Top Stories

Latest Releases and Resources: NVIDIA GTC 2022

Our weekly roundup covers the most recent software updates, learning resources, events, and notable news. This week we have several software releases.

Software releases 

Leveraging standard languages for portable and performant code with the HPC SDK

The NVIDIA HPC SDK is a comprehensive suite of compilers, libraries, and tools for developing accelerated HPC applications. With a breadth of flexible support options, users can create applications with a programming model most relevant to their situation. 

The HPC SDK offers a variety of programming models including performance-optimized drop-in libraries, standard languages, directives-based methods, and specialization provided by CUDA. Many of the latest enhancements have been in the area of standard language support for ISO C++, ISO Fortran, and Python. 

The NVIDIA HPC compilers use recent advances in the public specifications for these languages, delivering a productive programming environment that is both portable and performant for scaling on GPU accelerated platforms. 

Visit our site to download the new HPC SDK version 22.3 and read our new post on parallel programming with standard languages under our “Resources” section.

Get started: NVIDIA HPC SDK 22.3     

Accelerate quantum circuit simulation with the NVIDIA cuQuantum SDK  

cuQuantum – An SDK for accelerating quantum circuit simulation
NVIDIA cuQuantum is an SDK of optimized libraries and tools for accelerating quantum computing workflows. Developers can use cuQuantum while creating and verifying new algorithms more easily and reliably. Speeding up quantum circuit simulations by orders of magnitude, for both state vector and tensor network methods helps developers simulate bigger problems, faster.  

Expanding ecosystem integrations and collaborations
cuQuantum is now integrated as a backend in popular industry simulators. It is also offered as a part of quantum application development platforms, and is used to power quantum research at scale in areas from chemistry to climate modeling.

SDK available with new appliance beta 
The cuQuantum SDK is now GA and free to download. NVIDIA has also packaged up an optimized Beta software container, the cuQuantum DGX appliance, available from the NGC Catalog. 

Learn more: cuQuantum SDK 

Boost ray tracing application performance using Nsight Graphics 2022.2    

Nsight Graphics is a performance analysis tool designed to visualize, analyze, and optimize programming models. It also tunes to scale efficiently across any quantity or size of CPUs and GPUs—from workstations to supercomputers.

The latest features in Nsight Graphics 2022.2 include:

  • AABB Overlay Heatmap display
  • Shader Timing Heatmap (with D3D12/Vulkan + RT support)
  • Shows other processes using the GPU with ETW
  • Vulkan video extension 

Download now: Nsight Graphics 2022.2

Simplify system profiling and debugging with Nsight Systems 2022.2    

Nsight Systems is a triage and performance analysis tool designed to track GPU workloads to their CPU origins within a system-wide view. The features help you analyze your applications’ GPU utilization, graphic and compute API activity, and OS runtime operations. This helps optimize your application to perform and scale efficiently across any quantity or size of CPUs and GPUs—from workstations to supercomputers.

What’s new:

  • NVIDIA NIC Ethernet metrics sampling
  • Vulkan memory operations and warnings
  • Vulkan graphics pipeline library
  • Multireport view enhancements

Download now: Nsight Systems 2022.2

Advanced Deep Learning model design with Nsight Deep Learning Designer 2022.1

NVIDIA Nsight Deep Learning Designer is an integrated development environment that helps developers efficiently design and deploy high performance deep neural networks for in-app inference.

Features include:

  • Support for nested templates, allowing developers to easily create hierarchical model graphs.
  • Improved performance of the type-checking process in NeuralEditor and the overall user experience with the PyTorch exporter.
  • Support to save all tensors in the analysis mode.
  • New scalar-only Constant Input layer and new analysis layer – Signal Injector.
  • Performance improvement in the BatchNorm layer and the Upscale layer.
  • Support for down-scaling and fixed-size scaling in the Upscale layer.

Download Now

Enhanced CUDA 11.6, Update 1, platform for all new SDKs 

This CUDA Toolkit release focuses on enhancing the programming model and performance of CUDA applications. CUDA 11.6 ships with the R510 driver, an update branch. CUDA Toolkit 11.6, Update 1, is available to download.

What’s new:

  • GSP driver architecture is now default on NVIDIA Turing and Ampere GPUs.
  • New API for disabling nodes in instantiated graph.
  • Full support of 128-bit integer type
  • Cooperative groups namespace update
  • CUDA compiler update
  • Nsight Compute 2022.1 release

Learn more: CUDA Toolkit 11.6, Update 1                            

Deliver distributed and accelerated computing to Python with the help of cuNumeric 

NVIDIA cuNumeric is a Legate library that aspires to provide a drop-in replacement for the NumPy API on top of the Legion runtime. This brings distributed and accelerated computing on the NVIDIA platform to the Python community.

What’s new:

  • Transparently accelerates and scales existing NumPy workflow
  • Scales to up to thousands of GPUs optimally
  • Requires zero code changes to ensure developer productivity
  • Freely available, get started on GitHub or Conda

Learn more: cuNumeric

Warp helps Python coders with GPU-accelerated graphics simulation    

Warp is a Python framework that gives coders an easy way to write GPU-accelerated, kernel-based programs in NVIDIA Omniverse and OmniGraph. With Warp, Omniverse developers can create GPU-accelerated, 3D simulation workflows and fantastic virtual worlds!

What’s new:

  • Kernel-based code in Python
  • Differentiable programming
  • Built-in geometry processing
  • Simulation performance on par with native code
  • Shorter time to market with improved iteration time

Learn more: Warp

Discuss (0)