After anticipating at the beginning of Computex 2025, with the presentation of the Panther Lake Chips, the opening day of the fair has served as a framework to Intel to present your novelties And remember what he does 40 years that began its presence in Taiwan, Thought both for work with artificial intelligence and for their integration into work stations.
Among the novelties that the company has announced are the Intel Arc Pro Series B GPU, which for now are two: Arc Pro B60 y Arc Pro B50. In both cases they are destined to inference tasks of artificial intelligence and professional workstations. They are based on the XE2 architecture, and have the intel xe matrix nuclei extensions and units of advanced ray. They have different functions for AI and multigpu scalability, as well as 16 GB and 24 GB of memory, depending on the model.
Both are optimized for work stations designed for architecture, engineering and construction (AEC), as well as for inference. It has a wide variety of independent software suppliers (ISVS) and optimized software.
The company has also presented a configurable platform for Intel Xeon -based work stations, which has a code name Project Battlematrix and that is designed to facilitate the work of IA developers. It admits up to eight GPU Intel Arc Pro B60 of 24 GB to enable medium -sized AI models of up to 150,000 million parameters and high precision, with up to 192 GB of RAM video memory.
The Intel ARC Pro B60 GPU will begin its distribution in the form of samples to Intel card manufacturers from next June. Among the partners of the company that will receive them are Asrock, Gunnir, Lanner, Maxsun, Onix, Senao and Sparkle. As for the Intel ARC Pro B50, it will be available through distributors authorized by Intel since July 2025.
Another news is range of Intel Gaudi 3 accelerators 3already available in PCIS systems and rack scale, as open and scalable solutions for companies that perform AI inference tasks, both in local and Cloud. These cards support the scalable inference within servers environments in data centers.
The reference designs of large -scale Rack Gaudi 3 systems are developed to offer flexibility and scalability, and support up to 64 rack accelerators offer 8.2 TB of high speed memory, and thanks to their modular and open design they avoid blocking with suppliers.
Thanks to their scalable settings, customers who use it, from small to multinational companies, can run models that range from calls 3.1 8b to flame 4 scout or large -scale maverick. These PCIe accelerators will be available throughout the second half of 2025.
Finally, the Intel AI Assistant Builderalready publicly available in Github, allows developers to create local and specific agents, optimized of course for Intel platforms. It offers a light and open software frame to create and execute customs of PCS personalized in PCs based on Intel.
These releases, which coincide with the Intel 40th Anniversary in Taiwan, where Computex is celebratedThey mark an important point for the company. His CEO, Lip-Bu Tan, has taken advantage of the presentation to remind the sector that «During the last 40 years, the strength of our association with the Taiwan ecosystem has promoted an innovation that has changed our world for the better. This week we renew our commitment to our partners while we work to build a new Intel for the future. Together, we will create excellent products that will enchant our customers and take advantage of the exciting opportunities that are coming«.
