Hi all,
I'm very happy to announce that the octane v1.0 beta 2.3 will not have MLT (Metropolis light transport).
I'm pretty sure you're all in shock reading this.
We have worked tirelessly to try to adapt MLT to a GPU, and we have decided to abandon our effort as the results are not up to par with what we had hoped many months ago.
However, we always had a backup plan that was co-developed alongside our MLT effort, and we have created a new algorithm that is superior to MLT, and will be delivered at the same timeframe as the MLT promise, eg beta 2.3 final.
Classic MLT on a traditional physically based renderer is in most practical renders, much slower than pure pathtracing. a current octane render of an interior, cooked for 30 minutes, with fireflies on it, will cook firefly free to the same level of noise in at least 2-3 hours with MLT. only in special circumstances does it actually give faster convergence (eg the noise dissapears faster), and these are usually interior scenes with very small access to light.
(information above is not octane specific, but applies to all CPU based unbiased engines out there too)
MLT for exteriors is not recommended, it will just focus on invisible caustics and you end up with 5x more rendertime for the same result, but without the fireflies.
Since the new an algorithm we've developed is not based upon any solid scientific paper/concept, it might take a few pre releases or RCs to get it working perfectly, but the few weeks more have advantages, as the new algorithm is somewhat similar to bidirectional path tracing without fireflies, eg you can place an emitter in a glass lightbulb, covered by a diffuse transmission cloth lampcrown, and have beautifull caustics on your walls.
I hope everyone understands that developing octane is a difficult effort, most of our stability problems, slow development and limitations are not due to octane being in beta phase, but simply the hardware platforms (GPUs) and the APIs they use (Nvidia Cuda for us) are young technologies, unpredictable, and are very sensitive to various things, including OS/cuda toolkit versions/many different GPUs and many different drivers.
Since we're not marketers but developers, we decided to inform our customers of this decision as soon as possible, after we decided to make the switch a few days ago, so all our customers know what to expect.
We haven't decided on a name for our new algorithm, but we'll get back to you on that one.
in a nutshell:
MLT: slower render time, no fireflies, still difficult with interiors with hidden lightsources or small openings, not recommended for exteriors
Our new Alternative: The same rendertime as pathtracing, no fireflies, very good for caustics and hidden lightsources, with an option for firefly free, same speed exterior rendering.
Yours,
Radiance