#DENOISER 2 FREEZING DRIVERS#
#DENOISER 2 FREEZING WINDOWS 10#
#DENOISER 2 FREEZING WINDOWS#

Write with default values: The scene format now supports optionally writing parameters with default values.

Transform2D: The procedural now supports remapping UsdTransform2D to built-in Arnold nodes.

Purpose:В Usd Purpose is now supported in the procedural.

Crease Sets:В The procedural now supports crease sets on polymesh.

NodeGraph schemas:В The procedural now supports using the NodeGraph schema for shader networks.

Pause and Resume:В Pausing and resuming renders are now supported in the render delegate.

Half and Double precision:В Storing data using half or double precision is now supported in both the render delegate and procedural.

Motion blur for the Point Instancer:В The render delegate now calculates motion blur when using the point instancer.

Light and Shadow linking:В The render delegate now supports light and shadow linking.

PRef AOV: The PRef AOV is now available when rendering on GPU (core#10444)).

In particular, when the resolution is low or adaptive sampling has reduced the amount of work left to do (core#10405). This will result in an overall GPU render time reduction up to 1.7x in fast to render scenes.

Faster GPU render passes: Each render iteration on the GPU has some CPU overhead which has been reduced.

Scenes with heavy subdivisions will use around 33% less GPU memory for geometry (core#10462).

Reduced VRAM use for polymeshes: The amount of VRAM used by polymesh is significantly lower in this version.

Faster SSS on GPU: This overall GPU optimization is most beneficial to scenes with lots for subsurface scattering, where we have observed up to 1.2x speedups (core#9917).

This also produced a speedup due to the faster skipping of empty space possible with NanoVDB. This can lead to a very significant reduction in GPU memory usage in scenes with volumes, depending on the volume sparsity, compared to the previous dense grid implementation, as shown in the table below.

Sparse volumes on GPU: VDB volumes are now loaded on GPU in a sparse form using NVIDIA's NanoVDB library.

Adding new imagers that don't rely on dynamic AOV creation will non longer restart the render (arv#494).

OpenImageIO upgraded:В OIIO has been upgraded to version 2.3.2 (core#9705, #9481).

They will also now report to the log an estimated time before the render completes (core#10406).

Better progress status updates for progressive renders: Progressive mode renders, which are always enabled on GPU and on CPU are enabled with options.enable_progressive_render, now have more accurate progress reports when doing adaptive sampling.

Again previously this would lead to much-duplicated memory that is now shared (see second graph below). The memory gains are even more substantial when imagers chains are shared between drivers.

#DENOISER 2 FREEZING DRIVERS#

If a single output is sent to multiple drivers the memory is now shared, rather than duplicated (see the first graph below).

Improved output memory usage: The memory usage of outputs has been improved in a variety of cases.

#DENOISER 2 FREEZING WINDOWS#

Faster imagers: Imagers are faster, especially on many-core Windows machines where we have seen up to a 1.7x speedup with the imager_denoiser_noice.For high AA_max renders this can result in noticeably faster renders (core#9022).

Faster renders using progressive rendering with adaptive sampling on CPU: When using progressive rendering with adaptive sampling on the CPU, Arnold will now more quickly end the render when there is no more work left to do.

Toon shader (contour_filter) optimizations: The toon shader has been optimized, especially for Windows machines with many cores and many small buckets, wherein in some situations we've seen overВ a 4x speedup (core#10414, #10242).

MtoA 4.2.2 works with the following Maya plugins

Optixв„ў denoiser requires an NVidia GPU with CUDAв„ў Compute Capability 5.0В and above.

We recommend using the 460.39В or higher drivers on Linux and 461.40 (Quadro), 461.40 (GeForce),В or higher on Windows.В See Getting Started with Arnold GPUВ for more information.

GPU rendering works on Windows and Linux only and requires anВ NVIDIA GPU of the Ampere, Turing, Volta, Pascal, or Maxwell architecture.

CPUs need to support the SSE4.1 instruction set.

Linux with at least glibc 2.17 and libstdc++ 4.8.5 (gcc 4.8.5).

#DENOISER 2 FREEZING WINDOWS 10#

Windows 10 or later, with the Visual Studio 2019 redistributable.

Follow these installation instructions.

MtoA 4.2.2 introducesВ Arnold 6.2.1.0В and is aВ performance release bringing important optimizations on GPU, toon shading, and imagers.

Denoiser 2 freezing