ASUS ProArt P16 review: Nvidia RTX Spark debuts in a costly AI laptop

Engadget reports that the ProArt P16 starts at $2,800, with prices rising sharply as memory is added; the review configuration costs $7,000, largely because it carries 128GB of RAM. The laptop can play games well, the review says, but it costs far more than a dedicated gaming laptop. Its appeal is strongest for people running local AI models, since it fits up to 128GB of unified memory into a relatively slim case.

The review describes a sturdy aluminum case, a bright 16-inch OLED screen, a spacious keyboard and a large haptic touchpad. At 3.9 pounds, the ProArt P16 is lighter than Apple’s 16-inch MacBook Pro, which weighs 4.7 pounds, according to the review. ASUS includes three USB-C ports, USB Type A, HDMI, a headphone jack and a full-sized SD card reader; Ethernet is absent.

Inside, the RTX Spark system-on-a-chip combines a 20-core Arm CPU, a Blackwell GPU and up to 128GB of unified memory, a design similar to Apple’s M-series chips. Engadget says RTX Spark is not intended to replace Intel and AMD x86 hardware but to create a new branch of PC design for AI researchers and enthusiasts, alongside Qualcomm’s Snapdragon X chips for entry-level and midrange systems.

Engadget’s benchmarks were mixed. The ProArt P16’s PCMark 10 Applications score was 3,000 points lower than an ultra-slim ASUS A16 with Qualcomm’s Snapdragon X2 Elite that the publication tested earlier this year, and it lagged slightly in Geekbench 6’s single-threaded CPU test. The review says that suggests Qualcomm’s chips have more capable CPUs. The RTX Spark machine’s GPU, NPU and unified memory stood out: it scored 4,700 points higher than a MacBook Pro with M5 Max in Geekbench 6’s GPU benchmark, 3,895 versus 3,267 in UL Procyon’s AI Computer Vision 2 test, and 1,519 in Procyon’s AI Image Generation test, 600 points above Apple’s laptop.

In use, Engadget says the ProArt P16 felt as fast as an M-series MacBook Pro, downloading several large files at once without delay. The review says it ran OpenClaw and the large Qwen3.8 27B model, which alone takes 42GB of RAM, while leaving substantial memory for other tasks. For a $7,000 system with 128GB of RAM, the review adds, that speed is expected.

Engadget says RTX Spark may still benefit Windows PCs by increasing competition among Nvidia, AMD, Intel and Qualcomm, but it is unlikely to bring PC prices back to pre-AI-boom levels. The review notes that RTX Spark is another memory-hungry platform at a time when AI demand has already driven RAM prices sharply higher.