Logo

SemiVoice

Recent #AI Acceleration news in the semiconductor industry

semiwiki
3 months ago

CEO Interview with Fabrizio Del Maffeo of Axelera AI

➀ Axelera AI, founded in 2021, aims to provide scalable edge AI hardware and software solutions.

➁ The company has raised USD 120 million and has a world-class team of over 190 employees.

➂ Axelera AI's Metis platform offers high performance with low power consumption, targeting applications like computer vision, automotive, and healthcare.

AIAI AccelerationAI PlatformAxelera AIComputational EfficiencyDeep LearningEdge ComputingEnergy ConsumptionHardwareHealthcareIn-Memory ComputingIoTMetisNeural NetworksRISC-VRetailSDKSEMICONDUCTORSecurityautomotivesoftware
semiwiki
3 months ago

Harnessing Modular Vector Processing for Scalable, Power-Efficient AI Acceleration

➀ A new chiplet-based RISC-V vector processor architecture aims to balance performance, efficiency, and flexibility for AI workloads.

➁ The architecture utilizes modular chiplets, including a vector processing chiplet and a matrix multiplication accelerator, to optimize various AI computation aspects.

➂ The RISC-V solution addresses memory bottlenecks and enhances energy efficiency with time-based execution and dynamic power gating.

AI AccelerationChipletModular ComputingRISC-VVector Processing
semiwiki
7 months ago

Semidynamics: A Single-Software-Stack, Configurable and Customizable RISC-V Solution

➀ Semidynamics introduces a unified, programmable solution with a single software stack for Core, Vector, and Tensor processing units; ➁ The company's 'All-in-One IP' solution addresses the challenges of data availability and processing for AI applications; ➂ Semidynamics leverages RISC-V architecture for adaptability and control, offering configurable and customizable IP solutions.
AIAI AccelerationAI HardwareHardwareIP SolutionRISC-VSemidynamics
semiwiki
10 months ago

A New Class of Accelerator Debuts

1. Quadric introduces a new accelerator architecture, Chimera, which supports a wide range of applications including signal processing and GenAI, offering up to 864 TOPs. 2. The Chimera approach differs from traditional accelerators by using a common pipeline for all instruction types, simplifying software development and handling complex AI models more efficiently. 3. Quadric's platforms, ranging from QC Nano to QC Ultra, are supported by a mature Chimera SDK, allowing for easy integration of various AI models without extensive tuning.
AI AccelerationQuadricSignal Processing