Gigabyte AI TOP AERO features 20-core Grace CPU and 128 GB memory
★ Add us as a preferred source on GoogleTL;DR
Gigabyte announced the AI TOP AERO desktop on November 10, which uses the NVIDIA RTX Spark N1X Superchip and a one-liter chassis for local AI workloads and gaming.
What changed
- The processor features ten Cortex-X925 and ten Cortex-A725 cores, while the integrated graphics unit contains 6,144 CUDA cores comparable to a desktop RTX 5070. Club386 ↗
Gigabyte announced the AI TOP AERO desktop on November 10. The machine uses the NVIDIA RTX Spark N1X Superchip to handle local AI development, content creation, and gaming. It fits into a chassis of approximately one liter.
The system combines a 20-core NVIDIA Grace CPU, using ten Cortex-X925 and ten Cortex-A725 cores, with an NVIDIA Blackwell GPU. This GPU has 6,144 CUDA cores, which is equivalent to a desktop RTX 5070. The hardware uses up to 128 GB of unified LPDDR5x memory and can include up to 4 TB of PCIe Gen 5 SSD storage. It runs Windows 11 for Arm.
Gigabyte says the compact design includes cooling to help sustain performance during heavy AI inference or professional workloads. For connectivity, the desktop provides USB Type-C, 10GbE Ethernet, and Wi-Fi 7.
The platform supports NVIDIA DLSS 5 neural rendering for supported games. For AI tasks, the device is compatible with NVIDIA Nemotron models, LM Studio, and Ollama. Users can also use ComfyUI, NVIDIA Broadcast, and RTX Video for content creation. Gigabyte says the system can join a local network via NVIDIA PAIR to distribute inference tasks to other devices.
Gigabyte AI TOP AERO at a glance
- Processor
- 20-core NVIDIA Grace CPU
- Graphics
- NVIDIA Blackwell GPU
- Memory
- up to 128 GB unified LPDDR5x
- Storage
- up to 4 TB PCIe Gen 5 SSD
- Connectivity
- USB Type-C, 10GbE Ethernet, Wi-Fi 7
- Chassis
- approximately one liter
More on NVIDIAAI in gaming and hardwareGigabyte
Via Gigabyte ↗ · Also from TechPowerUp ↗, Club386 ↗ · Other coverage: Notebookcheck ↗, VideoCardz ↗, Wccftech ↗, TweakTown ↗
Spot a mistake? Tell us


