I’m testing my single (soon to be dual) Xeon 2696 v3 and i’m amazed at how fast Corona Interactive is comparing to Vray RT (cpu). Vray RT crashes at times and is not really usable in production to test the lighting assuming we keep materials and don’t use material override. No crashes on Corona.
I’m curious as to why can’t vray reach the level of speed of Corona Interactive? Production Render times are almost equal to me between the two since i still use IM+LC on vray. The differences of interactivity between them are really really noticeable though.
VrayRT was not a planned feature and was tacked on when it first started development. The main goal of VrayRT was clearly defined as GPU acceleration, CPU functionality was meant primarily as debug and a legacy function. The main problem was the two different functionalities. Vray used exclusively render buckets, which was not compatible with an interactive progressive refinement idea.
So they started deving it, which ment for the longest of times, Vray RT couldn’t do half of Vrays materials and features. Now VrayRT is a mature interactive GPU preview system.
Corona did stuff differently. It was planned from the get go to have an interactive functionality and progressive render nature only helped cement that. Basically, Corona IR is almost 100% the same as normal render, the only difference being, that AAvsGI is something super low and thus sampled faster, but doesn’t cast rays as efficiently and auto updates. That’s it, no real difference.
So this is basically the reason, no code was changed for CoronaIR, while Vray had to reinvent itself. (Which they did a very good job on the GPU side, for such a long running product)
Corona gains much speed by being built from the ground up on Embree, but loses a little bit, by always having Reflective caustics on, while reflective caustics are Default off in Vray and have to be enabled. So a comparison is not really possible. Only thing that really matter in the end is preference.
Thank you for the explanation SairesArt! As far as i know Vray 3.3 changed that. Reflective caustics are now on by default but still have the ability to be turned off when needed.
Do you use them in production ? I really wonder how well they work since they’re by default ON. (By I presume they get curbed by MIS as well).
Reflective GI caustics are my wet dream. It’s the ultimate holy grail of lighting that almost no renderer can tackle : / I mean, most can’t even tackle refractive, Corona included, but GI caustics, that’s what I want to see.
Very rarely i used reflective caustics in commissioned jobs that i can remember. I only recently started fiddling with the (now outdated…) vray 3.3 due to lack of time in between projects. New variance-based adaptive sampler is indeed faster and much more simple but i haven’t tried any scene with heavy caustics to be honest. Anyways nothing beats Corona Interactive… And this is with only one of my Xeon’s 2696 v3. I can’t imagine when the other cpu arrives. It’s realtime!
quick sidenote, reflective and refractive caustics are the same from the point of the view of algorithms. An algorithm that works on reflective ones will work on refractive and vice versa. Most differences in handling are observed between directly visible caustics and caustics visible in mirror/behind glass
Well, it seems i’ve had it with vray. Lightcache takes ages in most of the scenes with my 2x 2696 v3 due to a +48cores (known) bug in vray… Why i’m not hearing that same bug from other users beats me.
Interactive rendering in Corona is lighting fast in complete interior scenes where vray rt can’t simply manage. Not to mention the underexposure vray rt frequently gives me in interiors.
Ondra, is there any way to get interactive even faster? I generally set precision to 0,01 to improve feedback when testing general light and materials and then set it back to the default 1,0 before final rendering. Is that the only value we can crank down to improve interactivity?
I’m not compaining Rawalanche It’s damn fast even with default precision set to 1,0. It’s even faster with precision at the minimum of 0,01. I was just wondering if there was some other tweak i could make that could improve interactivity even further. I’m still on Max 2014 and i always use CoronaCameraMod so the max physical camera bug is not something i’ve come across. Good to know, btw. Thanks!
Excellent pointers Rawalanche! Thank you! Changing the secondary solver to PT definitely improves IR interactivity albeit taking longer to clear noise. It’s probably better than messing with UHD precision. To be tested…
Ondra, that’s VERY nice to hear! I’m loving Corona even more!
Thanks for the explanation SairesArt! As far as I know Vray 3.3 changes are now reflected caustics by default. It also has the ability to take off when needed.
So what is the state of caustics, reflective and refractive in the latest version of Corona for 3ds max? Are they available in the progressive production renderer or not?
I assume you know that reflective caustics in vray only works with BF+LC and works a lot better in BF+BF.
IT does not work with IM+LC. When you press render in IM +LC look at the log, it will say reflective caustics will be turned off or will not work something like that. Basically you have to use BF for it to work.
Just to let you know And Vray BF+LC is slower than corona. Tested
Are usually converging very slowly
There are no optimizations made for caustics rendering yet, so rendering them could be tricky, but the answer to your question is yes.
The question about the reflective/refractive caustics was with regard to Corona 1.5 (or 1.6 if there are improvements to caustics behavior/performance in that version).
If caustics are clamped by MSI, does a value of higher than the default of 20 make them more visible (and/or converge faster)? I don’t think that the MSI gets changed very often by users, in fact the tooltip advises against changing it if I remember correctly.