RRAM and RISC V based AI inference chip by TetraMem
Andes Technology and TetraMem are in for a strategic partnership to develop AI inference chip. AI inference chip for AI edge computing to advance computing in autonomous vehicles, smart cities, healthcare, cybersecurity, and entertainment by having on-device AI computing capability .
TetraMem, analog memristor tech and in-memory computing expert has licensed Andes RISC-V NX27V vector CPU with ACE (Andes Custom Extension) to create a cutting-edge solution that addresses the challenges of AI processing in power-constrained environments.
Both companies are collaborating to combine Andes' RISC-V Vector CPU with TetraMem's an analog RRAM based in-memory computing architecture through ACE to enable tight coupling for the best performance and a fast, energy-efficient AI inference breaking the memory wall and typical "Moore’s Law” constraints.
Features of the AI Accelerator Chip:
1. RISC-V Vector CPU Excellence: Andes RISC-V Vector CPU cores are known for their exceptional performance, efficiency, and configurability, making them ideal for a wide range of AI and edge computing applications. The addition of Andes powerful vector processor brings unparalleled performance capabilities to the accelerator chip.
2. Analog In-Memory Computing Prowess: TetraMem's unique, analog in-memory computing technology empowers the chip with massively parallel VMM computation without data movement, mitigating the energy overhead of conventional architectures as confirmed in TetraMem’s first commercially manufactured demonstration chip.
3. Energy-Efficient AI Acceleration: The joint effort aims to create a chip that is not only powerful but improves energy-efficient by at least an order of magnitude. By optimizing computations and eliminating transfer of weight data, the planned chip will significantly extend the battery life of edge devices and impose a near-zero impact on thermal budgets.
4. Flexible and Scalable: The AI accelerator chip will be designed from 22nm and beyond, to 7nm and below in the future, with versatility and scalability in mind for easy integration into various AI-powered products and applications. This adaptability ensures broad industry applicability. The TetraMem founding team has demonstrated scalability of the compute memristor to 2nm and below, ensuring a roadmap to future-proof solutions.
Mr. Frankwell Lin, Chairman and CEO of Andes Technology, expressed his enthusiasm for the partnership, saying, "Our collaboration with TetraMem represents a significant milestone in the advancement of AI accelerators. By combining Andes' world-class RISC-V vector processing technology with TetraMem's groundbreaking analog in-memory computing, we are poised to deliver a revolutionary solution that will empower the next generation of AI applications."
Dr. Glenn Ning Ge, CEO of TetraMem Technologies, echoed this sentiment, stating, "TetraMem's analog-RRAM-based in-memory computing technology changes the physics of how AI computations are performed, launching a new era in computing. Working hand in hand with Andes, we are confident that our joint AI accelerator chip will set a new standard for AI processing in terms of speed and energy efficiency."
Tetramem anticipates unveiling the AI accelerator chip and making engineering samples and development kits for the new 22nm “TetraMem MX Series” chip available to the public by the second half of 2024. The partnership between Andes and TetraMem signifies a major leap forward in the field of AI hardware, promising to unlock unprecedented possibilities for AI innovations.
TetraMem has announced a private beta of the world’s first memristor-based neural processing unit named MX 100 using analog in-memory compute architecture having 10 neural processing units with 64k 8-bit weights per NPU and a RISC-V processor boasting world-class MAC efficiency for small convolutional neural networks. For further information visit: https://www.tetramem.com/products
News Source: Andes Technology