Release Notes for Nuke and Hiero 15.0v4

Release Date

8 February 2024

New Features

There are no new features in this release.

Feature Enhancements

There are no new feature enhancements in this release.

Bug Fixes

 

• ID 564298 - Certain 2D nodes open the 3D toolbar

• ID 564372 - Nuke crashes on launch due to OpenCL Error

New Known Issues Specific to Nuke 15.0

This section covers new known issues and gives workarounds for them, where appropriate.

 

• ID 532048 - [3d Arch] DirectLight not illuminating scene objects in 3D Hydra

• ID 546692 - [OpenAssetIO] Nuke displays balReader not found with correct OpenAssetIO setup

• ID 547439 - [OpenAssetIO] color profile appears to have not loaded errored for braw files (sidecar issue?)

• ID 551945 - C_CameraSolver - Undo command not removing all Camera Ingest nodes

• ID 553640 - Specific MXF (op-1a 4:4:4 12bit) files fails to read in resolve

• ID 560317 - [PUBLIC] Playback/Caching of Prores files is worse in Nuke Studio/Hiero 14.1/15.0 than 14.0 on macOS

• ID 538645 - [Blink] Heavy Blinkscripts with multiple parameters will sometimes render incorrectly on Mac Arm

• ID 546734 - Unable to launch Studio, Hiero or Player on a specific machine with multi-GPUs

• ID 545782 - [Blink] Disabling/Enabling Tracks and Track Items causes the UI Spinner to appear

• ID 546803 - [Lens] Grid Detect only works on the first/left view of multi-view footage

• ID 548563 - [Lens] Moving the denominator knobs can cause UI to block

• ID 548998 - [Inference] Soft effects in error state stop Inferences in the soft effect stack from affecting the viewer

• ID 552644 - [Disk] Disabling a track item during auto timeline disk caching incorrectly invalidates the existing cached frames in the track item's range

• ID 553436 - [Blink] Black boxes appear in viewer when using BoxBlur2D kernel

• ID 560042 - [PUBLIC] The Burn-In soft effect evaluates before a Crop soft effect when above it in the stack

• ID 541985 - [Apple Silicon] Viewer becomes a solid colour when changing Ultimatte 'Screen Colour' from non-default to default

• ID 542483 - [Apple Silicon] Floating point errors between ARM and INTEL hardware

• ID 553337 - [CopyCat] The training Contact sheet doesn't reframe when selecting different trainings

• ID 554762 - [CopyCat] The training contact sheet will sometimes appear blank when getting training updates

• ID 554798 [CopyCat] When doing more than 1 training into the same directory, they will all be selected

• ID 554834 - [CopyCat] LiveUpdates causes script to be in modified state

• ID 527336 - [3D Arch] Camera 'Display at world Scale' knob does nothing

• ID 538007 - [3D Arch] NDK examples do not build

• ID 544629 - [3D Arch] Selected overlapping vertices not drawn

• ID 558775 - [PUBLIC] The inputs of the EnvironmentLight node do not match the classic Environment node

• ID 562002 - [PUBLIC] Changing certain GeoCard knobs will produce a "Failed to create spec" error while Viewing a GeoBindMaterial node downstream

• ID 538730 - [3D Arch] EnviromentLights scene input only accepts lights and cameras

• ID 542145 - [3D Arch] ScanlineRender2 crash when animating grade values

• ID 543275 - [3D Arch] Disabling and re-enabling a Geo node removes the bound material

• ID 537033 - OCIODisplay's display and view knobs not updated after changing OCIO config in Nuke

• ID 543135 - [OCIOv2.1] - Switching to another config loads wrong Input Transform

• ID 545900 - [PUBLIC] An OVRService (Open VR) error occurs when launching Nuke

• ID 548315 - [MO SDI] Timeline - 12 bit 4:4:4 on BMD shows incorrect output when there is an aspect ratio mismatch between the Output Resolution and the Display Mode (Studio)

• ID 554719 - Export only the knob values for the selected transform type

• ID 543621 - [AIR Tools] The AIR Tools don't work with CUDA 11.8 on CC 3.5 GPU

• ID 548132 - [Inference] Soft effect will process/display different results based on zoom level with auto proxy on

• ID 548734 - [Inference] Optimise for Speed is hidden from the inference node properties on OSX

• ID 549062 - [CopyCat][DisTraining] Executing 2 CopyCat nodes with different parameters works

• ID 550941 - [CopyCat] Crashes with unhandled exception when transform node with expression

• ID 551424 - [CopyCat] Progress bar is slow to appear (Local training)

• ID 552827 - [Dis Training] Cannot kill nuke processes in terminal win10

• ID 554790 - [CopyCat] Deleting trainings from datadirectory will cause unexpected behaviour in progress tab

• ID 555638 - [Inference/CopyCat] When hitting OOM on OSX GPU causes crash

• ID 555713 - [CopyCat][DistTrain] Distributed Training doesn't work across OS platforms

• ID 555714 [CopyCat][DistTraining] Linux cannot use IPv6 to train

• ID 555716 - [PUBLIC] [CopyCat][DistTraining] Windows cannot use IPv4 to train

• ID 558607 - [CopyCat] MultiRes OOM message doesn't appear when not caching

• ID 562775 - [CopyCat] Preview/contactsheet do not update to latest image if it saves on last step

• ID 553268 - Nuke ARM incorrectly prompts users to install through a Rosetta layer when opening Nuke X and Studio through the shortcuts on Macs without Rosetta installed

Qualified Operating Systems

• macOS Monterey (12.x), or macOS Ventura (13.x). Nuke 15.0 supports Apple's silicon hardware on M1 and M2 chips.

