NVIDIA RTX Spark laptops ship today with unified memory and local AI
Unified memory and local AI in a laptop — but discrete GPUs still rule for pure FPS
NVIDIA RTX Spark laptops are available today from six major OEMs. The new platform combines RTX graphics and NVIDIA's AI stack into a single superchip, with up to 128GB of unified memory and 1 petaflop of FP4 AI performance for running large language models locally.
The unified memory architecture is the big shift here — no split VRAM/RAM pools like traditional discrete GPU setups. That 128GB ceiling is shared across graphics, AI workloads, and system tasks. FP4 AI throughput hits 1 petaflop, targeting local LLM inference and AI agents without cloud dependencies.
NVIDIA frames this as a creation/AI/gaming hybrid. The gaming angle is less clear — no dedicated VRAM allocation mentioned, and unified memory designs often trade raw GPU bandwidth for flexibility. If you're chasing max FPS in competitive shooters, a traditional discrete RTX 50-series GPU with dedicated GDDR7 is still the safer bet. Spark looks more suited to content creators who also game, not the other way around.
Check BetterFPS's optimization guides to see if your current laptop still hits competitive frame targets before upgrading.