If this is your first visit, be sure to
check out the FAQ by clicking the
link above. You may have to register
before you can post: click the register link above to proceed. To start viewing messages,
select the forum that you want to visit from the selection below.
New! You can now log in to the forums with your chaos.com account as well as your forum account.
Announcement
Collapse
No announcement yet.
V-Ray Render Optimization - an in-depth Guide (call for Before/Afters)
If the noise threshold controls when the sampler stops adding additional rays.. what's the reason not to max out everything, max subdivs, max DMC sampler, and then use the noise threshold to limit the quality?
Thanks for taking the time to write this, a great reference to come back to. Can I ask, should you always do your testing at the full resolution you are planning to render at
No, there's no need to do tests at full res - you can optimize your scene at a lower resolution, and then once you've got it tuned, the speed/quality gains should scale accordingly as you go back to full-res. Just be careful you don't go TOO low res for testing, or V-Ray may have a harder time trying to resolve smaller details in the scene.
If the noise threshold controls when the sampler stops adding additional rays.. what's the reason not to max out everything, max subdivs, max DMC sampler, and then use the noise threshold to limit the quality?
hmm.
Well you've kinda just described the 'Universal V-Ray Settings'. And don't get me wrong, the universal settings have their place and work just fine for some scenes... but the reason why it isn't always the best choice is what the tutorial is all about: Basically the higher you set your AA settings, the lower all the other settings in your scene becomes (because of the way V-Ray balances the Image Sampler and DMC Sampler internally) - which can lead to noisier and slower renders than if you were to balance the Image Sampler and DMC Sampler manually according to the demands of the scene. Makes sense?
Thank you RokinAkin, really useful tutorial u realized. I was searching for something like this, to clarify as well as possible the way vray works. Especially for me with my bad english, a direct and specifically thinked tutorial is ver very useful.
Thanks a lot
Great write-up, and perfect as a sanity check for the brain. Mixed feelings though, as I had really hoped I would learn something new that would make my life easier, but happy that I haven't been completely wrong and stupid.
One aspect I would really like to see more info on in regards to this, is the "adaptive amount" value, how it works, and if it can be usefully tweaked to further optimize.
Also, could one just use the "Global Subdivs Multiplier" to up all the secondary samples to a certain point for the optimization stage, and then just slightly tweak the individual samples for lights and materials as needed afterwards?
I followed your procedure with a similar scene setup accompanied by one of my high resolution, sun-light hdri (which often is a challenge in and of itself regarding noise) to see how much faster I could get the renders while remaining the quality and level of noise. Bear in mind, I chose to allow for some noise to keep the renders around 5 minutes:
Top: Universal settings
Bottom: Optimized
As you can see, the noise are fairly identical in both renders, but the first render took 5.6x longer. That is just crazy.
Edit: Thumbnailed the image. Click on it for full size.
I followed your procedure with a similar scene setup accompanied by one of my high resolution, sun-light hdri (which often is a challenge in and of itself regarding noise) to see how much faster I could get the renders while remaining the quality and level of noise. Bear in mind, I chose to allow for some noise to keep the renders around 5 minutes:
Top: Universal settings
Bottom: Optimized
As you can see, the noise are fairly identical in both renders, but the first render took 5.6x longer. That is just crazy.
That is awesome! One of the biggest speed gains I've seen yet!
I'd be really interested to see if that gap in render time grew even more if you lowered the Noise amounts to really high quality settings.
Would it be alright to share your results in the tutorial? If so just shoot me an email - Akin@CGGallery.com
One aspect I would really like to see more info on in regards to this, is the "adaptive amount" value, how it works, and if it can be usefully tweaked to further optimize.
Also, could one just use the "Global Subdivs Multiplier" to up all the secondary samples to a certain point for the optimization stage, and then just slightly tweak the individual samples for lights and materials as needed afterwards?
I've got some thoughts regarding these, but I need to do a bit more testing before I can say anything about it for sure. Soon though!
Nice article Akin.
Between your writing and the video Grant made, I'm finally starting to feel a bit in control of vray.
My comparison is from my graduation film, which I renders using the universal settings. Top image is universal at 6m 58s and bottom is optimized at 2m 39s
That is awesome! One of the biggest speed gains I've seen yet!
I'd be really interested to see if that gap in render time grew even more if you lowered the Noise amounts to really high quality settings.
This I just had to try. And the results were pretty astonishing:
Vray Universal Settings: Click on image to see full resolution Image sampler: 1 min & 100 max Subdivs Lights, GI, and Materials: 8 Subdivs Noise Threshold: 0.001 Render time: 3h51m50s
Vray Optimized settings: Click on image to see full resolution Image sampler: 1 min & 32 max Subdivs Lights, GI: 128 Subdivs Materials: 164 Subdivs Noise Threshold: 0.001 Clr Threshold: 0.004 Render time: 18m3s
This I just had to try. And the results were pretty astonishing:
Vray Universal Settings: Click on image to see full resolution Image sampler: 1 min & 100 max Subdivs Lights, GI, and Materials: 8 Subdivs Noise Threshold: 0.001 Render time: 3h51m50s
Vray Optimized settings: Click on image to see full resolution Image sampler: 1 min & 32 max Subdivs Lights, GI: 128 Subdivs Materials: 164 Subdivs Noise Threshold: 0.001 Clr Threshold: 0.004 Render time: 18m3s
That's a 12.8x speed boost.
Can you post the sample rate pass? Are you hitting max 32 AA samples anywhere in the image? It seems like heavy overkill as there is no noisy textures.
Ivor 88 - Sweet results! Thanks for sharing! Jorq1H - While that speed increase is ASTOUNDING... I agree with Morbid Angel that we'd need to fix the noise in the teapot's highlight before calling it a success. But luckily we know there's a TON of room for improving the optimized render given the difference in render times. Even so, that's totally nuts.
Here's an animated gif I've put together to illustrate how allocating more samples to the DMC Sampler reduces the stress on the Image Sampler.
The Image Sampler's Min & Max Subdivs stay the same, while the scene's Lights / GI / and Materials Subdivs rise... resulting in less Primary Samples (AA) being needed while quality improves.
Ivor 88 - Sweet results! Thanks for sharing! Jorq1H - While that speed increase is ASTOUNDING... I agree with Morbid Angel that we'd need to fix the noise in the teapot's highlight before calling it a success. But luckily we know there's a TON of room for improving the optimized render given the difference in render times. Even so, that's totally nuts.
Here's an animated gif I've put together to illustrate how allocating more samples to the DMC Sampler reduces the stress on the Image Sampler.
The Image Sampler's Min & Max Subdivs stay the same, while the scene's Lights / GI / and Materials Subdivs rise... resulting in less Primary Samples (AA) being needed while quality improves.
(This scene generously provided by Peter Guthrie)
Akin, could you post the render times together with these images? Are they dropping as the subdivs are increased?
Comment