Intel's Arc Pro B70 shows gains

- Intel’s Arc Pro B70 workstation GPU, tested with 32GB of VRAM, was reported to be roughly twice as fast as the Arc B580 on average in gaming and prosumer workloads. - Benchmarks say the B70 beats Nvidia’s RTX 5060 Ti in some titles and can be up to ~66% faster than the previous Arc B580 in specific tests. - Stronger mid‑tier/workstation GPUs raise baseline expectations for local AI development and prosumer workflows that Apple silicon teams must contextualise against unified memory and efficiency advantages. (tomshardware.com) (club386.com)

Graphics cards are getting weird in a useful way. Intel’s Arc Pro B70 is a workstation card first, but the new gaming tests matter because they show what Intel’s bigger Battlemage silicon can really do when it isn’t cut down. The gap was obvious before — Intel had smaller Battlemage cards for mainstream buyers, but not the fuller-fat version gamers kept waiting for. Now a few independent runs are giving people a back-door look at that missing chip, and the result is simple: it’s a lot faster than the Arc B580, and much closer to Nvidia’s RTX 5060 Ti than a “pro” badge would suggest. ### What is the Arc Pro B70, exactly? The B70 is Intel’s new workstation GPU built on the Xe2 “Battlemage” architecture, using the larger BMG-G31 die rather than the smaller chip inside cards like the Arc B580. Intel positions it for AI, rendering, and high-end creative work — not gaming — and gives it 32GB of GDDR6 on a 256-bit bus, 32 Xe cores, 256 XMX engines, and up to 367 TOPS for AI workloads. That memory capacity is the headline feature, because it lets the card hold bigger models and heavier scenes than most midrange gaming GPUs can manage. ### Why are people testing it in games? Because this is basically the closest thing we’ve seen to a true big Battlemage gaming card. Intel never launched a consumer Arc card with this exact configuration, so the B70 ends up answering the question by accident — what if the B580 had a much bigger sibling? That makes every game benchmark feel like a preview of the product Intel didn’t quite ship. ### So how much faster is it? The cleanest takeaway is that the B70 is meaningfully ahead of the B580, but not a giant-killer. One roundup of the first full gaming review says it averaged 36% faster than the B580 in games and 46% faster in 3DMark. Another summary of the same test set says the lead over B580 was about 40% in raster and could stretch to roughly 65% in ray tracing-heavy results. That is a big jump, but it is not the “2x on average” claim floating around in early chatter. ### Does it really beat the RTX 5060 Ti? Sometimes — but only in the narrow sense that matters to benchmark nerds. In the tested games, the B70 won some individual titles and looked especially strong in ray tracing, where one summary put it 1% ahead of the RTX 5060 Ti on average across the sampled RT results. Across all tested raster and RT results together, though, the 5060 Ti still finished about 2.9% ahead overall. So the honest read is “trades blows,” not “beats Nvidia.” ### Why does the 32GB matter more than the FPS? Because workstation and local-AI buyers care about fitting the workload before they care about squeezing out 8 more frames. A 16GB gaming card can be faster in some scenes and still be less useful for model loading, heavier video timelines, or large textures. Intel’s whole pitch for the B70 is that it can run advanced AI models, scale across multiple GPUs, and bring workstation certifications on top of that memory headroom. Basically, the card is selling capacity and flexibility as much as speed. ### Is there a catch? Yes — price, drivers, and category confusion. The B70 is a pro card, so buyers are paying for memory, certifications, and reliability features, not just frame rates. It also draws up to 230W, which is not outrageous, but it means this is not some tiny efficiency miracle. And because these gaming numbers come from a small set of early tests, they are more useful as a signal than as a final verdict. ### Why does this matter beyond Intel? Because it raises the floor for what a mid-tier “workstation” GPU can look like in 2026. If Intel can put 32GB on a card that hangs near the RTX 5060 Ti in games while also targeting AI and creator work, then the baseline for local model tinkering and prosumer workflows just moved up a notch. Apple, Nvidia, and AMD still each have their own advantages — unified memory, software ecosystems, or raw gaming performance — but Intel just made the middle of the market more interesting. ### Bottom line The Arc Pro B70 is not secretly the best gaming GPU in its class. But it does show that Intel’s bigger Battlemage chip has real legs — especially when you pair decent performance with 32GB of VRAM. That combination is the story.

Get your own daily briefing

Scout delivers personalized news, insights, and conversations tailored to your role and industry.

Download on the App Store

Shared from Scout - Be the smartest in the room.