iRay Interior Tests

That's typical of MLT type renderers. The usual problem of needing many random samples is exacerbated by the fact that all light is gathered and... well, you already know this. Maxwell has the same problem.

FWIW it's more time consuming for iray to clear noise from an architectural interior scene like this that is illuminated by only a couple of small light sources. I changed HDR's on one of the scenes to an evening shot and rendered it again. Without much exterior illumination the two small light sources require far longer render times to clear noise. Of course you can scale the image down and run it through some noise reducing filters to help reduce the noise but it's still going to take longer to render than an interior scene that has larger light sources. This ran for six hours...So, for architectural interior scenes illuminated by a few small light sources I wouldn't expect fast render times out of iray. Adjusted image (scaled down, exposure increase + CA + slight color mod) [ATTACH=CONFIG]42924[/ATTACH] Original render (only converted to .jpg) [ATTACH=CONFIG]42925[/ATTACH]

have you noticed a significant different in the reflectance of glossy reflections when switching from MR to iRay. I have found iRay reflections tend to be shinier and more reflective. Or perhaps it due to being more sensitive / accurate regarding IOR?

Yes, I also noticed that. I've had to adjust a big chunk of my existing material libraries accordingly. I also figured it was due to the different lighting & calculations and probably my own miscalculations on the original material settings...I don't really know the IOR of a wood plank, etc., so I would typically use whatever looked right in my mental ray scene.

What I'm really trying to get at is the cost/benefit on this CUDA stuff - are you guys finding that it's at the point where it makes sense yet to use for still frame renders, where does one hit the limitations? Could you have rendered that at 3200 or 4800 pixels or does it overrun the memory?

With these scenes I had plenty of room left on the GPU's to increase the render output. I'm sure I could have pushed them into the 6k range easily, but larger sizes take longer to refine the noise and since these were tests I kept them fairly small. FWIW I hit the memory wall pretty early on with iray when my primary client changed my render output size requirements to 6k+. I had to switch to the 6GB GPU's in order to compensate for the larger frame buffer. My scenes typically range in the 10 to 15 million poly range and at 6k resolution with a few texture maps/HDR environment that usually weighs in around 4.2 to 4.8 GB range. So I still have a little headroom for larger frame buffers and/or more complex scenes. Is it worth it so far? For me, yes. When I first heard about iray I figured I wouldn't like it because of the lack of things to adjust/tweak that I'm so accustomed to having. Plus if I'm being totally honest I was a bit worried that the age of push button rendering was arriving and I should probably be a bit worried. I've been pleasantly shocked at how much time iray has saved me by NOT having to deal with those tweaks/adjustments. It sounds like my iray experiences have been similar to Samuel's experiences with Maxwell in terms of scene/material configuration time savings. It's just been a real surprise for me so far.

Yikes. That did come out more x vs y than intended. What I'm really trying to get at is the cost/benefit on this CUDA stuff - are you guys finding that it's at the point where it makes sense yet to use for still frame renders, where does one hit the limitations? Could you have rendered that at 3200 or 4800 pixels or does it overrun the memory?

Nice ones Jeff! As a fellow Mental Ray user I think Jeff makes an excellent point about considering setup/config/optimization time when making comparisons. If you start paying attention to the time spent there it really makes unbiased render times not such a big deal. I have been a Maxwell user since the original beta and from day 1 realized you spend much less time testing, tweaking etc. things just work. I imagine iRay to be much the same way. The fact it allows most MR shaders etc. means those of us familiar with Mental Ray don't have to re-learn a thing and can switch over to biased, optimize & tweak mode as needed without reconfiguring our scene. That's pretty sweet. On another note: Jeff, have you noticed a significant different in the reflectance of glossy reflections when switching from MR to iRay. I have found iRay reflections tend to be shinier and more reflective. Or perhaps it due to being more sensitive / accurate regarding IOR? Edit:Just noticed your edit. Right, agreed.

Those of you who have had a chance to work with iRay, do you see it replacing mental ray in the future?

There's many things that iray can't do at this point, like motion blur. So no, I don't see it replacing mental ray anytime soon.

So you used $10k of hardware to render screen res shots in 40 minutes? They're very nice but I'm not 100% excited yet. Couldn't you do the same thing faster or with less hardware in regular mental ray?

I get your point but no, I mainly use this hardware for my daily work with iray...not just for these tests. I actually don't work in the architectural-vis field so these are more so personal tests that I ran in order to help me provide accurate answers to the arch-vis people that send me questions about mental ray and now iray. EDIT: I removed the rest of my reply to help prevent this thread from turning into one of those render X vs Y threads. I'll be happy to try and answer any questions I can on iray or hardware but perhaps it would be better to do so in the new GPU Rendering Engines forum.

I'm anxious to see iRay. We haven't upgraded to Max 2012 yet. I'm still trying to get my head around mental ray settings. Those of you who have had a chance to work with iRay, do you see it replacing mental ray in the future? I wish technology would slow down just a tad. It's more than a full time job trying to learn software, much less stay on top of upgrades and new realeases. MAX seems to be a never ending mountain climb, and now I'm also having to spend a significant amount of time staying on top of Revit.

So you used $10k of hardware to render screen res shots in 40 minutes? They're very nice but I'm not 100% excited yet. Couldn't you do the same thing faster or with less hardware in regular mental ray?

is iRay viable for resolutions above 5k+ yet?

I've rendered several 6k+ images with iray so I'd say yes. Of course that is provided the GPU has enough memory to support the scene and the large frame buffer.

what resolution were these rendered at?

Just personal tests so they are relatively small at 1600 pixels original res.

is iRay viable for resolutions above 5k+ yet? what resolution were these rendered at?

Nice, nice work, Jeff! You have amazing talent dude! So are you liking the Iray stuff versus the old way of rendering?

Hey! Thank you sir. Yeah I'm digging iRay these days but still use mr/VRay for some things because it can't do everything I need yet.

Was there a rendering time test done without DOF? What was the image size rendered?

No, DOF doesn't really add much to the render time with these type rendering applications as it would with something like mental ray. Typically if the grain/noise level is where you want, the DOF will be be smooth as well at the same time.

Nice, nice work, Jeff! You have amazing talent dude! So are you liking the Iray stuff versus the old way of rendering?

No, DOF was rendered.

Was there a rendering time test done without DOF? What was the image size rendered?

The DOF is great. and materials and lighting - nice work

No, DOF was rendered.

Was the DOF effect done in post?