GigaIO, SourceCode develop Gryf AI supercomputer in a cabin luggage form factor
GigaIO, a company specializing in AI infrastructure, has joined forces with SourceCode, a provider of compliant IT systems, to create an edge solution–the Gryf AI supercomputer. This airline carry-on luggage-sized computer is designed to meet the needs of modern AI workloads at the edge.
GigaIO claims that Gryf’s performance can be scaled through its FabreX AI memory fabric, which integrates computing, storage, and communication IOs into a single cluster fabric. This architecture creates a unified, software-defined, and memory-centric infrastructure that is highly adaptable.
The Gryf supports server-class GPUs and can adjust its hardware configuration to meet the specific requirements of different workloads. This is made possible through FabreX’s ability to disaggregate and re-aggregate compute resources such as GPUs. Its capabilities enable it to handle vast amounts of data–over a petabyte.
“Gryf’s novel architecture, made possible by FabreX, our AI memory fabric, provides those customers the advanced compute, storage, and GPU capabilities they crave in today’s sensor-rich edge environments,” says Alan Benjamin, chief executive officer of GigaIO.
Gryf has a composable system that offers a high degree of adaptability to meet diverse customer needs. Its on-demand reconfiguration feature increases resource utilization, reduces hardware expenses, lowers power consumption, and improves availability.
The company claims that Gryf achieves maximum compute density by disaggregating GPUs and optimizing its ventilation system, ultimately enhancing performance and operational efficiency.
In terms of scalability, the Gryf units can be interconnected up to five, enhancing the system’s capability to handle large-scale, petabyte-size datasets. The interconnected units can share compute, storage, and other sleds, making them flexible in resource allocation.
“Engineering this level of power into such a compact form factor for mission-critical applications is a massive achievement,” says Parag Shah, chief technology officer and chief operations officer of Edge and Embedded AI at SourceCode.
Gryf has been made available for early access, with delivery of these units scheduled for June. According to the company, this phase will allow for real-world testing and feedback, which is crucial for refining the system ahead of broader deployment.
Comments