Source URL: https://www.theregister.com/2024/12/04/riscv_ai_champion_investment/
Source: The Register
Title: RISC-V’s AI champion just scored $693M cash infusion
Feedly Summary: Architecture may be good for more than cut-rate SoCs and high-volume microcontrollers after all
RISC-V chip designer Tenstorrent has won $693 million of investment – an endorsement of its plans to use the permissively licensed instruction set architecture for workloads like AI.…
AI Summary and Description: Yes
Summary: The text discusses Tenstorrent’s significant investment and its utilization of the RISC-V instruction set architecture for AI workloads. It highlights the potential of RISC-V in AI infrastructure and how Tenstorrent’s Blackhole accelerators outperform existing offerings from Nvidia, demonstrating the innovation and competition in the AI hardware space.
Detailed Description:
– Tenstorrent, a startup focusing on AI infrastructure and led by semiconductor expert Jim Keller, has secured $693 million in funding, indicating strong investor support for its plans.
– The company is leveraging the RISC-V instruction set architecture (ISA), which is permissively licensed and allows for both open-source and proprietary developments without royalty fees.
– RISC-V has seen broad adoption since its introduction in 2010, presenting itself as an attractive alternative to proprietary ISAs from companies like Arm, especially for microcontrollers and SoCs.
– Tenstorrent’s Blackhole accelerators consist of 768 RISC-V cores, including 16 high-performance cores for Linux operations and 752 smaller cores designated for memory and system management, showcasing a unique assembly for maximizing AI compute capabilities.
– The Blackhole architecture is engineered for peak performance of up to 745 teraFLOPS at FP8, positioning it slightly ahead of Nvidia’s existing solutions in terms of speed, although with less memory capacity.
– The accelerators will be integrated into Tenstorrent’s Blackhole Galaxy appliance, promising a significant total compute performance of 23.8 petaFLOPS.
– The funding garnered will not only enhance hardware development but also support the creation of an AI development cloud and an open-source software stack, indicating a commitment to a comprehensive AI ecosystem.
– Tenstorrent’s proactive approach contrasts with competitors like SiFive and Google, which also utilize RISC-V but focus more on licensing and custom chip design rather than full-scale hardware production.
Overall, this development reflects growing trends in AI hardware using open architectures, offering insights for security and compliance professionals about emerging technologies interpreting RISC-V’s role in secure AI infrastructure and the implications of its open licensing model.