Intel finally announces 'Big Battlemage' with new Arc Pro B70 and Arc Pro B65 graphics cards

It seems like we've been waiting for Intel to formally announce its 'Big Battlemage' graphics card for a couple of years now, with the good news being that the wait is finally over. However, according to recent rumors, the "bad" news is that 'Big Battlemage' isn't an Arc-based gaming product but an AI-focused workstation with an Intel Arc Pro B70 GPU that will ship with 32GB of GDDR6 memory and 608 GB/s of bandwidth.

3

It's set to launch with a $949 MSRP for the reference design, with third-party cards from ARKN, ASRock, Gunnir, Maxsun, Sparkle, and others on the way. Compared to the existing Arc Pro B60, the new Arc Pro B70 features 50% more memory and 60% more Xe cores, bringing the total to 32 Xe cores. And with that, AI performance across a wide range of workloads is said to be up to 60% higher.

In terms of the card's position in the market, Intel is pitting it against the more expensive NVIDIA RTX Pro 4000 graphics card, where, in a single-GPU setup, it offers up to 2X "tokens per dollar" on Linux with 367 TOPS of AI performance. In addition to the Intel Arc Pro B70, the company has announced the Intel Arc Pro B65, which is essentially the Arc Pro B60 with 32GB of VRAM instead of 24GB.

3

There's no price for the Intel Arc Pro B65; however, Intel notes that it's set to arrive in mid-April. "Intel Arc Pro has evolved from powering creative workflows to enabling AI builders at scale," Anil Nanduri, Vice President of AI Products & GTM at Intel, said. "With the Intel Arc Pro B70 and B65, we're delivering high‑performance workstation graphics and AI inference at exceptional value."

And yes, the Intel Arc Pro B70 is 'Big Battlemage' because it maxes out the architecture with 32 Xe cores, 256 XMX engines, 32 Ray Tracing Units, and a power draw between 160W and 290W. Sure, as much as we'd like to see this card arrive in an Intel Arc non-Pro form with 16GB of VRAM to compete in the mid-range PC gaming GPU market, with the current state of memory pricing, it's understandable that the focus has shifted to AI.