They are, actually.
Best regards,
Vlado
RedShift I can sort of see, but Octane would be odd… anyways, I hear you - will make a note to add that.
Best regards,
Vlado
Okay, great. As I said - this is not my opinion. But when people without a technical knowledge throw buzzwords at me I’d love to say something like “Yes, we can technically do that. Let’s decide if that’s the best for your project.”
And obviously even many people on this forum, who use vray every day, do not know how those things (GPU/CPU/Progr./…) are hooked up in VRay.
But thanks for looking into this! I think the possibility to render final frames on the GPU directly out of VRay for Maya would be a huge benefit, on both production and marketing side.
What about : rendering ---> render ---> render Animation with VRay RT ? (i don’t know if it is in the stable build)
Btw, in my current studio i have developed a job submitter for Deadline where you can choose Maya Batch/VRay standalone/VRay standalone RT.
Maya batch : render your scene in Maya batch
Standalone : translate your Maya scene into vrscene, render vrscene, delete vrscene
RT : translate your Maya scene into vrscene, render vrscene in RT mode, delete vrscene
When you start Maya you can choose what build of VRay you want to start. All VRay builds are on a network folder. The “starter” sets all environment to choose the “good” VRay.
Now we can add or remote render nodes easily without to install any VRay or other plugins.
This doesn’t work through Maya batch though. It can be done by exporting a .vrscene and rendering through the standalone, like you have done, but not everyone will want to do it.
Best regards,
Vlado
Yeah, we can of course write a submitter. I think renderpal already supports vray standalone submission. But this should really be implemented by chaos to make it available for everyone.
We also use the nightlies with some batch scripts for creating environment variables and stuff and this is really nice. But this doesn’t help with the gpu render in batch.
Corona render is also coming along for maya. It’s looking better and better in the maya beta, and I think most know what it’s capable of for Max. When it comes to progressive rendering, unfortunately I don’t think Vray balances things particularly well. Not sure technically what it is, if its the balance between direct and indirect light sampling or just the noise patterns, the way they are cleared?! Also would LOVE for the noise threshold thing to make more sense. I find it really hard to figure out why a noise threshold of 0.005 in adaptive takes X time, and using the same value in progressive takes Y time and looks different too. If they are different, change it to something I don’t have to trial and error so much.
The reason why renders like Corona and Octane are used more and more, is because they cut away the test and tweak on light subdivs etc… they make it simple, while still being just as a fast as vray. (corona at least).
I love using vray, and the strength of vray is the feature list, and the fact that it’s used in a lot of industries and places. Sorry if this is a little off topic. Just keep an eye out, because I have worked at 2-3 different companies this year already that has switched to corona full time over vray for arch viz.
Can you show me? In all the tests that I’ve made the two samplers work identically with respect to the noise threshold and produce similar results for similar render times with the latest builds.
Just keep an eye out, because I have worked at 2-3 different companies this year already that has switched to corona full time over vray for arch viz.I am keeping an eye out
Corona is a good renderer and for many people it will make more sense to use it. I don’t think at this point there’s much that I can do about it. Speedwise, V-Ray is capable of holding its own (still lots of room for improvement of course), but even if I remove all V-Ray settings tomorrow and make it as simple to use as it gets, you will have two render engines that are roughly equal. What would be V-Ray’s advantage then? A few more features that might be able to hold it afloat for while longer, but that’s it; then things will boil down to price. For this reason, I don’t think going for simple would be the best choice, even if that means some users will move to other solutions (that happens anyway, not least because of the fact that people like change). Of course, making V-Ray more robust so that there is less need to tweak settings is something very desirable - I think we’re not in a bad place even now, although things can certainly be improved and we are working on it. But, being able to tweak V-Ray can be an advantage in situations where other engines fail, even if those situations are relatively rare.
This is somewhat of a general trend though, as rendering gets “commoditized”. If you look at Arnold, RenderMan (RIS) and mental ray in Maya 2016, they look nearly identical, with identical controls. Sure, some of them handle some situations better than the others, and under the hood there are more differences, and each has different tools for some specific situations, but there are less and less reasons to prefer one over the others for the majority of cases.
