Loading...
 Start Page
 
 
 

NEOX™ GRAPHICS

 

The ultimate ultra-low-power RISC-V based GPU Processor

 
DESCRIPTION

NEOX™ is a parallel multicore and multithreaded GPU architecture based on the RISC-V RV64IMFC instruction set with adaptive NoC. The number of cores varies from 4 to 64 organized in 1-16 cluster elements, each configured for cache sizes and thread counts. Depending on cluster/core configuration, NEOX™ compute power ranges from 12.8 to 409.6 GFLOPS at 800MHz with support for FP16, and FP32, and SIMD instructions.

Additional information

Download NEOX™ Product Brief
MARKETS & APPLICATIONS

The various configuration possibilities of the NEOX™ IP with custom user instructions, make it flexible to configure key applications, such as computer graphics, machine learning, vision/video processing, and general-purpose compute. It serves a wide range between power, performance, and functional integration with different levels of SoC platforms sporting Microcontroller, Crossover Processors, and Application Processors. NEOX™ addresses multiple verticals in AI, IoT/Edge, and performance media processing in consumer and industrial markets.

  • GPGPU Compute
  • Artificial Intelligence
  • Vision Processing
  • Video Processing
  • Computer Graphics
NEOXTM
PERFORMANCE

Scalable to match multiple applications & performance levels

NEOX™ is a highly configurable IP supporting 1 to 4 cores per cluster. The cache sizes of each cluster and the thread count of the cores are parameterized and can be tailored to different applications and use cases. Multithreading hides long latency delays from the external memory controller maintaining high computation throughput for the entire array. NEOX™ represents a new era of smart GPU architectures with programmable compute shaders, running on a real-time operating system (RTOS) and supported by light-weight graphics and machine learning programming frameworks. The heavily multi-threaded system and its configurable programming libraries use the same hardware blocks and offer customized extensions by combining graphics, machine learning, vision/video processing and general-purpose compute workloads.
CLUSTER
CORES
MHz
GFLOPS FP32
GFLOPS FP16
TOTAL THREADS
 
 
NEOX™ 4
1
4
800
12.8
25.6
256

ARCHITECTURE

The NEOXTM architecture includes AI-specific ISA extensions, SIMD Vector in variable length datatypes including 8-bit and optionally Graphics ISA Extensions/Coprocessors: Unified Shader Architecture, Tile Based Rendering, Color/Vertex, Vector Support and contains dedicated hardware modules, such as rasterizer, texture unit, tile management unit and texture caches. By supporting a dedicated interface allows SoC architects to augment the instruction set with user-defined instructions to enable product differentiations and the ability to create custom unique designs.

DELIVERABLES, SOFTWARE & INTEGRATION*

NEOX™ SDK, System Verilog RTL, Integration Tests, LLVM C/C++ compiler, GCC C/C++ compiler. Custom instructions for Computer Graphics, Compute and AI, and user defined extensions. Evaluation on Xilinx SoC FPGA platform and SW Cycle Accurate Simulator. Supported OS: Linux, RTOS, Wear OS.

* Listed items represent a superset and are subject to change without further notice.

IN A NUTSHELL

ARCHITECTURE

  • RISC V64GC ISA
  • Multicore Array
  • Multithreaded
  • Adaptive NoC
  • Configurable 4-64 Cores

SOFTWARE

  • C/C++ LLVM Compiler
  • C/C++ GCC Compiler
  • Posix Threads
  • Open Graphics Frameworks

EVALUATION

  • Xilinx Zynq FPGA
  • Cycle Accurate Simulator

DELIVERABLES

  • System Verilog RTL
  • Configuration Tool
  • Verification Suite
  • Synthesis Scripts
  • Software Emulator
  • FPGA Prototype
  • SDK
Copyright © 2021 - 2024 Think Silicon | ULTRA-LOW POWER GPUs