Intel Unveils New GPUs, AI Accelerators, and Developer Tools at Computex 2025
At Computex 2025, Intel revealed its latest advancements in GPU technology, AI accelerators, and developer frameworks. Designed for professional workstations and scalable enterprise solutions, these launches extend Intel’s capabilities in AI and visual computing.
The Intel Arc Pro B-Series now includes the Arc Pro B60 and Arc Pro B50 GPUs. Built on Xe2 architecture, these GPUs are engineered for AI inference, design workflows, and ray tracing. The B60 model features 24GB of memory, while the B50 includes 16GB. Both support multi-GPU scalability and are certified by independent software vendors for professional applications.
These GPUs are optimized for architecture, engineering, and construction workflows. They are compatible with both consumer and pro drivers on Windows, while on Linux, they support containerized AI deployment environments.
To support high-performance model development, Intel introduced a workstation platform known as Project Battlematrix. It can support up to eight Arc Pro B60 GPUs and enables training of medium-sized AI models with 150 billion parameters using 192GB of combined VRAM.
Also read: QuickLogic Delivers eFPGA Hard IP Optimized for Intel 18A Technology
Expanded Gaudi 3 Accelerator Lineup for Scalable AI
Intel also launched new deployment options for its Gaudi 3 AI Accelerators. These now include PCIe card formats and full rack-scale configurations. The PCIe cards offer flexibility within existing data center environments and are designed for models ranging from Llama 3.1 8B to full Llama 4 variants.
The rack-scale reference systems support up to 64 accelerators and feature 8.2TB of high-bandwidth memory. A modular architecture ensures compatibility with open compute designs, while liquid cooling enhances performance and power efficiency.
Developer Tools Now Publicly Available
The Intel AI Assistant Builder is now accessible in public beta on GitHub. This lightweight framework allows developers to create and deploy localized AI agents tailored to Intel hardware. It simplifies agent deployment for both consumer and enterprise use cases.
With these updates, Intel continues to provide scalable, open, and performance-optimized solutions for AI and professional computing workloads.
