READ THE 275 TOPS HEADLINE CAREFULLY. NVIDIA's famous 275 TOPS for this module is a SPARSE INT8 figure and it is the SUM OF TWO SEPARATE ENGINES: 170 sparse TOPS from the GPU tensor cores plus 105 sparse TOPS from the two NVDLA v2.0 deep learning accelerators. This is a GPU catalog, so the INT8 row below carries the GPU-only figures, 85 dense with NVIDIA's own 170 sparse alongside, and the 275 is stated here in prose rather than stored as a GPU throughput it is not. POWER IS CONFIGURABLE from 15W to 60W; tdp_watts and max_power_watts both carry the ceiling. NVIDIA prints dense and sparse side by side for both INT8 and FP16 on its Jetson Orin product page, so neither dense figure is derived. The FP32 figure is the CUDA-core rate from the technical brief and is not a tensor rate.

NVIDIA Jetson AGX Orin 64GB

Ampere Module 2021 Spec confidence: Official
FP32
5.3
TFLOPS
VRAM
64 GB
LPDDR5
Bandwidth
205 GB/s
memory
TDP
60 W
88.3 TFLOPS/kW
Download datasheet

Overview

The Jetson AGX Orin 64GB is the largest of the Orin modules and the one that carried NVIDIA's edge line until Thor arrived. It pairs a 2048-core Ampere GPU with 64 third-generation tensor cores and two NVDLA v2.0 accelerators, and NVIDIA rates the whole module at 275 sparse INT8 TOPS, of which 170 comes from the GPU and 105 from the deep learning accelerators. On the GPU alone that is 85 dense INT8 TOPS and 43 dense FP16 TFLOPS, plus 5.3 TFLOPS of FP32 from the CUDA cores. The 64 GB of 256-bit LPDDR5 at 204.8 GB/s is the striking part for local model work, since it is eight times the frame buffer of the Orin Nano at twice the bandwidth, in a 100 by 87 mm module drawing between 15 W and 60 W. It remains the sensible choice for a robot that needs a lot of memory and does not need Blackwell.

Performance

Peak theoretical throughput by precision type

PrecisionDense2:4 Sparse
FP64
No verified data available Structured sparsity is a tensor-core feature; this vector precision has no sparse form
FP32
32-bit floating point
5.3TFLOPS Structured sparsity is a tensor-core feature; this vector precision has no sparse form
TF32
No verified data available The part supports 2:4 sparsity, but the vendor publishes no sparse figure for this precision
BF16
No verified data available The part supports 2:4 sparsity, but the vendor publishes no sparse figure for this precision
FP16
16-bit floating point
43TFLOPS 85TFLOPS
FP8
No verified data available The part supports 2:4 sparsity, but the vendor publishes no sparse figure for this precision
FP6
No verified data available The part supports 2:4 sparsity, but the vendor publishes no sparse figure for this precision
FP4
No verified data available The part supports 2:4 sparsity, but the vendor publishes no sparse figure for this precision
INT8
8-bit integer
85TOPS 170TOPS

The Jetson AGX Orin 64GB in the GPU landscape

Peak FP16 TFLOPS (dense) against TDP, single-GPU parts tracked by Flopper

06001,2001,8002,4003,0000 W250 W500 W750 W1000 W1250 W1500 WInstinct MI355XJetson AGX Orin 64GB

Higher and further left is better: more half-precision throughput for less power.

Specifications

Architecture

Ampere

Form Factor

Module

Launch Year

2021

Process Node

No verified data available

Memory

64 GB LPDDR5

Bandwidth

204.8 GB/s

TDP

60 W

Max Power

60 W

CUDA Cores

2,048

Spec Confidence

Official

Full Specifications

Compute Engine
CUDA Cores 2,048
Tensor Cores 64 (3rd Gen)
Streaming Multiprocessors 16
Boost Clock 1.30 GHz
Memory
VRAM 64 GB
Memory Type LPDDR5
Bandwidth 205 GB/s
Interface Width 256-bit
Power & Thermal
TDP 60 W
Max Board Power 60 W
Enterprise Features
Sparsity Yes
General
Form Factor Module
Architecture Ampere
Process Node No verified data available
Launch Year 2021

Datasheet & Resources

Data Provenance

Every figure traced to a source

Primary Source

Publisher
NVIDIA
Published
2022-07-01

Data Quality

Spec confidence
Official
Clock basis
Boost
Core precisions with figures
3 of 9
Normalization
All values in TFLOPS

Compare cloud providers offering on-demand GPU instances for AI training, inference, and HPC workloads.

Browse GPU Cloud Providers

Similar GPUs

NVIDIA RTX A6000 48GB

PCIe · 2020
FP32: 39 TFLOPS
Compare vs RTX A6000

NVIDIA A40 48GB

PCIe · 2021
FP32: 37 TFLOPS
Compare vs A40

NVIDIA A10G 24GB

PCIe · 2022
FP32: 35 TFLOPS
Compare vs A10G

Frequently Asked Questions

How many TFLOPS does the NVIDIA Jetson AGX Orin have?

The NVIDIA Jetson AGX Orin delivers 5.3 TFLOPS FP32 and 43 TFLOPS FP16 at peak. Flopper does not currently have a verified FP8 throughput figure for it.

What is the power consumption of the NVIDIA Jetson AGX Orin?

The NVIDIA Jetson AGX Orin has a TDP (Thermal Design Power) rating of 60 watts.

How much memory does the NVIDIA Jetson AGX Orin have?

The NVIDIA Jetson AGX Orin is equipped with 64 GB of VRAM with 204.8 GB/s of memory bandwidth.

What architecture is the NVIDIA Jetson AGX Orin based on?

The NVIDIA Jetson AGX Orin is based on the Ampere architecture, launched in 2021.

Stay Updated on GPU Releases

Get notified when new GPUs are added or specifications are updated.

Loading verification...

No spam, unsubscribe anytime.

Back to GPUs