InspireSemi has introduced the profitable tapeout of the Thunderbird I Accelerated Computing chip for fabrication at TSMC.
This extremely differentiated “supercomputer cluster-on-a-chip” options 1,536 customized 64-bit RISC-V CPU cores, tailor-made for high-level scientific computing and complicated knowledge processing.
Thunderbird I is designed to cater to a big selection of compute-intensive purposes, from AI and machine studying to graph analytics. Leveraging the open customary RISC-V CPU ISA, it permits simpler improvement and integration into present expertise frameworks, with entry to a strong ecosystem of software program, libraries, and instruments.
PCIe add-in card deliberate
The chip’s structure integrates a high-speed mesh community cloth that gives substantial bandwidth and minimal latency communication amongst cores, vital for purposes that rely on synchronized operations throughout a number of threads. This environment friendly community integration manages interactions inside the chip’s core array and reminiscence programs, making certain optimum efficiency with out the widespread bottlenecks.
The upcoming product launch will embrace a server PCIe add-in card internet hosting 4 Thunderbird chips, offering over 6,000 interconnected 64-bit CPU cores. This setup is supplied to deal with double-precision math, important for a lot of high-performance computing purposes in fields like local weather science, medical analysis, and complicated simulations.
Ron Van Dell, CEO of InspireSemi, mentioned, “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 phrase on pricing but, nonetheless.
InspireSemi additionally stresses Thunderbird I’s power effectivity, a carryover from its preliminary design for energy-sensitive blockchain computing purposes. The firm says this method affords more environmentally pleasant various to conventional knowledge middle GPUs.