Intel has introduced the Arc Pro B70, a workstation-focused GPU designed for local Artificial Intelligence workloads rather than gaming or large-scale data center deployments. Priced at $949, the Arc Pro B70 is positioned well below the $1,800 RTX Pro 4000 from Nvidia and the $1,299 Radeon AI Pro R9700 from AMD. Intel is betting that lower-cost hardware with substantial memory can make local inference more practical for enterprise and professional users.
The Arc Pro B70 has 32 Xe cores that deliver 22.9 TFLOPS of FP32 compute performance. It features 32GB of GDDR6 RAM, enabling 608 GB/s bandwidth. Intel claims the Arc Pro B70 can achieve 367 TOPS (trillions of operations per second). The company says that memory capacity allows a broader range of Artificial Intelligence models and larger context windows to run on a workstation GPU. Using the open-source Llama 3.1 8B model, Intel claims the Arc Pro B70 can handle a context window more than twice as large as the Nvidia RTX Pro 4000. Intel also says the GPU can produce faster responses even with many concurrent users, and multiple cards can be linked together for larger models.
Intel is framing the product as a strong value play in a market where Nvidia still benefits from CUDA and its entrenched software ecosystem. Compared to the RTX Pro 4000, Intel claims the Arc Pro B70 can deliver up to two times as many tokens per dollar, depending on the model. The company is also targeting users of professional and creative applications, not only dedicated Artificial Intelligence deployments.
Intel sees local Artificial Intelligence as a longer-term shift away from mega data centers, especially for enterprise inference workloads with predictable demand and smaller models. Running these systems in on-premises or private cloud environments could lower costs and keep data out of the public cloud. Intel also points to broader progress in personal computing hardware. The top-tier Panther Lake laptop chip, for example, can deliver 180 TOPS of performance via its built-in Artificial Intelligence accelerator and integrated graphics.
Memory pricing remains a major constraint on this vision. The ongoing memory shortage is described as a severe headwind, and lower memory chip prices will be needed before PCs can support more capable Artificial Intelligence models. Even so, Intel’s inclusion of 32GB of memory in the Arc Pro B70 highlights how it is preparing for a future in which more Artificial Intelligence computing power moves into workstations and PCs, alongside its separate push into data center hardware with the upcoming Crescent Island GPU.
