Samsung Exynos 2600: Xclipse 960 RDNA 4 graphics, ENSS, and 2 nm GAA

Samsung is moving its in‑house mobile graphics up a gear. The new Exynos 2600 is the first SoC to use a custom take on AMD’s RDNA 4—not a direct port, but a heavily reworked version tailored for phones and manufactured on a 2 nm GAA process. It’s a notable shift that hints at longer‑term ambitions for graphics under Samsung’s own roof.

At the heart of the Xclipse 960 GPU is the MGFX4 architecture—a scaled and modified iteration of RDNA 4 that replaces MGFX3 from the Exynos 2500. Samsung says the new chip doubles compute performance over the previous generation, while ray tracing performance is up by 50%. The gains stem from reworked compute blocks and better handling of AI workloads, a combo that should matter just as much in games as in increasingly AI‑heavy apps.

The Exynos 2600 carries eight WGPs, each bundling two compute units. The GPU tops out at 980 MHz—slightly below the Xclipse 950—but the architectural upgrades appear to make up for the lower clock. There’s also ENSS—Exynos Neural Super Sampling, Samsung’s upscaling tech akin to DLSS and FSR—which uses AI to boost image quality without a serious performance hit. On paper, that’s exactly the kind of assist modern mobile gaming needs.

Even so, the Exynos 2600 still trails Qualcomm. In Geekbench 6, the Xclipse 960 lags the Adreno 840 in the Snapdragon 8 Elite Gen 5 by roughly 10–20%, especially in Vulkan. Qualcomm’s edge comes from more compute units and higher frequencies. That said, the gap no longer looks as stark as it did in past generations, and that alone changes the tone of the rivalry.

The Exynos 2600 is expected in the Galaxy S26 and S26+, and Samsung has announced a new Heat Pass Block cooling system that cuts thermal resistance by 16 percent. How well this translates to real‑world gaming will become clear closer to launch, preliminarily slated for March 2026. For now, the direction looks promising—and more competitive than before.