Is this your business? Claim it to manage your IP and profile
The Avispado is a sleek and efficient 64-bit RISC-V in-order processing core tailored for applications where energy efficiency is key. It supports a 2-wide in-order issue, emphasizing minimal area and power consumption, which makes it ideal for energy-conscious system-on-chip designs. The core is equipped with direct support for unaligned memory accesses and is multiprocessor-ready, providing a versatile solution for modern AI needs. With its small footprint, Avispado is perfect for machine learning systems requiring little energy per operation. This core is fully compatible with RISC-V Vector Specification 1.0, interfacing seamlessly with Semidynamics' vector units to support vector instructions that enhance computational efficiency. The integration with Gazzillion Misses™ technology allows support for extensive memory latency workloads, ideal for key applications in data center machine learning and recommendation systems. The Avispado also features a robust set of RISC-V instruction set extensions for added capability and operates smoothly within Linux environments due to comprehensive memory management unit support. Multiprocessor-ready design ensures flexibility in embedding many Avispado cores into high-bandwidth systems, facilitating powerful and efficient processing architectures.
The Atrevido is a 64-bit RISC-V core designed for out-of-order processing, providing exceptional performance for applications needing high bandwidth and low latency. It features a 2/3/4-wide configurable out-of-order issue and completion mechanism, ensuring a seamless handling of complex, memory-intensive operations. The core is multiprocessor ready, equipped with direct hardware support for unaligned memory accesses, and supports various RISC-V extensions for enhanced functionality. This IP is particularly adept at handling machine learning workloads, key-value stores, and recommendation systems, thanks to its integration with Semidynamics' Gazzillion Misses™ technology. This technology enables the Atrevido core to sustain full memory bandwidth even with smaller processing cores, minimizing the need for a large silicon footprint. With support for the RISC-V Vector Specification 1.0, Atrevido is vector-ready, allowing for dense encoding of computational instructions and efficient handling of sparse tensor weights. Additional features of the Atrevido core include its Linux readiness, with full MMU support, and its compatibility with cache-coherent multiprocessing environments. This makes it beneficial for constructing systems on chips that require numerous cores, delivering scalability and performance tailored to extensive processing needs.
The Vector Unit from Semidynamics is an extensively configurable RISC-V processing component designed to handle multiple types of arithmetic and logic operations across various data types. Fully customizable for performance to specific application needs, it supports an adjustable data and vector path length, allowing for seamless integration into diverse processing environments. This unit is compatible with the RISC-V Vector 1.0 Specification and supports integer and floating-point operations across data widths from 8 to 64 bits and beyond. With a scalable design, the Vector Unit enables configuration of the vector core count and corresponding data path widths, catering to a variety of power-performance-area trade-off scenarios. The machine can integrate with up to 32 vector cores, enabling wide adoption across different segments of computing, from AI to standard processing tasks. Designed to interface efficiently with other Semidynamics technologies, such as the Tensor Unit, and equipped with the Gazzillion Misses™ for data flow optimization, this unit is fortified for advanced AI requirements and general-purpose processing tasks. Moreover, its seamless deployment within Linux environments underscores its flexibility and ease of integration into existing systems.
Semidynamics' Tensor Unit represents an advanced solution for executing intensive AI workloads, particularly when involving matrix multiplications, common in large-scale language and neural network models. This fully-coherent unit integrates with Semidynamics' RISC-V cores, providing fluid interaction and data transfer with their vector units. It is specifically optimized for matrix-heavy computations, delivering low-power yet high-speed computation capabilities directly from the core system architecture. The unit operates seamlessly within an RISC-V vector-enabled Linux environment and leverages existing vector registers to minimize additional hardware and programming complexity. This integration ensures enhanced performance by eliminating traditional bottlenecks and the need for cumbersome Direct Memory Access (DMA) programming. Instead, it relies on efficient designs and memory handling technologies, such as Semidynamics' Gazzillion Misses™, to optimize data flows internally within the system. With the Tensor Unit, developers gain the capacity to execute fast AI processes efficiently, boosting performance 128 times faster compared to standalone scalar core solutions. This makes it an indispensable component for running complex AI tasks such as neural network inference and training operations, underscoring its role in powering next-generation AI applications.
Gazzillion Misses™ is a groundbreaking technology designed to eliminate memory latency constraints in high-speed data operations. Integrated into Semidynamics' RISC-V cores, this innovation dramatically enhances the capability to handle vast numbers of memory requests, supporting up to 128 outstanding requests simultaneously. This vast increase in throughput enables processors to continue valuable computations while previously issued missing requests are processed, which is critical for high-bandwidth applications like big data and AI. Traditional processors often halt new request issuance under heavy memory demand, leading to significant idle times. However, Gazzillion Misses™ transforms this scenario by employing an implicit parallel overhaul of bandwidth management, ensuring that memory systems operate at maximum speed with almost no idle time. Consequently, this technology facilitates optimal performance, translating into increased application speed without additional hardware complexity. Developers targeting applications such as machine learning, HPC, and real-time processing can benefit tremendously from Gazzillion's ability to handle large datasets seamlessly, overcoming the conventional 'memory wall' problem. Its integration assures performance boosts without unnecessary software complication, thereby streamlining development and deployment procedures for next-generation semiconductors.
Join the world's most advanced semiconductor IP marketplace!
It's free, and you'll get all the tools you need to discover IP, meet vendors and manage your IP workflow!
To evaluate IP you need to be logged into a buyer profile. Select a profile below, or create a new buyer profile for your company.