Mike A

Where that A6000 /:48GB memory comes in useful is with 'Disney' level rendering to handle massive datasets. Eg: https://www.disneyanimation.com/resources/moana-island-scene/ and that example, as you can see, is 6 years old. Or not for rendering at all, but for number crunching in simulations, mathematical modelling, AI, etc.

It will be interesting to see what they deliver for the desktop. The market could do with some more competition.

...and of course that's on the basis that you are rendering all the time at 100% power - which you won't be - by a long way - as a freelancer / solo operator . For a solo user I think the A6000 payback in terms of power cost is... never : )

I'll jump in to this conversation guys as I've been through the same deliberations. I seriously considered the A5000 too. For Redshift the CUDA core count is the main factor that determines rendering speed. This makes the 3090 about 25% faster than the A5000. On the positive side the A5000 is 230 watts vs 350 watts for the 3090, and it's a two slot blower. On the negative, its a bit more expensive. Fortunately I managed to obtain two of the 'end of line' Gigabyte 3090 two slot blowers. So my plan is to run those at considerably reduced power, maybe 70% or so, for everyday less demanding work. That will give me reduced electricity costs - and noise : ). But it will also enable me to ramp up the power when I need to. I'm hoping to get the best of both : )

I've got a big client who is Maya based and I have wondered about getting Maya indie to aid file transfer etc. Does Indie work with the same format as the full commercial version? In other words - could I open my clients file in Indie?

I did listen : ) It's absolutely true that this method takes a lot of AI training and is scene specific - and is not practical for general 3D work currently, but this sort of tech is on its way...

A couple of things are screwing you up here: 1. your alpha texture is 'alpha1.jpg' - but your file path is 'alpha1.png' : ) 2. You need 'Reflect / Refract IDs' enabled for this sort of reflection / refraction type effect to show in a puzzle matte. Using reflect/refract IDs means that the floor reflection will also be in the puzzle matte. It that's a problem you could take an alternative route: 1. Add an Rs 'Store color to AOV' node between the material and output nodes. Connect material output to Beauty input. 2. In that node set the AOV input 0 colour to white. Set AOV Name 0 to 'Add new custom AOV' and give it a name of your choice eg: 'ripped box'. 3. You'll now have a 'ripped box' AOV in the render settings AOV manager. In the settings for that AOV set 'secondary ray visibility' to Refraction/Transparency. This will give you a new 'clean' AOV of your ripped box : )

I'm really not sure who will buy this over a standard 3090 - for GPU rendering at least. 2.5% more CUDA cores at the cost of an extra 100 watts of power. Am I missing something?

Zsolnai-Fehér's 'two minute papers' are always an interesting watch. I'm sure that real time rendering at the quality of current 'off line' rendering will be with us sooner than we expect.

Welcome Theob!

Welcome Turbolello ! I'm pleased to report we're a household treating our pasta correctly LOL : )

If you're using a Dome light with a Texture / HDRI in it, that will bock the alpha by default. Select 'Environment > Replace Alpha Channel' in the Dome light settings. If this is not the issue uploading your scene here will enable us to help.

I'm sorry to say: you can't. The Dome light is not currently open to colour (RGB) inputs, only texture inputs ('RSFile'), due to 'performance implications'. The Rs team understand that users want this and they do have it on their long distance radar. Currently the only external input you can make is via a C4D shader node as shown below. Tex 0 is the dome light, Tex 1 the backplate. Tex 0 generally not a lot of use as any C4D non-image shader input won't be 32bit HDR. The only other thing to note is that Rs does support the use of the C4D physical sky if that might meet your needs.

The new Rs 'standard surface material' node is simply an update/upgrade to replace the existing Rs 'material' node and will work in the same fundamental way. It has better shading algorithms and a better structure. The result of that will be better looking materials that are easier to build.

Yes, there are some good Redshift improvements about to be delivered alongside R26. The new standard material, significant improvements in energy conservation and rough surface shading, thin film built in, improved material blending,,, and long awaited SSS improvements scheduled to arrive a few months later - fingers crossed. I'm looking forward to all of those : )

This is my favourite thread of the year. So much fun to review come mid April or so : )

The issue is more one of 'materials conversion' rather than capability. Yes - you can do 3D linear gradients with a bit of node work, but for a simple Y / height based gradient you can use the 'ray position' output of the Rs state node. I've attached very simple example. The Rs state node 'ray position' supplies the position of the shaded pixel in object or world space (node setting). The Y data is extracted with the vector to scalars node. The change range node 'old maximum' value is set to the height of the landscape - roughly 100 units in this case. The node then outputs a 0 to 1 gradient over that height. That gradient can be manipulated with a ramp node. There are lots of uses for this type of technique in Rs 🙂 Rs state landscape.c4d

I wouldn't be - you'll have more spare change in the bank this week : ) LOL! No matter how much you spend, or when you buy you can be absolutely certain of one thing: The Ferrari you just bought will be a Reliant Robin* next month. * Google it : )

I decided to go with 24GB / 3090's, although if there had been a suitable 16GB alternative I would have taken that. 12GB feels a little too tight for my work. Having said that, I don't think I'm in the middle of the distribution as a C4D user: visualisation work, high poly counts, high res textures, environments, displacements, volumes etc. They are all memory hungry. While I doubt I'll need it - for the foreseeable future at least - the option that does become available with the 3090's, but not with the lower spec cards, is that of NVlink - giving a potential pool of 48GB.

I'm just awaiting a new AM4 + 3090 delivery myself. It's a cost / performance compromise of course, but it seemed the best option for me. You pay a lot to climb up that last few percentage points on the performance charts... Have fun with your new kit : )

I suspect what's going on here is not post effects... Are you rendering in ACES colour space? Unfortunately C4D picture viewer is not ACES - or colour management in general - aware, so at the end of a render the colours can get screwed up if you have the wrong settings. If you are using EXR/ ACES rendering, and outputting via the C4D picture viewer, you should render with the following option off: Render settings > Redshift (Advanced) > Globals > Color Management > Compensate for View Transform = OFF If this is relevant - post here and I'll try and direct you to some more info.

To get refraction values in an alpha channel you need to enable alpha channel - and in the Redshift Render settings enable 'Refraction Affects Alpha Channel' under: Render Settings > Redshft > System (tab) > Legacy. ...but of course you won't have any refraction behind the glass in compositing, so it won't look right. That's why this option of 'off' by default.

Personally I'd be using a high res 32bit displacement and/or normal map for this, ie texturing, not modelling.

I would try the latest version 3.0.66. There have been a few bugs fixed since .62, maybe it's one of those causing your issue.