Gigabyte AI Top Atom is a compact personal AI supercomputer designed to enable local AI model training, fine-tuning, and inference. Built around the NVIDIA GB10 Grace Blackwell Superchip, each individual hardware unit delivers up to one petaflop of FP4 computing power and features 128GB of coherent unified LPDDR5X memory. At Computex 2026, the platform demonstrated its clustering capabilities by connecting four units together to form a unified four-petaflop workstation system.
—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—
To enable high-speed clustering, the hardware uses an NVIDIA ConnectX-7 network interface card connected to a high-bandwidth switch. This configuration supports 200 GbE bandwidth using fiber ethernet cables terminated with QSFP connectors to ensure low latency and high throughput. Although GIGABYTE officially announces cluster configurations of up to four units, the networking infrastructure allows scaling to eight units if an eight-port switch is deployed, enabling even larger model footprints.
The system is designed to run large language models and other complex computational tasks locally. A single Gigabyte AI Top Atom box can run models with up to 120 billion parameters, while a clustered four-unit setup scales the capacity to handle models containing up to 480 billion parameters. Users can choose to cluster the units together to operate as a single supercomputer or partition them to run separate AI agent workloads concurrently.
A key application demonstrated at the event is the autonomous orchestration of thermal simulations using the open-source OpenClaw agent framework. In this workflow, one Gigabyte AI Top Atom acts as the commander agent running OpenClaw, which manages and automates the simulation tasks. The commander distributes workload instructions to three subordinate units running GROMACS molecular dynamics simulation software to compute thermal interface material properties for 1000W Blackwell processors.
By automating over 100 iterative simulation runs, the OpenClaw agent optimizes the material formulas without manual intervention. This distributed workflow reduces simulation times from a full week on a single unit to just two days across three units, representing at least a threefold speedup. The resulting recipe optimized the heat dissipation channels, converting them into straight lines that allow heat to escape the chip efficiently and improving thermal performance from 22 to 4.2.
