Huawei has announced its SuperPoD interconnect technology, designed to link thousands of chips, including its Ascend AI chips, to create powerful computing clusters. This move arrives as China aims for technological self-reliance amid restrictions on access to advanced chips from companies like Nvidia.
The SuperPoD architecture allows multiple physical machines to operate as a single, logical unit for AI tasks. Huawei's Atlas 950 SuperPoD, featuring 8,192 Ascend NPUs, and the Atlas 960 SuperPoD, with 15,488 Ascend NPUs, deliver industry-leading performance in computing power, memory capacity and interconnect bandwidth. Huawei also plans to launch upgraded Ascend AI chips over the next three years.
Huawei's interconnect technology, UnifiedBus 2.0, addresses the challenges of linking numerous chips over long distances while maintaining high-speed, low-latency connections. The company hopes industry partners will adopt the protocol to foster an open ecosystem. Huawei also introduced the TaiShan 950 SuperPoD, a general-purpose computing SuperPoD.
