← News
Hardware

GIGABYTE announces AI TOP AERO: up to 128GB unified memory is not a verified GPU budget

The November 10 announcement specifies RTX Spark N1X, a 20-core Grace CPU, Blackwell RTX GPU and an approximately one-liter chassis. Regional configurations, pricing and purchase availability remain unverified; other systems’ benchmarks do not apply.

Source: GIGABYTE via PR Newswire

GIGABYTE announced AI TOP AERO through PR Newswire on November 10 at 14:00 UTC, inside this window. This is a new desktop announcement, not the first launch of the RTX Spark platform. The Windows 11 system uses NVIDIA's RTX Spark N1X superchip, combining a 20-core Grace CPU and a Blackwell RTX GPU.

  1. Separate maximum configuration from an actual unit. Advertised configurations reach up to 128GB LPDDR5x unified memory and a 4TB PCIe Gen5 SSD. This does not specify every SKU, nor the GPU allocation available under the target operating system. AraoAI creates a separate desktop dictionary entry but leaves VRAM blank rather than equating maximum system memory with GPU-usable memory.

  2. Platform numbers do not transfer to a new desktop. The approximately one-liter chassis, 10GbE, Wi-Fi 7 and USB Type-C describe form and connectivity. Sustained behavior depends on cooling, power policy and software. The announcement contains no independently verifiable inference measurements, memory-bandwidth specification or power figure. Results from another N1X laptop or development kit cannot fill those gaps or establish a GPU ranking.

  3. Announcement is not universal retail availability. The release directs readers to regional announcements for configuration, prices and availability. We leave purchase date and price unverified. Buyers should first identify the exact SKU, memory-allocation policy, drivers and runtime versions, then confirm stable service at the required context length.

CUDA, Nemotron, LM Studio, Ollama and ComfyUI are mentioned as ecosystem directions, not proof that every tool and model has been independently validated. The verified facts are the announcement and its stated specifications. A useful next report needs actual free memory after loading, model quantization, context, prefill, decode and runtime build. Without these, 128GB is not a performance conclusion.

Publisher announcement: https://www.prnewswire.com/news-releases/gigabyte-expands-its-local-ai-portfolio-with-the-launch-of-ai-top-aero-powered-by-nvidia-rtx-spark-302934919.html