对比
EyeQ™6H vs. NVIDIA Jetson AGX Orin
EyeQ™6H vs.
NVIDIA Jetson AGX Orin
EyeQ™6H 在边缘侧实现高性能 AI 运算,支持部署深度神经网络(DNN)、 Transformer 模型及多种主流机器学习框架,能够实时完成图像分类、目标检测与图像分割等任务。 作为我们迄今为止最先进的系统集成芯片(SoC), EyeQ™6H 可将原始传感器数据高效转化为精准且可执行的洞察,赋能驾驶自动化系统做出快速、可靠的决策。
图像分类
EyeQ™6H
NVIDIA Jetson AGX Orin 64GB*
TOPS
34 TOPS
275 TOPS
单帧延迟(msec)
模式
Power
Power (MaxQ)
Unconstrained (MAXN)
ResNet-50
0.5 msec
1.64 msec
0.64 msec
Power 模式 指芯片在原始功耗状态下测试,无任 何修改。
Unconstrained 模式
指在无外部限制功耗或散热 条件下测试。
* Nvidia 与 MLPerf 公布的公开数据。
Vision Transformer
模型
分辨率
参数
MACs
EyeQ™6H
(Int8)*
NVIDIA Jetson AGX Orin 32GB
(FP16)**
模式
Power
Unconstrained
EfficientViT-B1
224x224
9.1M
0.52G
0.564 msec
1.48 msec
EfficientViT-B2
224x224
24M
1.6G
0.932 msec
2.63 msec
*EyeQ™6H 支持 Int8 精度计算,并通过先进的 QAT(量化感知训练)和 PTQ(训练后量化)技术,在性能、能效与精度之间实现理想平衡。
** Orin 数据基于 FP16。
图像分类
Power 模式 指芯片在原始功耗状态下测试,无任 何修改。
Unconstrained 模式
指在无外部限制功耗或散热 条件下测试。
* Nvidia 与 MLPerf 公布的公开数据。
Vision Transformer
*EyeQ™6H 支持 Int8 精度计算,并通过先进的 QAT(量化感知训练)和 PTQ(训练后量化)技术,在性能、能效与精度之间实现理想平衡。
** Orin 数据基于 FP16。
高效架构设计
凭借独特且高度高效的多元加速器架构, EyeQ™ 在低功耗的前提下,实现了业界领先的计算机视觉性能。
异构计算
为每项任务匹配最合适的计算内核
从通用 CPU 到高计算密度的加速器,包括深度学习神经网络, 能灵活调度不同类型的处理单元,最大化计算效率与性能表现。
CPU
Central Processing Unit
MPC
Multi-threaded Processor Cluster
More versatile than a GPU, and with higher efficiency than any CPU
VMP
Vector Microcode Processor
A wide vector (VLIW and SIMD) machine with exceptional performance for short integral types common in computer vision and deep learning algorithms.
PMA
Programmable Macro Array
A CGRA dataflow machine. Its unique architecture delivers outstanding performance for dense computer vision and deep learning algorithms that are unachievable in classic DSP architecture.
XNN
Deep Learning Accelerator
Dedicated high-performance AI engine. The main source of horse power for convolutional neural networks.
GENERAL-PURPOSE COMPUTE
CPU
Central Processing Unit
MPC
Multi-threaded Processor Cluster
More versatile than a GPU, and with higher efficiency than any CPU
VMP
Vector Microcode Processor
A wide vector (VLIW and SIMD) machine with exceptional performance for short integral types common in computer vision and deep learning algorithms.
PMA
Programmable Macro Array
A CGRA dataflow machine. Its unique architecture delivers outstanding performance for dense computer vision and deep learning algorithms that are unachievable in classic DSP architecture.
XNN
Deep Learning Accelerator
Dedicated high-performance AI engine. The main source of horse power for convolutional neural networks.
DL PERFORMANCE / COMPUTE DENSITY