Nvidia Announces Jetson Xavier NX, World’s Smallest Supercomputer for AI

Navin Bondade 1 Comment Artificial Intelligence, Data Science, Deep Learning, Machine Learning

Nvidia has announced Jetson Xavier NX this week, the company claims that it’s the world’s smallest, most powerful AI supercomputer for robotic and embedded computing devices at the edge.

Nvidia said that Xavier NX has achieved the fastest results across five benchmarks in the MLPerf Inference Suite. The new Jetson Xavier NX is physically smaller about the size of a credit card, 70 x 45mm.

“We’ve taken Xavier and tuned it for extremely high performance but in a very targeted power budget of 10 to 15 watts, which is really important for a lot of the segments of the autonomous and robotics embedded world,” Nvidia VP Deepu Talla said in a briefing before its release.

Jetson Xavier NX delivers up to 21 TOPS (INT8) or 6 TFLOPS (FP16) of AI performance while consuming only 15 watts of power. When limited to 10 watts, it can still perform at 14 TOPS.

That compares to the .5 TFLOPS (FP16) of the similarly-sized Nano ($129) and almost matches the specs of the much larger Jetson AGX Xavier ($599). By comparison, Intel’s Movidius Myriad X AI accelerator and Arm’s Ethos-N77 NPUs, are limited to 4 TOPS but support lower power operation.

The Jetson Xavier NX is pin-compatible with the current Jetson Nano, which should mean that current embedded devices will quickly be able to benefit from an order of magnitude increase in performance in essentially the same power and size envelope.

The Xavier NX features Nvidia Volta with 384 CUDA cores, along with 48 tensor cores, and 2 Nvidia Deep Learning Accelerators, thanks to Nvidia’s engineering and design, the Jetson Xavier NX provides up to 15x higher performance than its Jetson TX2 in a smaller form factor with the same power draw.

The Santa Clara, Calif.-based company said the Jetson Xavier NX is made for performance-hungry devices that are “constrained by size, weight, power budgets or cost,” such as commercial robots, drones high-resolution factory sensors, portable medical devices, and industrial IoT systems.

“AI has become the enabling technology for modern robotics and embedded devices that will transform industries,” Deepu Talla, vice president and general manager of edge computing at Nvidia, said in a statement.

“Many of these devices, based on small form factors and lower power, were constrained from adding more AI features. Jetson Xavier NX lets our customers and partners dramatically increase AI capabilities without increasing the size or power consumption of the device.”

Like Nvidia’s other Jetson products, Jetson Xavier NX runs on the chipmaker’s CUDA-X AI software architecture that the company said can speed up development and lower costs. It’s also supported by the company’s JetPack software development kit, which provides a “complete AI software stack.”

The Jetson Xavier NX has twice the RAM of the Nano with 8GB and similarly supplies 16GB eMMC. It’s pin-compatible with the Nano, enabling Nano-based carrier boards and other hardware to upgrade.

The Xavier NX is equipped with 6x ARMv8.2 Carmel CPU cores with 6MB L2 and 4MB L3 cache compared to 8x ARMv8.2 Carmel cores with 8MB L2 and 4MB L3 cache on the Xavier AGX.

AnandTech has benchmarked the Xavier AGX’s CPU and found it to be roughly equivalent to a Cortex-A75, compared to Cortex-A57 on the quad-core Nano. No clock rates were mentioned, and it’s possible these are throttled down Carmel cores optimized to run at lower power.

The Xavier NX is limited to dual 4Kp30 encoding instead of the AGX’s dual 4Kp60.

The module supports up to 6x MIPI-CSI cameras (or 36 via virtual channels) at 1080p60 resolution and 12 lanes (3×4 or 6×2) of MIPI-CSI-2 at up to 30Gbps. By comparison, the AGX supports 16 lanes of CSI-2 and 8x lanes of SLVS-EC for up to 16 simultaneous cameras.

The module ships with 8GB 128-bit LPDDR4x at 51.2GB/second compared to 16GB 256-bit LPDDR4 on the AGX. On-module storage is similarly halved to 16GB eMMC 5.1 down from

The Xavier NX supports Nvidia’s popular JetPack SDK and runs CUDA-X AI software architecture, developers who want to get a head start on developing for Xavier NX can work with the Jetson AGX Xavier Development Kit after applying a patch to emulate the Xavier NX.

The Nvidia Jetson Xavier NX module will be available in Mar. 2020, priced at $399. More information may be found in Nvidia’s announcement and product page.

“In a world where AI chips are announced on what seems like a daily basis, I believe NVIDIA raised the bar with its Jetson Xavier NX — showing that exceptional performance at small size and low power, together with a consistent and powerful software architecture, is what matters in embedded edge computing,” said Patrick Moorhead, president and principal analyst of Moor Insights & Strategy.

Here’s a quick comparison of the Jetson lineup, ranked in order of ascending processing power:

Jetson Nano: 69.6 x 45mm; 4x -A57 @ 1.43GHz CPU; 128-core Maxwell GPU; 4GB LPDDR4; 16GB eMMC
Jetson TX2: 87 x 55mm; 2x Denver and 4x -A57 CPU; 256-core Pascal GPU; 8GB LPDDR4; 32GB eMMC; opt. WiFi/BT
Jetson Xavier NX :69.6 x 45mm; 6x ARMv8.2 Carmel CPU; 384-core Volta GPU; 48 tensor cores; 8GB LPDDR4; 16GB eMMC
Jetson Xavier AGX — 105 x 87mm; 8x ARMv8.2 Carmel (-A75 like); 512-core Volta GPU; 64 tensor cores; 16GB LPDDR4; 32GB eMMC