According to news on the evening of May 19, at Computex 2025, Intel released a new graphics processor (GPU) and AI accelerator product series designed for professionals and developers. include:
The new Intel Sharp Pro B series GPUs: Intel Sharp Pro B60 and Intel Sharp Pro B50 GPUs are equipped with configurations tailored for AI inference and professional workstations, expanding the Intel Sharp Pro product series.
Intel Gaudi 3 AI Accelerator: The Intel Gaudi 3 AI accelerator is now available in PCIe and rack-scale system deployment options, providing a scalable, open solution for enterprise and cloud AI inference.
Intel AI Assistant Builder: Intel AI Assistant Builder is now available on GitHub, allowing developers to create purpose-built native AI agents optimized for Intel platforms.
Intel CEO Chen Liwu said: "With the opportunity of Computex, Intel will continue to strengthen its commitment to partners and build a new Intel for the future. We will work together to create excellent products, meet customer needs, and seize future opportunities."
New Intel GPU to meet the performance needs of AI and workstations
The new Intel Ruixuan Pro B60 and B50 GPUs are based on the Xe2 architecture, equipped with Intel XMX AI core and advanced ray tracing unit, providing powerful performance support for creators, developers and engineers.
The launch of new products enriches Intel's professional-grade GPU product line. Intel Radar Pro B60 and Intel Radar Pro B50 GPUs are designed for demanding AI inference workloads and workstation applications. The Intel Ruixuan Pro B series is equipped with AI features, equipped with 24GB and 16GB of video memory respectively, and supports multi-GPU expansion, providing diversified solutions for creators, AI developers and professionals.
Both GPUs are optimized for AEC (architecture, engineering, construction) and inference workstations, and provide stable and reliable performance with a large number of ISV certified and optimized software. Intel Ruixuan Pro B series GPUs are compatible with consumer-grade and professional-grade drivers on Windows, and support a containerized software stack on Linux. This software stack can simplify AI deployment and support gradual upgrades and feature optimization. By combining high graphics memory capacity with critical software compatibility, the Intel Sharp Pro B Series provides creators and AI developers with a scalable, cost-effective solution. In addition, Intel also released a configurable workstation-level Intel Xeon platform (codenamed Project Battlematrix) to help AI developers solve the problems they face. With up to 192GB of video random access memory (VRAM), it can support up to eight Intel Sharp Pro B60 24GB GPUs to run medium-scale and high-precision AI models of up to 150 billion parameters.
Vivian Lien, vice president and general manager of client graphics at Intel, said: "The launch of the Intel Radar Pro B series fully demonstrates Intel's persistence and commitment in the field of GPU technology and ecosystem. With the advanced performance of the Xe 2 architecture and the growing software ecosystem, the new Intel Radar Pro GPU provides an accessible and scalable solution for small and medium-sized enterprises that have been looking for targeted solutions."
AIB partners, including but not limited to ASRock, Lanner, Lanner, Maxxuan, Aoshi, Senao and Hanhe Technology, will begin to provide Intel Ruixuan Pro B60 GPU samples in June 2025. Intel Radar Pro B50 GPU will be available at Intel authorized resellers starting in July 2025.
Intel Gaudi 3 supports PCIe and rack-scale deployment
Intel launches a new deployment solution for the Intel Gaudi 3 AI accelerator to further advance its AI strategy:
Intel Gaudi 3 PCIe cards enable scalable AI inference in existing server data center environments. Taking the large AI model LLaMa as an example, thanks to the scalable configuration, AI models of various sizes from LLaMA 3.1 8B to LLaMA 4 Scout or Maverick can be flexibly run regardless of the size of the enterprise. Intel Gaudi 3 PCIe cards are expected to be available in the second half of 2025.
The Intel Gaudi 3 rack-scale system reference design brings extreme flexibility and scalability, with up to 64 accelerators per rack and up to 8.2TB of high-bandwidth memory. The open modular design eliminates vendor constraints, while wired backplanes and blind-mate 2D full-rack cabling make installation and maintenance simpler. With liquid cooling, these systems deliver powerful performance and aid in TCO control.
The Intel Gaudi rack-scale architecture is deeply optimized for large AI models, enabling excellent real-time inference performance with low latency. These configurations support the customized needs of cloud service providers (CSPs) and the Open Compute Project (OCP) design specifications, further demonstrating Intel's strong commitment to building open, flexible and secure AI infrastructure.
Intel AI Assistant Builder is now available
After debuting at CES 2025, Intel AI Assistant Builder is now available in public beta on GitHub. This lightweight open software framework is designed for building and running custom AI agents natively on Intel-based AI PCs. The innovative solutions presented by Acer and ASUS at Computex fully demonstrate the power of AI Assistant Builder. Developers and partners can use this tool to quickly build and deploy AI agents for their organizations and customers, thereby creating more business value.
At Computex 2025, Intel showcased its latest processors and graphics cards.