Elma JetKit-3010, Руководство пользователя

Страница: 10 / 44

USER MANUAL JetKit-3010 ELMA.COM
www.elma.com EMBEDDED BOARDS
2 Hardware
2.1 Overview
The Elma JetKit-3010 is based on the NVIDIA ® Jetson AGX Xavier™ Industrial System-on-Module.
It provides the ability to integrate the NVIDIA ® Jetson AGX Xavier™ seamlessly into existing
CompactPCI Serial infrastructure. With its industry leading performance, power efficiency,
integrated deep learning capabilities and rich I/O, the NVIDIA
® Jetson AGX Xavier™ enables
emerging technologies with compute-intensive requirements. Designed for applications converging
on artificial intelligence (AI), computer vision (CV) and advanced graphics, the AGX Xavier™ is
ideal for (but not limited to): Intelligent Video Analytics (IVA), Robotics, Virtual Reality (VR),
Augmented Reality (AR), Portable Medical Devices and autonomous driving.
2.2 NVIDIA ® Xavier™ Software
Intentionally, the design of the JetKit-3010 was oriented on the design of the NVIDIA ® AGX
Xavier™ Developer Kit. This allows Software reutilization between platforms. Main differences
(influencing the Device Tree) are leaving out the audio codec and usage of a different USB type C
power delivery controller. The modified device tree file can be provided on request.
For more information on the NVIDIA
® Xavier™ module, please consult the NVIDIA ® online
documentation:
Jetson™ Download Center:
https://developer.nvidia.com/embedded/downloads
NVIDIA ® Jetson™ Linux Developer Guide:
https://docs.nvidia.com/jetson/l4t/
Getting Started:
https://developer.nvidia.com/embedded/learn/getting-started-jetson
2.3 Technical Specification
The Elma JetKit-3010 is based on the NVIDIA ® AGX Xavier™ Industrial module and inherits the
module’s features. The module adds further functionality and extendibility as well as it adapts the
interface to the CompactPCI Serial infrastructure. The technical specification of the JetKit-3010 can
be found in Table 1.
Table 1 - Technical Specification
Features Specifications
SoC
GPU Volta GPU
Tensor Cores
End-to-end lossless compression
Tile Caching
OpenGL® 4.6
OpenGL ES 3.2
Vulkan™ 1.0
CUDA® 10
CPU Carmel CPU Complex
ARMv8.2 (64-bit) heterogeneous multi-processing (HMP) CPU
Architecture
Dual-core CPU clusters connected by a high-performance system
coherency interconnect fabric
L3 Cache: 4MB (shared across all clusters)
NVIDIA
® Carmel (Dual-Core) Processor: L1 Cache: 128KB L1
instruction cache (I-cache) per core; 64KB L1 data cache (D-cache)
per core
L2 Unified Cache: 2MB per cluster