Intel oneAPI

Why choose Intel oneAPI?

The use of high-performance computing technologies such as AI, Machine Learning, and Deep Learning requires developers to deploy and scale HPC workloads in a way that unleashes the full capabilities of Intel-powered processors, gas pedals, and HPC clusters.

Intel recently introduced the Intel oneAPI family, the next generation of Intel software development tools, based on current standards for creating and deploying high-performance, data-centric applications for various architectures.

INTEL ONE API AND HPC TOOLKIT COMBINE TO CREATE ONE HPC PRODUCT:

One API and HPC Toolkits include state-of-the-art vectorization, multithreading, multi-node parallelization, and memory optimization techniques. Deploy applications and solutions on shared-memory and distributed-memory systems (such as clusters) using development tools, a standardized MPI library, MPI analyzer, and cluster tuning and monitoring tools.

INTEL oneAPI TOOLKIT

Intel oneAPI Toolkit is a set of tools and libraries for developing high-performance, data-centric applications targeting a variety of Intel architectures (CPU, GPU, FPGA).

• Use one single programming model and one  cross-architecture language Data Parallel C++ (DPC++) to develop high-performance, data-centric applications that enable code reuse on hardware targets such as Intel’s CPUs, GPUs and FPGAs and allow custom tuning for each gas pedal.
• Develop efficiently using the DPC+/C++/C compiler, performance libraries, enhanced design-assist profiling and debugging tools, and accelerated Python.
• Migrate your CUDA code to the new DPC++ code that can be used on multiple architectures using a compatibility tool, and adapt it as needed for specific implementations.
• Develop future-ready code to facilitate adoption of future innovative HPC architectures.

Build, analyze, optimize and adapt fast HPC applications for a variety of architectures with vectorization, multithreading, multi-node parallelization and memory optimization techniques

• Build. Simplify the deployment of cross-architecture HPC applications on Intel processors and gas pedals using Intel’s industry-leading compilers and libraries. Efficiently create fast parallel code and improve application performance by leveraging the cutting-edge features of current and future Intel architecture.
• Analyze. Quickly assess application performance, resource utilization, and areas to optimize to ensure fast performance across different architectures.
• Optimize. Discover how resource utilization affects your code, including compute, memory, I/O, etc. to make smart cross-architecture design decisions.
• Supports HPC standards, including C/C++, Fortran, Python, OpenMP, and MPI, for easy integration with existing code.

What’s new in Intel oneAPI 2025

Intel oneAPI evolves with its latest 2025 release, offering even more features and innovations. This unified platform for high-performance and AI application development pushes the boundaries of performance on the latest Intel architectures. It simplifies access to industry-standard tools while broadening available hardware options, giving developers optimal flexibility and efficiency to fully leverage Intel technologies.

Unmatched performance on Intel platforms to maximize your AI and HPC projects

• Up to 3x higher generative AI (GenAI) performance on 6th Gen Intel® Xeon® processors with oneDNN, Intel-optimized AI frameworks, and Intel® AMX.
• Up to 2.5x higher HPCG performance powered by MRDIMM and oneMKL.
• High-performance AI development on PCs (including large language models like LLMs) using tools optimized for Intel® Core™ Ultra processors.
• Enhance security and encryption with the Intel® Cryptography Primitives Library.

Simplified and enhanced access to industry-standard tools

• Harness leading AI frameworks and performance libraries with expanded Intel optimizations, including native support for PyTorch 2.5 on CPUs and GPUs.
• Optimize performance for LLMs (Llama 3.2, Qwen2, Phi-3) using Intel AI tools.
• Streamline tool installation with a selector to configure full kits or tailored subsets.

Expanded hardware flexibility to maximize multi-architecture resources

• Improved support for multi-vendor and multi-architecture environments, including faster CUDA-to-SYCL migration with the Intel® DPC++ Compatibility Tool.
• Near-native CPU and GPU performance for numerical computing with Intel® Distribution for Python.
• 4x faster GPU kernel acceleration for algorithms with oneDPL, and future flexibility with cross-hardware AI libraries (Triton, JAX, OpenXLA).

Is your PSXE maintenance still in progress? Take advantage of the free upgrade to oneAPI .