Intel Arc Pro B60chat_bubble_outlineFrom real comments

Huge VRAM at a reasonable price is the card's biggest draw. Multiple commenters highlight that 48GB (via dual 24GB GPUs) at under $1000 undercuts Nvidia badly, since Nvidia 'gives you 8GB and tells you to get lost.' Buyers see it as a cheaper alternative to used 3090s or awkward multi-GPU builds, ideal for LLMs, CFD simulations, and hybrid gaming/AI home rigs, provided pricing stays reasonable and stock is available.

Software and driver support remain a major weak point. Several users struggled getting llama.cpp, PyTorch, and CUDA-equivalent workflows running, calling it a 'huge opportunity lost' compared to Nvidia and AMD. OpenVINO backend updates are described as behind, and one user directly says PyTorch just won't translate CUDA calls to Intel. This software gap is seen as the card's core limitation versus its hardware potential.

Real-world AI and gaming benchmarks are inconsistent and sometimes disappointing. Some testers report memory latency worse than AMD/Nvidia, an underclocked 3090 outperforming it, and a competing card getting double the tokens/second despite lower bandwidth. ComfyUI video generation and gaming results were called 'rough' and 'disappointing,' with one commenter comparing it unfavorably to the failed XGI Volari Duo multi-chip experiment from the past.