RenderMan Pro Server 不只快而且比以往更加真實。不但可以處理多種幾何形狀、更高效、更逼真的燈光和陰影，產生更卓越的成品，適合全世界的3D藝術工作者簡化工作流程是一款最棒的工具。

Faster Re-Rendering

Providing the fastest feedback possible for shaders and lighters is a key goal for RenderMan, so RPS 18 delivers Path Tracing for greatly enhanced interactive re-rendering, allowing lighters and shaders to focus on the artistry of their craft, and to hopefully get home by suppertime.
Neon created with Geometric Area Lights

Better Area Lights
RPS 18 allows you to turn any geometric primitive into a light source with Geometric Area Lights. These lights are not only highly realistic and capable of creating incredibly complicated lighting effects, but geometric area lights are much easier for lighters to control with a feature called "balanced sampling" ... and compared to standard area lights, they're simply faster.

Enhanced Volumes
RenderMan's ability to process volumes has been optimized with hybrid sampling techniques. Because volumes near the camera don't need to be sampled as much as other elements in the scene, the REYES hider has been leveraged to deliver outstanding speed-ups whenever you need to fly a camera through a few explosions, a tornado, or whatever volumetric hijinks a shot demands.

Better Subsurface Scattering
Creating realistic human skin is hard, and the quality of the final result really depends how the subsurface scattering is being calculated. With RPS 18 there's a better way to create realistic skin with Beam Diffusion Subsurface Scattering. This new model is only slightly more expensive than standard SSS, but it produces much more satisfying results (results similar to "quantized diffusion," but that are numerically stable for animation). Now your humans can look real, and not waxy. 

And Much More ...
With other significant improvements like Correlated Multi-Jittered Sampling and core optimizations to geometric primitives, RenderMan Pro Server 18 delivers amazing tools for VFX and animation. For the complete feature set please check out the RenderMan Pro Server 18 Release Notes.
More Efficient Polygons
Polygonal meshes have been overhauled, reducing memory requirements significantly. Depending on the scene, users can expect anywhere from a 4x to 10x overall memory reduction when rendering complex polygonal meshes. The new implementation ray traces complex polygonal meshes up to 2 to 3 times faster with displacement and 20% to 35% faster when there is no displacement.

Additional features include:

1. Correlated Multi-Jittered Sampling
2. New textureatlas() shadeop
3. Higher performance for instanced geometry
4. Improved statistics

Tractor 是一個網路工作分配制度，用來取代Alfred, which is now de-coupled from RenderMan Studio。

Tractor 提供了許多 Alfred具備的功能，但更加強大和可擴展性。

他並不侷限於只渲染 RenderMan，Tractor可以輕鬆地被用來從各種來源分配網路任務，成為一個中央生產工作分配系統。

Components

Tractor contains 3 key components:

• Tractor Engine
• Tractor Blade
• Tractor Dashboard

Tractor Engine
The engine is the core component of Tractor. It manages the task assignments across the network and keeps a log/database of the spooled jobs. Only one Tractor Engine is usually running on your network, and jobs can be spooled to it via python. There is a sub process of the Engine, called the Monitor. The Monitor handles the requests, generally from clients, to the web interface (Dashboard)

Tractor Blade
The blade is the remote execution server. It is a python based server which asks for tasks and jobs to be handed to it by the Tractor Engine. Unlike the previous alfred/maitre-d configuration, the blades themselves monitor their host's RAM, disk & CPU usage to check if it should be assigned tasks. If the host is heavily loaded with other tasks, then it will not participate in job acceptance. When a Blade asks for work from the Engine, the Engine matches the blade against a set of profiles and assigns it jobs if it matches the criteria in the profiles.
Tractor Dashboard
The Dashboard is the central web interface for your Tractor renderfarm. It lists all the jobs, blades and history of your current setup. Users can customize their own views, including window panes and other properties. This is also where control over existing spooled jobs can be achieved as well as viewing the status of remote execution blades. For more information, see the Tractor User Interace documentation.

Tractor Architecture

Tractor Spooling
Rendering to Tractor from within RenderMan Studio is extremely simple. There is a new switch on the render options, which you can toggle between Alfred and Tractor

Just like Alfred, the Spooling and submitting of jobs is painless. You can select the standard set of rendering methods (local or remote execution, min/max slots etc) just as you could with Alfred.
The jobs are automatically spooled via a script on the tractor engine called tractor-spool.py In actual fact, RenderMan Studio, will create an Alfred file for your job when you do an external render, and it passes this .alf file as a source file for the tractor-spool script.

Operating System Compatibility

64-bit Linux
Compatible with CentOS/RHEL 7.2 - gcc6.3 and glibc 2.17

64-bit Mac OSX
Compatible with versions 10.13 and 10.14

64-bit Windows
Compatible with Windows 8.1 and 10

DCC Compatibility

Maya
Compatible with 2018.3+

Houdini
Compatible with 17.0+

Katana
Compatible with 3.0+

Hardware Requirements

CPU
RenderMan requires x86-64 and SSE4.2 instruction set (introduced 2007). RenderMan threading scales well with multiple CPU cores and Hyper-Threading; results vary by scene.

GPU
Using the GPU Denoiser requires CUDA 8.0 (compute capability 2.0 or later) and a capable NVIDIA graphics card with support. If not, Denoise will fallback to CPU automatically.

Memory
RenderMan needs a minimum of 4GB of RAM. For best performance we recommend 8GB of RAM. Maximum RAM depends on scene complexity.

Plugins
Custom plugins can utilize C++14 and must be compiled with gcc6.3

VFX Reference Platform: CY2018 (OpenVDB 5.2, Alembic 1.7.x).