Article:  For more information on Foundry products and supported macOS versions, see Foundry Knowledge Base article Q100592.

• Windows 10 (64-bit) or Windows 11 (64-bit)

• Linux Rocky 9.0 (64-bit)

Warning:  Nuke requires libnuma to run under Linux distributions, the library is required by the Nablet H264 Codec SDK.

Note:  The currently supported version of VFX Reference Platform includes library versions that are only compatible with Rocky 9.0.

Other operating systems may work, but have not been fully tested.

Requirements for Nuke's GPU Acceleration

If you want to enable Nuke to calculate certain nodes using the GPU, there are some additional requirements.

NVIDIA

An NVIDIA GPU with graphics drivers capable of running CUDA 11.8, or above. A list of the compute capabilities of NVIDIA GPUs is available at https://developer.nvidia.com/cuda-gpus

Note:  The compute capability is a property of the GPU hardware and can't be altered by a software update.

With graphics drivers capable of running CUDA 11.8, or above. On Windows and Linux, CUDA graphics drivers are bundled with the regular drivers for your NVIDIA GPU. Driver versions 522.06 (Windows) and 520.61.05 (Linux), or above are required. See https://www.nvidia.com/Download/Find.aspx for more information on compatible drivers.

Note:  We recommend using the latest graphics drivers, where possible, regardless of operating system.

AMD

Note:  Bitwise equality between GPU and CPU holds in most cases, but for some operations there are limitations to the accuracy possible with this configuration.

• On Windows and Linux, an AMD GPU from the following list:

Note:  Other AMD GPUs may work, but have not been fully tested.

• AMD Radeon PRO W7900

• AMD Radeon PRO W6600

• AMD Radeon PRO W6800

• AMD Radeon Pro W5700

• AMD Radeon RX 6800 XT

Note:  For information on the recommended driver for each GPU, see https://www.amd.com/en/support

• On Mac, integrated AMD GPUs are supported on the following Intel CPU Macs:

• Any late 2013 Mac Pro onward (including 2019 Mac Pro),

• Mid-2015 MacBook Pros onward, and

• Late 2017 iMac Pros onward.

All supported Mac Pros include a multi-GPU support option, where applicable. Bitwise equality between GPU and CPU holds in most cases, but for some operations, there are limitations to the accuracy possible with this configuration.

Warning:  Although AMD GPUs are enabled on other Mac models, they are not officially supported and used at your own risk.

Multi-GPU Processing

Nuke's GPU support includes an Enable multi-GPU support option. When enabled in the preferences, GPU processing is shared between the available GPUs for extra processing speed.

Note:  Multi-GPU processing is only available for identical GPUs in the same machine. For example, two NVIDIA GeForce GTX 1080s or two AMD Radeon™ Pro WX 9100s.

GPU Requirements for the Machine Learning Toolset

Training using the CopyCat node requires an NVIDIA GPU, with compute capability 3.5 or above; or MacOS Apple silicon integrated GPUs.

If an appropriate GPU is not available, Inference and other machine learning plug-ins can run on the CPU with significantly degraded performance.

Developer Notes

As Nuke develops, we sometimes have to make changes to the API and ABI under the hood. We try to keep these changes to a minimum and only for certain releases, but from time to time API and ABI compatibility is not guaranteed. See the following table for the situations when you may have to recompile your plug-ins and/or make changes to the source code.

Release Type	Example	Compatibility	Recompile	Rewrite
Version	14.0v1 to 14.0v2	API and ABI
Point	14.0v1 to 14.1v1	API
Major	14.0v1 to 15.0v1	-

Additionally, node Class() names occasionally change between major releases. While these changes do not affect legacy scripts, you may not get the results you were expecting if a node class has been modified. The toolbars.py file, used to create Nuke's node toolbar, contains all the current node class names and is located in <install_directory>/plugins/nukescripts/ for reference.

As an example, between Nuke 13 and Nuke 14, the Axis node Class() changed from Axis3 to Axis4. In the toolbars.py file for the two releases, the entries for the Axis node appear as follows:

m3Dclassic.addCommand("Axis", "nuke.createNode(\"Axis3\")", icon="Axis.png", tag=MenuItemTag.Classic, node="Axis3", tagTarget=MenuItemTagTargetFlag.TabMenu)

m3D.addCommand("Axis", "nuke.createNode(\"Axis4\")", icon="Axis_3D.png", tag=MenuItemTag.Beta, node="Axis4")