The Midrange GPU War Just Got Complicated
Intel’s Arc B770 was not supposed to be a serious threat. When Intel first entered the discrete GPU market, the Arc A-series cards earned a reputation for driver instability, confusing performance in older DirectX titles, and a general sense that the company was learning in public. The B770 changes that narrative in ways that will make Nvidia uncomfortable, especially at the $300-$350 price point where the RTX 5060 Ti is supposed to dominate without much argument.
The RTX 5060 Ti launched with Nvidia’s full suite of advantages: DLSS 4, Frame Generation, and a brand recognition that effectively functions as a pricing floor. Nvidia knows that a significant portion of buyers will pay a premium just for the green logo and the ecosystem that comes with it. But raw rasterization performance is harder to brand your way out of, and that is exactly where the B770 is applying pressure in 2025.
Specs on Paper vs. Specs in Practice
The Arc B770 is built on Intel’s Xe2 architecture, the same foundation as the Battlemage mobile chips, and it carries 20 Xe2 cores alongside 16GB of GDDR6 memory on a 256-bit bus. That memory bandwidth advantage over the base RTX 5060 Ti configuration is not subtle. Nvidia’s 5060 Ti ships with either 8GB or 16GB depending on the variant, and the 8GB version has already drawn criticism for being undersized at the exact resolutions buyers in this tier are targeting.
Nvidia counters with a higher core count in shader throughput terms and the architectural maturity that comes from being on its fourth major generation of Ada and Blackwell refinements. The GB206 chip powering the 5060 Ti brings Blackwell’s architectural improvements to the mid-tier, including better ray tracing hardware and the Transformer Model-based DLSS 4 upscaling. On paper, these are real advantages. In practice, whether they matter depends entirely on what you are playing and at what settings.
Benchmark Performance Across Key Titles
In Cyberpunk 2077 at 1440p with ray tracing disabled, the B770 trades blows with the RTX 5060 Ti within margins that are essentially noise. Both cards sit in the 85-95 fps range depending on system configuration, and the B770 occasionally edges ahead in raw rasterization benchmarks when DLSS is removed from the equation. That is a genuine result. Intel’s XeSS 2 upscaling has also matured considerably, and in titles where it is properly implemented, the quality gap between XeSS 2 and DLSS 4 is narrower than it was a generation ago.
Ray tracing is where the gap opens up. Turn on path tracing in Alan Wake 2 or enable RT reflections in Hogwarts Legacy, and the RTX 5060 Ti pulls ahead by margins that start to matter – 15 to 25 percent depending on the scene. The B770’s ray tracing hardware is improved over the Arc A770, but Nvidia still holds a structural lead in RT workloads, and DLSS 4’s Frame Generation option essentially doubles perceived frame rates in supported titles. Intel’s XMX AI engines can run XeSS but Frame Generation support in Arc drivers remains more limited by comparison.
Elden Ring, Counter-Strike 2, and Call of Duty: Warzone are all titles where the B770 performs at or above the RTX 5060 Ti at 1440p. Competitive shooters running on older or well-optimized engines tend to favor raw shader throughput and memory bandwidth, and the B770’s 16GB buffer means it does not run into the texture streaming issues that have been documented on the 8GB 5060 Ti in some open-world titles. For a player whose library skews toward esports and non-RT games, the value math genuinely favors Intel right now.
The situation at 4K changes the calculation again. At 3840×2160, the RTX 5060 Ti with DLSS 4 Quality mode can maintain playable frame rates in demanding titles where the B770 starts to struggle with raw pixel throughput. DLSS 4’s reconstruction quality at 4K is still ahead of XeSS 2 in side-by-side comparisons, though the gap has closed enough that casual players are unlikely to notice without freeze-frame analysis.
Driver Maturity and Software Ecosystem
Intel’s driver situation has improved dramatically since the A770 launch, but it is not yet at Nvidia’s level of refinement. Game compatibility is broader, day-one driver support has become more reliable, and the Arc Control software is functional rather than frustrating. Some DX9 and DX11 titles that caused issues on the A-series have been patched or resolved, though edge cases still exist. For players running a modern library of DX12 and Vulkan titles, this is largely a non-issue.
Nvidia’s GeForce Experience and driver pipeline remains the industry benchmark for software support. Reflex, DLSS, Broadcast, and the broader Nvidia app ecosystem all integrate tightly, and for streamers or content creators who use GPU-accelerated encoding, NVENC still outperforms Intel’s Arc equivalent in quality at equivalent bitrates. These are real considerations that matter to a subset of buyers, even if they do not show up in raw gaming benchmarks. The RTX 5060 Ti’s rasterization standing against AMD’s RX 9060 XT already showed Nvidia is under pressure from multiple directions in this tier.
Price, Availability, and the Real Decision
The B770 is targeting a street price around $299-$329, depending on partner board configurations. The RTX 5060 Ti 16GB variant has been hovering around $449-$499 at launch due to supply constraints, while the 8GB version sits closer to $379-$399. At those numbers, the value argument for the B770 is not close – it is decisive, assuming the buyer’s game library aligns with the card’s strengths.
The 8GB 5060 Ti is the version most buyers will encounter first, and it is the version that makes the B770 look its best. A 16GB Arc card competing against an 8GB Nvidia card at a $100-$150 premium gap is not a close fight in rasterization workloads at 1440p. If Nvidia’s 16GB 5060 Ti stabilizes at its MSRP over the coming months, the gap narrows on paper but the B770 still undercuts it on price.
Where this gets unresolved is longevity. Arc drivers have historically required patience, and Intel’s commitment to long-term driver support at the scale Nvidia provides is still being demonstrated rather than proven. Buyers who kept their RTX 20-series cards through four years of driver updates had a clear support track record to rely on. B770 buyers in 2025 are making a partial bet on Intel following through over a similar span.
The B770 is a better card than the midrange market expected Intel to produce this cycle, and in a straight rasterization fight at 1440p without RT or DLSS in play, it wins or ties more often than Nvidia would like to admit. The question Intel cannot answer yet is what happens when a game releases two years from now with heavy DLSS 4 Frame Generation integration and no XeSS equivalent. That gap might not exist, or it might matter enormously – and there is no benchmark that settles it today.
