OTOY Forums

thanks mykola ..... normally i know windows quite well, i guess you don't work with support.

if i rightclick and use save as, then there is no problem - other than it says that the zip file is corrupt :-/

Thanks we will waiting the render elements,thank you again.

I had a chance to spend about an hour doing tests on the speed improvements of v2.1 vs v2.06 here are my results. The test were done using a fairly simple interior scene that a had few different shaders that were using diffuse, spec and bump images, no displacement, no motion blur just straight interior type renders. I only tested the path tracing kernel.

Step 1, Just loading the new version and setting path term. power to 0 (default was .3) and leaving coherent mode off I got about a 12.5% increase in speed over v2.06. Yes this was exciting news! It proves without any new assistances in features the kernel is faster.

Step 2, and I did not know this at the time I went for the extreme case. I set path term. power to 1 and coherent mode to be checked on. Wow this is rendering crazy fast, almost 80% faster. Something must be up, so I let the same image from step 1 get to the same sample level (4000) and then saved and directly compared. As I feared the result were not good. The image had much more grain at the same sample level. I would guess it would need to go to 6000-7000 samples to look like 4000 if that would even help. I didn't wait to see - what would be the point.

Step 3, was it this new coherent mode that caused this bad result in step 2? I shut off coherent mode and left path term. power at 1. Still rendering very fast about 68% faster but still looking grainy. In fact the amount of grain was just about the same as step 2. That sucks.

Step 4, ok so maybe coherent mode is not so bad then. next I turned on coherent mode and set path term. power to 0. This result is looking great other then the weird gradation at the start of the render similar to PMC this rendered about 19.5% faster then v2.06 and the amount of grain is the same, it is just in a different place then with out it checked. I assume this is from it not being a random path trace. For still renderings this is the way to go for sure. Not sure about animation?

Step 5, and my last test. I left coherent mode on and set path term. power back to the default .3 (obviously Karba knew something I didn't) Again another render increase over Step 4. although a small one this time. It brought the speed up to about 22.5% faster then v2.06. And comparing the final images it was hard to tell a difference in the graininess accept in the darkest shadows. This is where I will be keeping my settings for interiors.

Conclusion v2.1 is a great improvement in speed about 22.5% on the scene I tested without any difference in results of the final image. I am sure on certain scenes you may be able to crank up the path term. power closer to 1 and get even faster results but be careful with this number as it can have the opposite results as well.

Great job Karba and Otoy!!!!!! At 22.5% you just bought me a new Titan black and I don't even have to pay for the electric usage.

did you have any luck fixing this issue?

http://render.otoy.com/forum/viewtopic. ... 7&start=20

i could download the file here at home.

please delete my posts. and sorry for any disturbance.

Here are my tests :

Image 1 - Octane 2.06 - 9min45 - 5000 Samples/pixels
Image 2 - Octane 2.10 (few more seconds... sry !) - 10min38 - 1811 Samples/pixels
Image 3 - Octane 2.10 with Coherent mode (PathTerm 0.3)- 6 min - 1347 Samples/pixels
Image 4 - Octane 2.10 with Coherent mode - 9min45 (same as image 1 this time) - 2186 Samples/pixels

Thank you boys, that's a really nice release, the noise disappears insanely fast !

it's about 30% faster for me! great job Karba!

finally static noise on static objects Thank you!

coilbook wrote:finally static noise on static objects Thank you!

hmmm... does it mean that it will render static objects also faster ??

Still fails when trying to open a complex scene, anything over 3GB fails. Works fine in 1.x