Maxwell is amazing, the quality of light is impressive, thanks a lot for those tests Mihai
From: Mihai Iliuta Sent: Saturday, February 16, 2013 8:05 AM To: softimage@listproc.autodesk.com Subject: Re: Octane render Hi there, Hopefully this will be in the right thread....here are some classroom tests with Maxwell Render. Ronald, I just took your scene and applied Maxwell materials trying to match the look from your tests and for the lighting I just used Skydome. My machine is an i7 2600K @ 3.4GhZ, and according to our benchwell test (www.maxwellrender.com/benchwell) your machine should run it in about 10 min, mine took 14min, so 40% slower. Time 15min: https://www.dropbox.com/s/2uw6akqot5yun4d/15min.jpg 45 min: https://www.dropbox.com/s/110ec8qwea2dl39/45min.jpg 2h: https://www.dropbox.com/s/d3ty0h696nxwbuh/2h.jpg 7h: https://www.dropbox.com/s/yqdclhiwj36a9b6/7h.jpg The last one was to try and match your 45min Octane render, although some areas looked cleaner sooner (DOF areas especially). It's true that it takes a much longer time to get a super clean image, but usually a small amount of denoise is enough instead of waiting... So to compare the 7h render would take about 5 hours on your machine. the 2h render matches in noise your Arnold 1.33min render and would take about 1h20min on your machine. For the 7h render I turned up the intensity of the ceiling lights using Multilight to match your render (but they would never look that strong in reality if you expose for daylight, unless you want the electric company to come arrest you...). Quality wise, perhaps in my not so unbiased opinion, there is more life in the Maxwell render, mainly due to not limiting bounces. I don't understand what the point is to have only 2-3 bounces of GI because it starts looking dead and like an AO pass - greyness everywhere. Look especially in the chrome material, radiator area, in between the wood boards on the desk. All these things add up in my opinion and other tests will show it better. I prefer to know I'll get a render that looks and feel like a photo, not a render that looks like yet another....render. Also most of our users will take up 2GB of RAM just for loading their textures, no matter the resolution of the render and many want to render at at least A4 print size which is about 3500px wide at 300dpi. Finally, maybe we should maybe do a test with thousands of particles, hair, DOF, MB, sharp caustics??? Bring it ON!!!
<<wlEmoticon-smile[1].png>>
<<330.gif>>
