slow vray Next vs 3.6

[Karonte]

Hi guys. I switched all my4 licenses to Vray next4 yesterday. I opened a simple scene and the render was 2 times SLOWER. I cannont belive it. Am I wrong with something? the scene is a simple white plane/wall and a sofà for a still life render. My 3.6 render is 2.7 min and with gray Next 4.20 minutes! I know that the scene was built with 3.6 originally but 2 time slower is unacceptable because my studio has a lot of files that need to reopen and re-render. for us is a big problem...Any help? ( I search in the forum...but found nothing)

[JonasNöll]

I experienced the same, contacted support and apparently there is no easy fix to render old scenes setup with 3.6 with the same speed in V-Ray Next. Apparently the introduction of new technologies results in an overall decrease in rendertime which you can reduce by enabling new optimizes technologies such as adaptive domelight, denoiser and so on, but it depends on your scene if that makes a huge difference. Was also a bit disappointed when I tried out Next for the first time and expected an overall speedup which resulted in the exact opposite.
As said that is so far my experience but I would be happy to be corrected

PS: I opened a thread about it in the V-Ray 5 Beta forum, so if you got access you can maybe find some information there as well:
https://forums.chaosgroup.com/forum/...getting-slower

[vlado]

It would be best to get us some example scenes. From the tests that I've made with the scenes provided so far (including those by Jonas), I can conclude that the overall raytracing speed is exactly the same or faster; however the defaults are different and in some cases we have decided to make the engine more robust (f.e. less flickering with the light cache) at the expense of some additional work.

Best regards,
Vlado

[Karonte]

Thanks for the answer. I was VERY disappointed after years of vray ( since its first beta). I expected to be fast as the same old (or slower,2/3%). Not twice! Lot of studies have project that needs to be reopened. I did non used domelight and denoiser. The thing that i don't like also, is that i tried corona demo, convert the scene with one click script and the time was faster than Next. I have the doubt that I bought the wrong one. . In the meanwhile, finding a solution, i willi use 3.6.
In attachment a small set with only vray lights . Vray3.6=2.07min, Vray next = 4.20 min.

[vlado]

Like I said, it would be best to provide examples. My own tests don't indicate that Next is any slower.

Best regards,
Vlado

[JonasNöll]

It would be best to get us some example scenes. From the tests that I've made with the scenes provided so far (including those by Jonas), I can conclude that the overall raytracing speed is exactly the same or faster; however the defaults are different and in some cases we have decided to make the engine more robust (f.e. less flickering with the light cache) at the expense of some additional work.

Best regards,
Vlado

Could you explain how to test and compare the overall raytracing speed between 3.6 and Next, as I would be very interested in that. If the defaults are different (even though the rendersettings are exactly the same?) it would make it very hard to compare anything or? It's just a bit frustrating for the user to upgrade to the latest version, open his old scenes and see everything render slower as before, so marketingwise this is maybe not the best move

[JonasNöll]

Thanks for the answer. I was VERY disappointed after years of vray ( since its first beta). I expected to be fast as the same old (or slower,2/3%). Not twice! Lot of studies have project that needs to be reopened. I did non used domelight and denoiser. The thing that i don't like also, is that i tried corona demo, convert the scene with one click script and the time was faster than Next. I have the doubt that I bought the wrong one. . In the meanwhile, finding a solution, i willi use 3.6.
In attachment a small set with only vray lights . Vray3.6=2.07min, Vray next = 4.20 min.

I downloaded both images and visually they seem to have the same amount of noise. If I put them in Photoshop and set the layer to "difference" it also shows that there is almost no difference in terms of noise. Somehow if I boost up the exposure there is something odd happening, so maybe it can give you a hint of what's wrong, from my understanding those different colored buckets shouldn't be there and maybe something went wrong during the rendering... (?) But also could have something to do with jpeg compression, I just found it odd...

[Karonte]

1) I opened the scene with 3.6. put pixel to 1500 and hit render ( no DR render)
2) uninstall and install Vray Next.
3) open the same scene, put to 1500 pixel and hit render ( no DR render).

Even if my file is a test and is not optimized, the same thing has to be in both cases ( 3.6 and Next). So the difference is too much in my opinion. Sorry if I write so quick but this morning I have deadlines with job . In the afternoon if I have to try something else let me know. Thanks to all.

[vlado]

Could you explain how to test and compare the overall raytracing speed between 3.6 and Next, as I would be very interested in that. If the defaults are different (even though the rendersettings are exactly the same?) it would make it very hard to compare anything or? It's just a bit frustrating for the user to upgrade to the latest version, open his old scenes and see everything render slower as before, so marketingwise this is maybe not the best move

Sure, I did several tests in 3ds Max 2019 through 2021. First, I opened your scenes without changing any settings. The render times were very similar, as well as the overall number of rays traced (see render_comp_interior.png and render_comp_teapot.png). This generally means that the raytracing speed is not significantly worse.