Best regards,
Vlado
First of all, thank you for your response Vlado! Secondly I totally agree with you.
The trend right now is to make things simpler, the “make pretty buttons” have finally arrived (sort of). And I’m not saying Vray should fold and change what it is to try and accommodate, but the interesting part for me is WHY do all the big old renders do this. Arnold, mental ray, and even renderman must feel that it’s the right thing to do, or that they need to do it. There must be a reason for this? My guess is that in all industries the infamous turn around times are always getting shorter, with the same or higher quality demands. So when render engines gives you a “no-brainer” solution where you don’t have to sacrifice artistic time to meet the deadlines, because you have to think and tweak settings based on your shot, people take it.
And yes, there are a ton of settings in Vray that, if you just left them alone, you would probably be totally fine, but I think that people have a tendency to feel like they need to tweak.. just for the sake of it. There are so many options and I sometimes feel like I need to tweak something.. but what particularly takes time to figure out, so you end up just adjusting all the “usual” things a bit, probably, often without really having had to. But rendering, to then see something is wrong or too noisy, adjust… rendering again… just takes a lot of time. What the other render engines are doing is giving you no choice but to trust that the settings you can’t access are the right ones. Then it purely becomes a matter of time, but at least you know exactly what you get after 15mins.. 30min and 1hour, because there are no variables.
I do think though, that Vray is in a great spot, and I am guessing that the upcoming update (service pack?) of Vray for maya will look somewhat similar to the great update we got for max recently? (SP1). I may be wrong, I only very recently partly switched from max to maya so I’m not 100% up to date on the version numbers and if their compare. But the SP1 for max was a great release!
I will try and setup a test for you, either to show you or to prove that I was wrong 
So I just ran a super quick test, where I put the noise threshold at 0.01 in the global dmc using adaptive. Afterwards I used progressive and put in 0.01 and the two renders look very close. 3m 16s for adaptive and 3m 25s for progressive. But what I think throws me off is the fact that when using progressive you have access to both the progressive noise threshold AND the global DMC… do they need to be the same? what happens if they are not?
I have attached (tried to at least) the settings I am talking about.
One threshold is for primary image rays (the one in the image sampler). The other one is for secondary effects (GI/reflections/lights etc).
I will eventually move all but the “Lock noise pattern” options in the DMC sampler into an “Expert” section.
Best regards,
Vlado
So how do I go about getting the same look as when using adaptive?
If I leave the image sample threshold at default, so the DMC threshold is the same between the two, is that all I need?
or should both be the same? like how do I know without rendering several times, isolating layers of reflection, GI, light sources etc to try and guess? and how do I know when it’s actually the light (subdivs) itself? and not the sampler threshold?
I understand the two controls, but I have no feel for what the values actually mean. like if I put global dmc at 0.01 and image sampler at 0.0001 do I get nicer direct light? and noisy secondary? (outside of the light subdivs) and what happens if I swap the two? of course I can just test this, but from a technical point what is meant to happen? or what’s the intended use for these options?
update:
Or is half my problems with noise really more about BF/LC settings? I usually use 35-60 BF depending on how much time I have, and 1800-2500 LC. But i’m not sure if my global noise really should just be 0.01 and my BF should be at 120.
+1 for Vlado
Personally, i like to have a lot of settings. I don’t want a “toy” where when you have a problem, you can not solve the problem because you do not have the setting for this (example in Arnold few year ago : i have added DOF on my scene. The DOF had noise. To remove noise, i had to oversample all (SSS, reflection, all…). Normally i had to oversample the AA, but the option was not here. Render time jump from 15min to 2h. I was few years ago, perhaps now this problem was fixed).
I prefer to see VRay with some “advanced” users than “no brain” users.
The solution would be to hide a lot of menu for “no brain” users, and keep an advanced mode.
