Computex 2025: Intel Unveils New GPUs for AI and Workstations

Intel Arc Pro B60: 2025 Enterprise GPU Revolution { "@context": "https://schema.org", "@type": "TechArticle", "headline": "Intel Arc Pro B60: 2025 Enterprise GPU Revolution", "description": "Explore Intel’s Arc Pro B60 GPU—high‑memory, low‑power inference for 2025 enterprises. Compare with NVIDIA Ada, benchmark ROI, and learn deployment best practices.", "author": { "@type": "Person", "name": "Senior Technology Journalist" }, "datePublished": "2025-12-01", "dateModified": "2025-12-01", "publisher": { "@type": "Organization", "name": "Tech Insights Network" } } figure {margin:1.5rem 0;} figcaption{font-size:.9em;color:#555;} table{border-collapse:collapse;width:100%;margin:1rem 0;} th,td{border:1px solid #ddd;padding:.75rem;text-align:left;} th{background:#f7f7f7;} Intel Arc Pro B60: 2025 Enterprise GPU Revolution Published on: December 2025 In the wake of NVIDIA’s Ada dominance and AMD’s incremental gains, Intel has finally announced a full‑size workstation GPU— Arc Pro B60 —that could reshape how mid‑market enterprises build inference pipelines. The 24‑GB HBM3e card delivers nearly double GPT‑4o throughput over an RTX 4060 Ti while drawing 35% less power, making it a compelling low‑capex alternative for data scientists and AI architects. Executive Summary Arc Pro B60 “Battlemage” : 24 GB HBM3e, XMX matrix cores, $299–$600 per GPU; ~1.8× GPT‑4o inference throughput over RTX 4060 Ti with Gaudi 3 Rack‑Scale Accelerator : 64 Xe² cores per rack, 8.2 TB HBM, >10× TFLOPs/W vs. Gaudi 2; open‑compute compatible and liquid‑cooled. Intel’s oneAPI , OpenVINO, and AI Assistant API mirror CUDA but maintain tighter x86 integration. Price/performance and energy density make Intel attractive for SMEs, research labs, and edge deployments that cannot afford NVIDIA’s flagship cards. Key risk: lack of independent long‑term benchmarks; software maturity still catching up with emerging frameworks (PyTorch 1.13+, JAX). Strategic Business Implications Intel’s dual launch targets two distin