Intel Unveils Arc Pro B70 and B65 GPUs for Advanced AI and Professional Workloads
Intel has introduced two new graphics cards, the Arc Pro B70 and Arc Pro B65, designed specifically for demanding AI compute workloads and professional visualization on workstations. These GPUs are engineered to accelerate local AI inferencing, software development, and multi-GPU deployments in rack-scale environments, offering robust solutions for professionals seeking high-performance AI acceleration.
Arc Pro B70: Flagship Performance with Xe2 "Battlemage" Architecture
The Intel Arc Pro B70 stands out as the most powerful discrete GPU built on the Xe2 "Battlemage" architecture. It features 32 Xe cores (Xe2-HPG), 256 XMX engines for AI workloads, and 32 dedicated Ray Tracing Units. The card is equipped with 32 GB of GDDR6 memory on a 256-bit wide interface, delivering an impressive 608 GB/s of memory bandwidth. This configuration enables the Arc Pro B70 to achieve a peak throughput of 367 TOPS (INT8), making it highly suitable for intensive AI and compute tasks.
Connectivity is ensured through a PCI-Express 5.0 x16 interface, providing ample bandwidth for data-intensive applications. The Arc Pro B70 supports a comprehensive range of graphics and compute APIs, including DirectX 12 Ultimate, OpenGL 4.6, Vulkan 1.3, Intel oneAPI, OpenCL 3.0, and OpenVINO. Its advanced media engine offers hardware-accelerated encoding and decoding for AV1, HEVC, VP9, and H.265 formats. For display connectivity, the card includes four DisplayPort 2.1 outputs, supporting high-resolution, multi-monitor setups.
Power consumption for the Arc Pro B70 varies between 160 W and 290 W, depending on the specific implementation, with the Intel reference design rated at 230 W. Certified drivers will be available for Windows 11, Windows 10, and Linux, ensuring broad compatibility across professional environments.
Arc Pro B65: Optimized for Versatility and Efficiency
The Intel Arc Pro B65 offers a compelling alternative for professionals who require strong performance with a slightly different configuration. It features 24 Xe cores, 160 XMX engines, and 20 Ray Tracing Units—similar to the Arc B580 gaming GPU but with enhancements tailored for workstation use. Notably, the Arc Pro B65 is equipped with 32 GB of GDDR6 memory on a 192-bit interface, likely achieved through an asymmetric memory chip configuration.
Unlike the Arc B580, which uses a PCI-Express 5.0 x8 interface, the Arc Pro B65 utilizes the full PCI-Express 5.0 x16 interface, maximizing data throughput for professional workloads. The card maintains the same API support, media acceleration features, and display output options as the Arc Pro B70, including four DisplayPort 2.1 ports. The typical board power for the Arc Pro B65 is rated at 200 W.
Expanding Options for AI and Professional Visualization
With the launch of the Arc Pro B70 and B65, Intel is expanding its portfolio of workstation GPUs, providing advanced solutions for AI compute, software development, and professional visualization. These new graphics cards leverage the latest Xe2 architecture and offer robust memory configurations, high-speed interfaces, and comprehensive software support, positioning them as strong contenders in the professional GPU market.
