Home / Component / Graphics / ASRock unveils 32GB Intel Arc Pro B70 Creator and Passive GPUs

ASRock unveils 32GB Intel Arc Pro B70 Creator and Passive GPUs

ASRock has officially rolled out its custom take on the Intel Arc Pro B70. Targeting the growing demand for local AI inference and professional rendering, ASRock has introduced two distinct models: the Intel Arc Pro B70 Creator 32GB and the Intel Arc Pro B70 Passive 32GB. 

The Intel Arc Pro B70 Creator 32 GB is designed for high-performance workstations that require sustained reliability. It features a classic 2-slot blower-style cooler built around a vapour chamber, making it suitable for multi-GPU setups. To ensure optimal heat transfer, ASRock is using Honeywell PTM7950 phase-change material, a high-end thermal solution chosen to prevent the performance throttling that often plagues workstation cards during long rendering or training sessions. For high-density server environments and industrial PC (IPC) applications, the Intel Arc Pro B70 Passive 32GB offers a fanless, maintenance-free alternative with a minimalist metal shroud.

Both cards leverage Intel's latest Xe2-HPG microarchitecture, which brings significant improvements to AI-driven creative tools and OpenGL-heavy professional software. Both cards have a 32GB memory pool and a 256-bit bus, providing 608GB/s of bandwidth and a peak AI performance of 367 INT8 TOPS. With 32 Xe cores clocked at 2,540MHz, these B70 cards are positioned as a cost-effective rival to Nvidia's RTX Pro 4000 series.

KitGuru says: Intel's “Big Battlemage” has finally arrived. Unfortunately for gamers, it's currently limited to the Pro series. 

Become a Patron!

Check Also

Intel launches Core Ultra Series 3 CPUs, new Arc Pro graphics cards and more

Intel has introduced its latest commercial client portfolio, led by the new Intel Core Ultra Series 3 processors and a pair of professional‑grade Arc Pro GPUs. The lineup targets business laptops, desktops and workstations, with a focus on efficiency, security and as you would expect in 2026, AI‑accelerated workloads.