NVIDIA Jetson Thor Developer Kit Blackwell 2070T - Top for Humanoid Robotics and Edge AI

• ✔ Unrivaled AI performance with 2070 TFLOPS (FP4—Sparse)
• ✔ NVIDIA Blackwell architecture for cutting-edge graphics and computing power
• ✔ 14-core Arm® Neoverse®-V3AE CPU for robust parallel processing
• ✔ Stunning 128GB of 256-bit LPDDR5X memory with 273GB/s bandwidth
• ✔ Integrated 1TB NVMe M.2 slot for lightning-fast storage
• ✔ Support for up to 20 cameras for comprehensive vision and interaction
• ✔ Multi-Instance GPU (MIG) for efficient resource utilization

Introducing the NVIDIA Jetson Thor Developer Kit with Blackwell Architecture, the ultimate platform designed to push the boundaries of humanoid robotics, autonomous systems, and comprehensive Edge AI. Combining unprecedented AI performance with extensive connectivity, this development kit is ready to propel your most ambitious projects into a new era.

🤖 What can this robotic brain do?

• It delivers peak AI performance of 2070 TFLOPS (FP4—Sparse) for the most demanding AI models.
• It uses the advanced NVIDIA Blackwell architecture and a 14-core Arm® Neoverse®-V3AE CPU.
• Integrated 128 GB LPDDR5X memory and 1 TB NVMe storage provide ample capacity and speed.
• Supports Multi-Instance GPU (MIG) for parallel processing of multiple AI tasks.
• It offers a wide range of I/O interfaces, including support for up to 20 cameras for vision systems.

🔍 Explore every detail

The NVIDIA Jetson Thor Developer Kit is designed for developers and engineers who demand uncompromising performance and flexibility. Its 2560-core Blackwell GPU and 96 5th-generation Tensor cores are optimized for accelerating machine learning and deep neural networks. Combined with a powerful multi-core CPU and massive memory, it is capable of processing huge amounts of data in real time, which is crucial for applications such as autonomous vehicles, smart factories, and sophisticated humanoid robots.

With extensive connectivity options, including support for high-speed cameras and PCIe Gen5, Jetson Thor provides the necessary infrastructure for integration with the latest sensors and peripherals. Its compact size and efficient thermal management with an integrated heat transfer plate (TTP) make it an ideal solution for edge deployments where space and reliability are critical.

🚧 Are you bothered by limited performance and slow AI development at the edge of the network?

Current AI and robotics development platforms often face limits in processing power, memory, and I/O throughput, slowing innovation and making it difficult to implement complex autonomous systems. Optimizing and deploying advanced AI models on edge devices requires robust hardware that can process data with low latency and high efficiency.

🚀 Get unmatched power for your AI and robotics innovations!

The NVIDIA Jetson Thor Developer Kit addresses these challenges by delivering cutting-edge hardware that accelerates your development cycles and enables you to tackle projects previously considered too difficult. With Jetson Thor, you get a tool that lets you unleash the full potential of AI and push the boundaries of what’s possible at the edge of the network.

💡 More than just features, real benefits

• Features: 2070 TFLOPS of AI performance -> Benefit: Perform complex AI inference and training directly on the device with unprecedented speed, reducing latency and cloud dependency.
• Features: NVIDIA Blackwell Architecture & 14-core Arm Neoverse CPU -> Benefit: You get a platform with the latest technologies that is ready for future challenges in AI and robotics, with optimization for diverse tasks.
• Features: 128GB LPDDR5X memory -> Benefit: Conveniently manage large datasets and run memory-intensive models without compromising performance.
• Feature: Multi-Instance GPU (MIG) -> Benefit: Efficiently distribute GPU resources among multiple tasks or users, maximizing the throughput and flexibility of your system.
• Features: Extensive camera support (up to 20) -> Benefit: Integrate sophisticated visual systems for advanced perception of surroundings, key for autonomous navigation and interaction.

🎯 Who is it the ideal choice for?

• For humanoid robotics developers and researchers who need the highest AI performance and extensive connectivity for complex movement, perception, and interaction.
• For engineers developing autonomous vehicles and drones who require fast and reliable real-time processing of sensor data at the edge of the network.
• For industrial innovators and startups focused on Edge AI solutions, machine vision, and predictive maintenance, where low latency and high reliability are key.