In my current studio for the render setting we use the easy ways :
- subdivs mult to 0
- ADMC 1 - 10
- min shading rate to 10
- DMC adaptive amount to 0.9
No more, no less, work with 90% of scenes, and we do not need to tweak any materials/lights
If we have some noise, we set ADMC to 1-20. Another noise ? 1-30…
This is no brain, and it is working.
The only problems than we see with this settings is :
- IR map doesn’t work with this. Perhaps you have to ignore IR map when subdivs mult is to 0.
- if we use DomeLight or DirtNode, we have more noise by default (with ADMC to 1-10). Perhaps you have to add more subdivs by default when subdivs mult is to 0 for DomeLight or DirtNode (and i think there is another nodes).
After this, why you do not add this “settings” for “no brain” users ?
Yeah the no irradiance map is kind of a drag
So bigbossfr you prefere to have global subdivs off? in some of my scenes I accidentally used this option and got surprisingly better render time! But as you said it does not work with irmap…so Kinda limiting.
You would generally keep the “Global DMC” noise threshold the same for both samplers (f.e. 0.005 or something like that); and then make sure that the progressive sampler “noise threshold” is the same as the adaptive sampler “color threshold” and that the min/max subdivs match. Here is a render done with progressive and adaptive samplers, along with the respective sample rate elements - you can see that render times and noise are similar.
Best regards,
Vlado
Yeah, someone already requested that the subdivs==0 should not affect the irradiance map.
- if we use DomeLight or DirtNode, we have more noise by default (with ADMC to 1-10). Perhaps you have to add more subdivs by default when subdivs mult is to 0 for DomeLight or DirtNode (and i think there is another nodes).For VRayDirt, this could work, yes. In hindsight, making the VRayDirt dependent on the AA subdivs was a big mistake as it caused a lot of trouble. Hopefully not too late to fix it
Best regards,
Vlado
Vlado, why the progressive sampler took longer to finish? The difference, if projected to a complex scene with longer render time, is quite big. Is it just about the progressive is relative new and still being optimized?
Thanks for testing Vlado, I do believe you, it’s more about understanding the settings, since especially the thresholds 0.1 up or down can make huuuge difference in quality and render time.
The solution would be to hide a lot of menu for “no brain” users, and keep an advanced mode.
In my current studio for the render setting we use the easy ways :
- subdivs mult to 0
- ADMC 1 - 10
- min shading rate to 10
- DMC adaptive amount to 0.9
No more, no less, work with 90% of scenes, and we do not need to tweak any materials/lights
So the global noise and the DMC noise are both put to 0.0 ? o.O
Because of the intermediate image updates. Showing the intermediate image, and keeping something ready to show you at any moment, requires extra resources and CPU time. F.e. color mapping and gamma correction need to be performed for every update; this is a small thing but across many pixels it adds up. The adaptive sampler needs to do this just once after a bucket is complete; the progressive one needs to do it for every pass through the image (there were about 1600 of them for this render).
The difference, if projected to a complex scene with longer render time, is quite big.This is not related to scene complexity at all. Only image-related properties have an effect on this (i.e. resolution, number of render elements and number of passes through the image). So for more complicated scenes, this additional time will be actually still the same and will be a smaller portion of the total render time.
Is it just about the progressive is relative new and still being optimized?There are some optimizations that can be done for sure, but there is just a bit more work to do generally.
Best regards,
Vlado
IR map : yes, i think the fix can be quick (if 0 blabla). If you have some time for a nightly for this, just tell me.
This is never late to fix ;).
What about the DomeLight ?
I don’t know why but by default with a domeLight we have a lot of noise. So, if i use the technique of subdivs mult to 0, all renderElement are ok (no noise in reflection. refraction, GI…) but the lighting render element has a lot of noise.
I have to add more AA max subdivs to have a clean result, but it looks like i will over sample other render elements (reflection, GI…) for nothing (because of the Domelight).
It is not possible to add more sample to domelight by default ?
Or perhaps do not affect domelight with subdivs mult to 0 ?
We are not far from an automated way…