Gigabyte launches AI Top Aero mini PC with 128GB of memory
The new compact desktop combines Nvidia's RTX Spark chip with high-capacity memory to run local AI models and gaming.
The short version
- Powered by the Nvidia RTX Spark N1X chip with 6,144 CUDA cores.
- Features 20 Arm cores (Cortex-X925 and Cortex-A725) and up to 128GB of LPDDR5X memory.
- Runs Windows 11 for Arm, allowing for both local AI workloads and traditional PC gaming.
- Includes high-speed connectivity like 10GbE LAN, Wi-Fi 7, and PCIe Gen 5 SSD support.
Why it matters
This hardware provides a path for users to run powerful AI models privately on their own machines without relying on the cloud. It also positions a small-form-factor PC as a viable alternative to high-end workstations for both creators and developers.
A compact desktop for AI
Gigabyte has introduced the AI Top Aero, a small desktop mini PC built specifically to handle the demands of the modern AI era. The device is designed to bridge the gap between standard computing and heavy machine learning tasks. It aims to provide a complete package for users who want to run sophisticated models without needing a massive tower.
The hardware is centered around the Nvidia RTX Spark chip, which provides the necessary horsepower for various machine learning applications. This chip allows the system to manage complex workloads that typically require significant computing resources. By focusing on local execution, the device appeals to those who prioritize data privacy or want to avoid the costs of cloud-based services.
Despite its small physical footprint, which is approximately one litre in volume, the machine is built for sustained performance. Gigabyte designed the cooling system specifically to manage temperatures during heavy AI inference workloads. This ensures that the hardware can maintain its speed even when processing large amounts of data over long periods.
The architecture of the Spark
Under the hood, the AI Top Aero utilizes a unique combination of 20 Arm cores to drive the operating system and applications. These cores are split between ten Cortex-X925 and ten Cortex-A725 units. This configuration helps the PC handle diverse tasks, from basic background processes to demanding multi-threaded software.
The graphics and AI processing are handled by a 6,144 CUDA-core Blackwell GPU, which the manufacturer notes is equivalent to a desktop RTX 5070. This GPU shares the system's up to 128GB of unified LPDDR5X memory. Unified memory allows the CPU and GPU to access the same pool of data quickly, which is a critical requirement for large language models.
Because the system runs Windows 11 for Arm, it maintains compatibility with a wide range of regular PC applications and games. This distinguishes it from some specialized AI hardware that might be locked into specific environments. Users can switch between developing AI tools and using everyday software on the same machine.
Performance for developers
For those interested in AI development, the Nvidia RTX Spark N1X chip provides the muscle needed for several popular tools. The system can run applications like LM Studio and Ollama, as well as Nvidia Nemotron models. These tools are commonly used to host and interact with large language models locally.
The hardware is particularly well-suited for large prefill and prompt processing. This is due to the high matrix multiplication capabilities of the Blackwell GPU architecture. Matrix multiplication is a fundamental mathematical operation that AI models use to process and generate information from their training data.
However, the hardware has specific trade-offs when compared to other high-end systems like the Mac Studio. While the RTX Spark excels at certain processing types, the Mac Studio may remain faster at token generation. This is because the Mac Studio benefits from different memory bandwidth characteristics that favor rapid sequential output.
Gaming and content creation
While the primary marketing focus is on AI, the AI Top Aero is also a capable machine for high-end gaming and professional work. The Blackwell GPU supports the same feature set found in gaming GeForce cards. This includes DLSS upscaling and DLSS 5 neural rendering, which use AI to improve image quality and frame rates.
Content creators can also leverage the CUDA cores to improve their productivity in compatible software. These cores are standard in many professional rendering and video editing tools. By using this hardware, creators can achieve faster render times while keeping their workspace compact and organized.
The PC is rounded out with high-end connectivity options to ensure it can handle large files and fast networks. It includes support for up to 4TB of PCIe Gen 5 SSD storage for rapid data access. Additionally, it features USB-C connections, fast 10GbE LAN, and the latest Wi-Fi 7 standard for high-speed internet connectivity.
What to watch
- How the AI Top Aero compares to the Apple Mac Studio in real-world token generation speeds.
- The availability of new Nvidia Nemotron models for local deployment on the RTX Spark chip.
Try it
Check if your current AI workflow requires high memory bandwidth or high matrix multiplication to see if this hardware fits your needs.
Sources and links
- ReportingClub386: the article we sourced
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by Club386 · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