When the settings are reset (render_comp_interior_reset.png), V-Ray Next and V-Ray 5 tend to trace more rays, which increases the render time a bit. However, I should also note that with V-Ray Next and V-Ray 5, the scene rendered slightly brighter and more accurate, so V-Ray had a bit more work to do than in V-Ray 3.7.

I also tried other scenes, like archinteriors 55, scene 2, which rendered a lot longer in 3.7 due to stuck buckets on some glass objects, which were gone in V-Ray Next and V-Ray 5 (render_comp_ai55_002.png).

One obvious result from these tests is that the hash map based light cache traces a bit more rays. Like I mentioned in the other post, this is intentional as we want it to be more stable in animations out of the box without users needing to fiddle with settings. It is possible that this can be improved and we will be looking into it.

If there are other scenes that you want me to examine, please send them over to vlado@chaosgroup.com

Best regards,
Vlado

[JonasNöll]

1) I opened the scene with 3.6. put pixel to 1500 and hit render ( no DR render)
2) uninstall and install Vray Next.
3) open the same scene, put to 1500 pixel and hit render ( no DR render).

Even if my file is a test and is not optimized, the same thing has to be in both cases ( 3.6 and Next). So the difference is too much in my opinion. Sorry if I write so quick but this morning I have deadlines with job . In the afternoon if I have to try something else let me know. Thanks to all.

If you can upload the scene I can also make a test, I have 3.6, Next and 5 Beta all installed on the same computer. As said I'm also trying to track this problem for a while and so far didn't find some solution but I'm interested to investigate...

[JonasNöll]

Sure, I did several tests in 3ds Max 2019 through 2021. First, I opened your scenes without changing any settings. The render times were very similar, as well as the overall number of rays traced (see render_comp_interior.png and render_comp_teapot.png). This generally means that the raytracing speed is not significantly worse.

When the settings are reset (render_comp_interior_reset.png), V-Ray Next and V-Ray 5 tend to trace more rays, which increases the render time a bit. However, I should also note that with V-Ray Next and V-Ray 5, the scene rendered slightly brighter and more accurate, so V-Ray had a bit more work to do than in V-Ray 3.7.

I also tried other scenes, like archinteriors 55, scene 2, which rendered a lot longer in 3.7 due to stuck buckets on some glass objects, which were gone in V-Ray Next and V-Ray 5 (render_comp_ai55_002.png).

One obvious result from these tests is that the hash map based light cache traces a bit more rays. Like I mentioned in the other post, this is intentional as we want it to be more stable in animations out of the box without users needing to fiddle with settings. It is possible that this can be improved and we will be looking into it.

If there are other scenes that you want me to examine, please send them over to vlado@chaosgroup.com

Best regards,
Vlado

Amazing, I will go through it and try to understand everything. Just one quick question though, did a lot changed from 3.6 to 3.7? I didn't try 3.7 but 3.6 and there it seemed to render faster than Next. In this thread here the comparrison is also about 3.6 to Next. So 3,6 and 3,7 should render the picture in same speed?

[^Lele^]

The images posted above show an average noise level which is over double for the 3.x than it is for Next.
Both are ofc below the visible threshold, but that doesn't mean there are no differences.

In fact, once one takes into account noise levels, and normalises time, the emerging scenario is rather different.

In the linked contact sheet, you'll want to check the value of average noise level, in the middle row of values, the last of four (V).
0.001047 Vs. 0.002697 is a difference of 2.57X, while the rendertimes were slower by a factor of 2.04X, making V-Ray Next somewhere around 25% faster for this scene (hello marketing claims.)

Notice that doing these kind of assessments with integer images is bad, and it's made worse by using Photoshop.
You'll want Nuke, Fusion, or equivalent FP imaging tool, with FP imagery, to make sense of this stuff.

the contact sheet is here.

p.s.: for the scene you posted, instead, the noise is visually much lower for Next/v5. An A/B in the VFB shows it even without need for post measurements.

[vlado]

Did a lot changed from 3.6 to 3.7?

No, there are no major changes. It's mostly just a recompile for 3ds Max 2020.

Best regards,
Vlado

[kosso_olli]

p.s.: for the scene you posted, instead, the noise is visually much lower for Next/v5. An A/B in the VFB shows it even without need for post measurements.

That is what I wanted to express in this post: https://forums.chaosgroup.com/forum/...40#post1072640

People focus too much on a certain threshold number, without actually comparing the noise distribution. In nearly all our scenes, Next/5 shows a much cleaner image for a given noise threshold than the same threshold in 3.6/3.7

[Karonte]

Ok guys, we are not all tech people...What's your suggestion Lele, for example, to quick reach more or less the same" render time" and the "same amount of noise" if I open the old scene? you suggest to increase the noise threshold? or what?