Supercomputer-on-a-chip goes live: single PCIe card packs more than 6,000 RISC-V cores, with the ability to scale to more than 360,000 cores — but startup still remains elusive on pricing

InspireSemi has announced the successful tapeout of the Thunderbird I Accelerated Computing chip for fabrication at TSMC. 

This highly differentiated "supercomputer cluster-on-a-chip" features 1,536 custom 64-bit RISC-V CPU cores, tailored for high-level scientific computing and complex data processing.

Thunderbird I is designed to cater to a wide array of compute-intensive applications, from AI and machine learning to graph analytics. Leveraging the open standard RISC-V CPU ISA, it allows easier development and integration into existing technology frameworks, with access to a solid ecosystem of software, libraries, and tools.

PCIe add-in card planned

The chip's architecture integrates a high-speed mesh network fabric that provides substantial bandwidth and minimal latency communication among cores, important for applications that rely on synchronized operations across multiple threads. This efficient network integration manages interactions within the chip's core array and memory systems, ensuring optimal performance without the common bottlenecks.

The upcoming product release will include a server PCIe add-in card hosting four Thunderbird chips, providing over 6,000 interconnected 64-bit CPU cores. This setup is equipped to handle double-precision math, essential for many high-performance computing applications in fields like climate science, medical research, and complex simulations.

Ron Van Dell, CEO of InspireSemi, said, “We are proud of the accomplishment of our engineering and operations team to finish the Thunderbird I design and submit it to our world-class supply chain partners, TSMC, ASE, and imec for production. We expect to begin customer deliveries in the fourth quarter.” There's no word on pricing yet, however.

InspireSemi also stresses Thunderbird I’s  energy efficiency, a carryover from its initial design for energy-sensitive blockchain computing applications. The company says this approach offers more environmentally friendly alternative to traditional data center GPUs.

More from TechRadar Pro

• Intel's formidable 288 core CPU now has a proper family name
• World's first wireless CPU prototype could lead to seamless modular computing
• 'Virtually unhackable' chip could make GPU much faster at AI