🤔 Do you have concerns? We'll answer them

• Concern: Is integrating such a complex system too challenging? -> Answer: NVIDIA provides robust development tools, such as the JetPack SDK, and extensive community support that simplify development and deployment.
• Concern: Will the platform be powerful enough for future AI models? -> Answer: With Blackwell architecture and 2070 TFLOPS, Jetson Thor is ready for the AI ​​future. Additionally, MIG enables flexible scaling and workload optimization.
• Concern: How is cooling handled at such high power? -> Answer: The integrated thermal transfer plate (TTP) with heatpipe effectively dissipates heat, and there is also the option to add an optional fan or cooler.

📊 Technical specifications

Parameter	Value
AI Performance (FP4—Sparse)	2070 TFLOPS
GPU Architecture	NVIDIA Blackwell
GPU Cores	2560-core with 96 fifth-generation Tensor Cores
Multi-Instance GPU (MIG)	With 10 TPC
CPU	14-core Arm® Neoverse®-V3AE 64-bit
CPU L2 Cache	1 MB per core
CPU L3 Cache	16 MB shared system
Memory	128GB 256-bit LPDDR5X
Memory throughput	273 GB/s
Storage	1TB NVMe M.2 Key M Slot
NVMe support	Via PCIe
SSD support	Via USB3.2
HSB camera	Up to 20 cameras via HSB (QSFP slot)
MIPI CSI-2 camera	Up to 6 cameras via 16x MIPI CSI-2
Camera Virtual Channels	Up to 32 cameras using C-PHY 2.1 virtual channels (10.25 Gbps)
M.2 Key M slot	x4 PCIe Gen5
M.2 Key E slot	x1 PCIe Gen5 (WLAN/BT, USB2.0, UART, I2C, I2S)
PCIe Gen5	Up to x8 lines (Root port C1(x1), C3(x2) | Root Point/Endpoint C2(x1), C4(x8), C5(x4))
xHCI host controller	Integrated PHY
Ethernet	1x 5GBe RJ45 connector, 1x QSFP28 (4x 25 GbE)
Display	VESA DisplayPort 1.4a—HBR2, MST
CAN Header	2x 13-pin
Automation Header	2x 6-pin
JTAG connector	2x 5-pin
Fan connector	1x (12V, PWM, Tach)
Audio panel header	2x 5-pin
RTC battery connector	2-pin
Low-speed I/O	5x I2S/2x AHUB, 2x DMIS, 4x UART, 4x CAN, 3x SPI, 13x I2C, 6x PWM outputs
Dimensions	243.19mm x 112.40mm x 56.88mm
Thermal management	Integrated Thermal Transfer Plate (TTP) with heatpipe
Connector	699-pin B2B

❓ Frequently Asked Questions (FAQ)

What is the NVIDIA Jetson Thor Developer Kit?

It is an advanced AI and robotics development kit that uses the NVIDIA Blackwell architecture and offers cutting-edge computing power for Edge AI applications, humanoid robotics, and autonomous systems.

What AI performance does Jetson Thor provide?

The kit has AI performance of up to 2070 TFLOPS (FP4—Sparse), making it one of the most powerful on-device AI acceleration solutions.

Which CPU and GPU architecture is used?

The kit is equipped with a GPU with NVIDIA Blackwell architecture (2560 cores, 96 fifth-generation Tensor cores) and a 14-core Arm® Neoverse®-V3AE CPU.

How much RAM and storage is available?

NVIDIA Jetson Thor has 128GB of 256-bit LPDDR5X memory with 273GB/s throughput and an integrated 1TB NVMe M.2 slot for fast storage.

What applications is this kit best suited for?

It is ideal for developing humanoid robotics, autonomous systems, industrial AI solutions, advanced machine vision, and other applications that require high AI performance and extensive connectivity at the edge of the network.

What are the camera connection options?

The kit supports up to 20 cameras via the HSB (Host Side Bus) slot, up to 6 cameras via 16x MIPI CSI-2 lines and up to 32 cameras using C-PHY 2.1 virtual channels.

What is Multi-Instance GPU (MIG) and what is it used for?

MIG technology allows you to split the GPU into smaller, isolated instances, which maximizes the use of GPU resources and allows multiple AI tasks to be run in parallel or serve multiple users.

How is cooling handled?

The kit features an integrated Thermal Transfer Plate (TTP) with heatpipe for efficient heat dissipation. There is also the option to connect an optional fan or heatsink for further increased cooling performance.

What network interface does the kit offer?

There is 1x 5GBe RJ45 connector and 1x QSFP28 interface supporting 4x 25 GbE, ensuring high-speed network connectivity.

What are the physical dimensions of the development kit?

The kit dimensions are 243.19 mm x 112.40 mm x 56.88 mm, allowing its integration into various space-constrained applications.