but It seems V-Ray is so slow when rendering in comparison to V-Ray 3.7.
Even BR-LC on 3.7 seems a LOT faster. Also Rendering Interiors with 5 is exceptional slow and noisy. What do I do wrong? Where is it possible to optimize performance?

Thx in advance!

Settings:

Using HDRI as Light.
Bucket Sampler Min 1/Max 10

Everything else is default I think.

HDRI has 16 K. Its nothing in the scene except Grass with 2sided and white geometry (no texture applied)
see screenshot. Even the white geometryparts render so slow but theres nothing to render except a white material.

UPDATE:
Do you see the estimate? 1h18 minutes? For that Image? (5000x2812px)

Iam rendering with one host System:
AMD Ryzen 9 3900X 12-Core Processor 3.80 GHz
64 Gig Ram (its at 37 Gigabyte used)

and with a DR:
AMD Ryzen Threadripper 2990WX 32-Core Processor 3.20 GHz
64 Gig Ram

CPUs are all at 99%....

[sirio76]

In my experience overall speed is pretty similar, it is just that denoiser on 3.7 was working much better and that helps to cut render time significantly.
I never used 0.01 threshold on 3.7, normally I keep threshold between 0.02 and 0.06 and let the denoiser do his job, on V5 I'm often forced to use 0.01 to 0.005 because the new denoiser is largely ineffective for retaining sharp texture details.

[PdZ]

Hey sirio,

thx for your response. Then I dont know what iam doing wrong. It seems extreme more slow. And we decided not to render interiors with V-Ray 5 at all because in 3.7 its hella fast with Irradiance map LC. In V-Ray 5 it takes ages and is noisy as hell...

[sirio76]

I can not compare render done with Irradiance Map since I haven't used biased solution in ages. Back when I've switched to BR/LC (seems to remember it was with a 3.x update several years ago) I've done a few tests and in the end the BR coupled with a well setup denoiser was usually as fast as the IM/LC solution. Not only it was equally fast, it was also immediate to setup while to get nice detail with IM you often need to fiddle with the settings wasting more time, so in the end BR was probably faster overall.
Without using denoiser, for the same render time you will indeed get a much noisier image compared to IM, but as said the old denoiser was very effective with that.
IMO it is all about the denoiser, just to be sure of the render speed create an identical scene in 3.7 and 5 using BR/LC on both, my guess is that you will get similar results.

[aleksandar.kasabov]

Hello PdZ,

Could you please share the scene which you use to compare the performance of V-Ray 3.7 and V-Ray 5? What render times are you getting with both render engines?

You can also check the following article, explaining the best render settings for fast renders and good render quality.

[andreas_hummel]

Hello, I have also very long rendertimes for a clean rendering, but maybe I make something wrong. The model i a church with lofts. This Lofts limit the light. The choir windows are stained glass windows. I had a VrayObjectProperties with no visibility for GI. The normal windows outside the choir have VrayObjectProperties with no shdows and no visibility for GI. The light source is only Vray sun and sky. The dome light have a Diffuse contribution of 3 and a specular Contribution of 3. This was for avoid too dark area per exemple under the first loft. The test1 is a cutout from 70% an this was rendering 90 minutes with 136 passes. For a clean 360° panorama 9000x 4500 pix with denoisier I need more than 30 h with AMD Ryzen threadripper 3960 X 24 core.

Has someone a tipp? thanks andreas

[aleksandar.kasabov]

Hello andreas_hummel,

Could you please share the scene in question with all assets for it so we can have a look at what might be the reason for the long render times? You can send it to us using our contact form. Please make sure to mention this thread in your ticket. Thank you.

[andreas_hummel]

Hello aleksandar.kasabov , I was seending you a message because of the upload.
Too long rendertimes: What I found and it was maybe not a good idea from me: The dome light have a diffuse contribution of 3 and a specular contribution of 3. That increase the rendertime of around 200%. Better is to work with exposure and filmic tonemap to avoid too dark areas. The next test is arealights instead the dome light (vray sky) for complicated indoorscene. But maybe I have also wrong materials ( too light color) or another incorrectly setting?

[RwVision]

Hi all,

Also Rendering Interiors with 5 is exceptional slow and noisy. .

Hi

Yes, you are completely right.

Vray 5 is not capable of producing noise free renderings.

And in interiors mostly, it is due to forced BR method. The poor denoiser is also a big issue, but mainly an uneven, high noise when using BR.
I wrote about this on the forum, my recent noise topic with my scene also talks about that.

I still use 3.7 and will be using it, because of many aircraft interior animations I do.
IR+LC + Detail enhancement could produce the same or even better detail result in interiors, where it was clean from noise and rendered 2x faster. As a result, in 3.7 when using denoiser - I'm still at least 2-3 times faster and better than V5.

It does of course depend on the scene, but for me V5 is a serious downgrade in speed vs image quality.

In my experience overall speed is pretty similar, it is just that denoiser on 3.7 was working much better and that helps to cut render time significantly.
I never used 0.01 threshold on 3.7, normally I keep threshold between 0.02 and 0.06 and let the denoiser do his job, on V5 I'm often forced to use 0.01 to 0.005 because the new denoiser is largely ineffective for retaining sharp texture details.

100% same here - Every time I'm wondering what's wrong, because I keep pushing the threshold below 0.02, 0.01... and the image is still noisy! 90% of it is using BR method, but since it is forced now, Vray should do some progress on it?
I end up mostly using 0.01, and still have to add that blur-ish denoiser on like 0.6-0.7 opacity.

The worst part is that even pushing the threshold to 0.05 still produces noise - where in 3.7, even on BR I was able to get rid of it only by going higher with the min AA.

[sirio76]

Have you tried to denoise a V5 image using the 3.7 standalone denoiser?
It looks like the V5 denoiser don't read some of the needed channel.

About render time and noise in general, I've noted that on V5 I often get better results (quality/rendertime) by lowering the max AA to 16/12/8 and rising the threshold to 0.005.

[RwVision]

Have you tried to denoise a V5 image using the 3.7 standalone denoiser?
It looks like the V5 denoiser don't read some of the needed channel.

About render time and noise in general, I've noted that on V5 I often get better results (quality/rendertime) by lowering the max AA to 16/12/8 and rising the threshold to 0.005.

Yes, you can't use old denoiser on V5 sadly.

The Max AA on 10-12 is my default setup ever since 3.7, so I'm already optimized. I really don't get why 24 is a default value. Rarely a scene benefits from AA max at 12+.

[pardalis]

why don't you use an external denoiser?

[sirio76]

Because a good integrated denoiser (like on 3.7) is superior

[jeremiah_jones]

Yes, you can't use old denoiser on V5 sadly.

The Max AA on 10-12 is my default setup ever since 3.7, so I'm already optimized. I really don't get why 24 is a default value. Rarely a scene benefits from AA max at 12+.

We are facing a similar issue with the noise being a major issue in Vray 5 to the point that I'm also thinking of downgrading back to 3.7. The problem is that while a universal setting approach would be great, we can't seem to get clean results without really cranking up the bucket image sampler settings and in turn the render times are ridiculous. We tend to do animations quite often as well so having renders time increase drastically is a big issue for us. The other issue I'm having is that the hdri solution we have been using now produces a ton of white specks in the specular highlights. This was once resolved by just turning on clamp output and sub pixel mapping. Now that 5 removed clamp output that is also an issue. I have been trying to resolve the issue but I'm not having much luck.