ASUS Brings NVIDIA RTX Spark to ProArt PCs With Up to 128GB of Unified Memory

ASUS is bringing NVIDIA's new RTX Spark platform to its ProArt lineup, putting unusually large pools of unified memory and serious local AI performance into two creator laptops and a compact mini PC.

The new ProArt P16, ProArt P14 and ProArt GR1X combine a Grace CPU, Blackwell RTX GPU and unified memory architecture. ASUS says configurations can reach 128GB of unified memory and up to 1 petaflop of FP4 AI performance.

The pitch is straightforward: run larger AI models, creative generation pipelines and agentic workflows locally instead of sending every task to the cloud.

Three ProArt PCs built around RTX Spark

The ProArt P14 weighs around 1.48kg and offers up to a 14-inch 3K 120Hz Lumina Pro OLED display. The ProArt P16 moves to a 16-inch chassis weighing around 1.77kg, with configurations offering up to a 4K 120Hz Lumina Pro OLED display.

Both laptops can be configured with as much as 128GB of unified memory.

The ProArt GR1X puts RTX Spark into a compact 1.1-liter desktop measuring roughly 150 x 150 x 51mm. It supports 64GB or 128GB of unified memory, dual M.2 SSDs, 10GbE, Wi-Fi 7 and as many as four 4K displays.

See ASUS' RTX Spark ProArt lineup

Why 128GB of unified memory matters

The headline specification isn't simply raw GPU performance. It's memory.

Traditional PCs normally divide system RAM and dedicated GPU VRAM into separate pools. Large local AI workloads can quickly hit the GPU's VRAM limit even when plenty of system memory remains available.

RTX Spark instead uses unified memory shared across the CPU and GPU. With configurations reaching 128GB, ASUS says the machines can handle larger language and vision-language models, local image and video generation, LoRA customization, agentic workflows and complex creative projects.

That doesn't automatically mean every 128GB workload will be fast. Model architecture, quantization, context length, memory bandwidth and software optimization still matter. But having that much addressable memory dramatically expands what can fit locally.

Up to 1 petaflop of AI performance

RTX Spark combines an NVIDIA Grace CPU with a Blackwell RTX GPU and fifth-generation Tensor Cores. ASUS rates the platform at up to 1 petaflop of FP4 AI performance in its highest configurations.

It also retains one of NVIDIA's biggest advantages for developers: CUDA.

ASUS is positioning these machines for local language models, image and video generation, creative applications, development environments and autonomous AI agents.

ASUS wants AI agents running on your own PC

ASUS is integrating its ProArt software with tools including ComfyUI and Hermes Agent from Nous Research.

Hermes can work across applications, terminal environments and local files, while ComfyUI provides visual workflows for image and video generation. ASUS' MuseTree application also supports local generative workflows.

The local-first approach offers obvious benefits: privacy, offline capability and freedom from per-token or per-generation cloud charges.

Cloud AI isn't disappearing. Frontier-scale models will still make sense remotely, and ASUS describes the platform as capable of local, cloud or hybrid workflows.

Pricing starts around ₱163,123 (US$2,599.99)

These aren't inexpensive creator PCs.

ASUS says its RTX Spark ProArt laptops start at approximately ₱163,123 (US$2,599.99) based on the USD-to-PHP conversion at publication time. The peso figure is a conversion, not an official Philippine MSRP.

The base ProArt P14 starts at approximately ₱163,123 (US$2,599.99) in the United States, while the ProArt P16 starts at roughly ₱175,671 (US$2,799.99) using the same conversion basis.

At the top end, a 128GB ProArt P16 configuration has been listed for as much as roughly ₱439,178 (US$6,999.99), putting these systems firmly into workstation territory.

Philippine buyers have a more useful local reference for the desktop model. A local ASUS representative told REVU Philippines that the 64GB ProArt GR1X is expected to start at around ₱260,000 (US$4,160) when it reaches the Philippines.

Unlike the converted laptop figures, that ₱260,000 number is a locally reported expected price, although final retail pricing can still change before launch.

Read REVU Philippines' look at the ProArt GR1X

When can you buy them?

Global preorders for the RTX Spark ProArt laptops opened on October 7, 2026.

ASUS says the first U.S. units are expected from October 16. Higher-capacity 64GB and 128GB configurations will primarily be sold through selected retailers and ASUS' own store.

The ProArt GR1X is scheduled to arrive later in 2026, primarily in 64GB and 128GB configurations.

Philippine availability for every laptop configuration hasn't been announced, so local buyers should treat converted U.S. prices as references rather than expected Philippine retail prices.

Who are these machines actually for?

The pricing makes the target audience clear.

These aren't mainstream laptops designed primarily for browsing, office work or occasional AI features.

A 128GB unified-memory machine makes more sense for AI developers running large local models, creators using demanding generative pipelines, studios handling sensitive client assets, researchers experimenting with local inference, or businesses that want predictable local compute rather than recurring cloud usage.

The GR1X could be particularly interesting as an always-on local AI server. Its small footprint, 10GbE networking and high memory ceiling make it suitable for hosting local models or agents that other machines on a network can access.

The bigger shift: AI PCs are becoming workstations

The first wave of AI PCs largely focused on NPUs and features such as transcription and lightweight assistants. RTX Spark represents something different.

Instead of asking whether a PC has an AI accelerator, the question becomes whether the machine can run genuinely large models and complex AI workflows entirely on-device.

Up to 128GB of unified memory, CUDA compatibility and Blackwell-class graphics put ASUS' new ProArt systems much closer to portable AI workstations than conventional AI PCs.

They're expensive. But for developers and creators already spending heavily on cloud inference or constrained by conventional GPU memory, that price has a different comparison point.

The real test will be independent performance.

For now, ASUS has made one thing clear: local AI on Windows is getting much bigger than an NPU.