The new Synopsys ARC® NPX Neural Processing Unit (NPU) IP family delivers the industry’s highest performance and support for the latest, most complex neural network models. The Synopsys ARC NPX6 NPU IP features up to 96K MACs with enhanced utilization, new sparsity features and new interconnect for scalability that address the demands of real-time compute with ultra-low power consumption for AI applications.
A single instance of the ARC NPX6 NPU IP delivers up to 250 TOPS at 1.3 GHz on 5nm processes in worst-case conditions, or up to 440 TOPS by using new sparsity features that can increase the performance and decrease energy demands of executing a neural network. Both the NPU IP and the new MetaWare MX Development Toolkit integrate connectivity features that enable implementation of multiple NPU instances to achieve up to 3,500 TOPS performance on a single SoC.
Neural Processor IP for Functional Safety
The Synopsys ARC NPX6FS NPU IP meets stringent random hardware fault detection and systematic functional safety development flow requirements to achieve up to ISO 26262 ASIL D compliance. The processors, with comprehensive safety documentation included, feature dedicated safety mechanisms for ISO 26262 compliance and address the mixed-criticality and virtualization requirements of next-generation zonal architectures.
High-Productivity Software Tools
The ARC MetaWare MX Development Toolkit includes compilers and debugger, neural network software development kit (SDK), virtual platforms SDK, runtimes and libraries, and advanced simulation models. MetaWare MX offers a single toolchain to accelerate application development and automatically partitions algorithms across the MAC resources for highly efficient processing. For safety-critical automotive applications, the MetaWare MX Development Toolkit for Safety includes a safety manual and a safety guide to help developers meet the ISO 26262 requirements and prepare for ISO 26262 compliance testing.
ARC NPX Neural Processing Unit (NPU) IP supports the latest, most complex neural network models and addresses demands for real-time compute with ultra-low power consumption for AI applications
Overview
Key Features
- ARC processor cores are optimized to deliver the best performance/power/area (PPA) efficiency in the industry for embedded SoCs. Designed from the start for power-sensitive embedded applications, ARC processors implement a Harvard architecture for higher performance through simultaneous instruction and data memory access, and a high-speed scalar pipeline for maximum power efficiency. The 32-bit RISC engine offers a mixed 16-bit/32-bit instruction set for greater code density in embedded systems.
- ARC's high degree of configurability and instruction set architecture (ISA) extensibility contribute to its best-in-class PPA efficiency. Designers have the ability to add or omit hardware features to optimize the core's PPA for their target application - no wasted gates. ARC users also have the ability to add their own custom instructions and hardware accelerators to the core, as well as tightly couple memory and peripherals, enabling dramatic improvements in performance and power-efficiency at both the processor and system levels.
- Complete and proven commercial and open source tool chains, optimized for ARC processors, give SoC designers the development environment they need to efficiently develop ARC-based systems that meet all of their PPA targets.
Benefits
- ARC processors are highly configurable, allowing designers to optimize the performance, power and area of each processor instance on their SoC by implementing only the hardware needed.
- The ARChitect wizard enables drag-and-drop configuration of the core, including options for Instruction, program counter and loop counter widths
- Register file sizeTimers, reset and interrupts Byte ordering Memory type, size, partitioning, base address Power management, clock gating Ports and bus protocol Multipliers, dividers and other hardware features Licensable components such as a Memory Protection Unit (MPU), Floating Point Unit (FPU) and Real-Time Trace (RTT).
- Adding/removing instructions
Technical Specifications
Maturity
Available on request
Availability
Available
Related IPs
- Enhanced Neural Processing Unit for safety providing 32,768 MACs/cycle of performance for AI applications
- Enhanced Neural Processing Unit for functional safety providing 49,152 MACs/cycle of performance for AI applications
- 2D (vector graphics) & 3D GPU IP A GPU IP combining 3D and 2D rendering features with high performance, low power consumption, and minimum CPU load
- Small-size ISP (Image Signal Processing) IP ideal for AI camera systems.
- General Purpose Neural Processing Unit (NPU)
- RISC-V ARC-V RMX-100 Ultra-low Power 32-bit Processor IP for Embedded Applications