Intel has introduced its Arc Pro B70 graphics card alongside a new board partner called ARKN, and this launch expands Intel’s professional GPU lineup with a stronger focus on AI workloads and workstation use, while also adding more partner options beyond the usual names like ASRock, Gunnir, Maxsun, and Sparkle.
ARKN enters the market as a new player, but it already sells Arc Pro cards, including a 24GB B60 variant listed earlier this year, and now it plans to offer three models across B50, B60, and B70 tiers, which gives Intel a wider presence in the professional GPU space.
Intel positions the Arc Pro B70 as its most capable Battlemage GPU so far, built on the Xe2 architecture and powered by the larger BMG-G31 chip, and it brings 32GB of GDDR6 memory, 32 Xe cores, and up to 608 GB/s bandwidth, which helps it handle demanding workloads like local AI inference and content creation.
Here are the key specs:
- 32GB GDDR6 memory
- 32 Xe cores
- 367 INT8 TOPS AI performance
- 608 GB/s memory bandwidth
- 160W to 290W board power
The Arc Pro B65 and B60 sit below it, with reduced core counts and memory, while the B50 targets lighter workloads with a smaller configuration.
Intel clearly targets developers and professionals who want to run AI models locally, and the large VRAM allows users to load bigger models or longer context windows without relying on cloud services, which reduces ongoing costs for repeated workloads.
The company also highlights multi-GPU support under Linux through oneAPI, which lets users scale performance by running several GPUs together in workstations or servers.
Competes with Nvidia RTX Pro 4000
Intel positions the B70 against Nvidia’s RTX Pro 4000, and instead of focusing on raw compute numbers, it claims better value in AI inference workloads, especially when comparing tokens per dollar, while also offering more memory at a lower price point.
Intel makes it clear that Arc Pro GPUs are not designed for gaming, even though they can run games, and the company continues to keep its gaming GPU plans separate while focusing this lineup on professional use cases.
