The RTX 4000 series is the big announcement from Nvidia this Tuesday (20), but not the only one. Among the company’s other novelties is a small but invoked device: the new Jetson Orin Nano. Despite the reduced size, the device features a 512 or 1,024-core GPU and a hexa-core CPU.
We can even say that the Jetson Orin Nano is a “owned” version of the Raspberry Pi cards. But both lines follow different proposals. While the latter is focused on cost-effectiveness, the Nvidia line prioritizes performance in robotics or artificial intelligence applications, for example.
Just to give you an idea, the new line is capable of up to 40 trillion operations per second in artificial intelligence tasks.
For that, Nvidia equipped the card with an Ampere architecture GPU. This is the architecture that appears in GeForce RTX 3000 graphics chips. GeForce RTX 4090 and 4080 cards have Ada Lovelace architecturejust for the record.
Jetson Orin Nano series has two versions
Nvidia follows the nomenclature that ranks every trillion operations per second in artificial intelligence as TOP. As is already clear to you, the Jetson Orin Nano comes in two versions. One works with 20 TOPs; the other, with 40 TOPs.
In the most basic version, we find a GPU with 512 Ampere cores and a maximum frequency of 625 MHz, in addition to 16 Tensor cores for artificial intelligence. The same model features a CPU with six Arm Cortex-A78AE cores and clocked at 1.5 GHz.
This processor also equips the 40 TOPs Jetson Orin Nano. However, this version is more powerful because it features a GPU with 1,024 Ampere cores and 32 Tensor cores.
One version also differs from the other in RAM. The Jetson Orin Nano 20 TOPs has 4GB of LPDDR5 memory; the 40 TOPs version, 8 GB of LPDDR5.
Storage? There is not. But the cards support NVMe to work together with external storage.
The Jetson Orin board line
The Jetson Orin Nano models are the simplest versions of a line of high-performance cards from Nvidia. The signs Jetson Orin NX appear as intermediate options, while units Jetson AGX Orin appear as the most powerful.
For you to have an idea, the Jetson AGX Orin with 64 GB of RAM has 2,048 Ampere cores and CPU with 12 Arm Cortex-A78AE cores. This lineup also has internal storage — 64GB of eMMC 5.1.
In all cases, the objective is one: to shape projects in robotics, artificial intelligence, research and development, automation and any other activity that benefits from embedded computing.
According to Nvidia, more than 1,000 customers and 150 partners already have projects based on Jetson AGX Orin cards since they were released six months ago.
The Jetson Orin Nano line completes the portfolio by offering simpler but still advanced configuration options.
Jetson Orin Nano: when and how much?
The first units of Jetson Orin Nano boards are expected to start shipping in January 2023. In the United States, prices will start at $199.
For a change, we still don’t have availability information in Brazil.
Nvidia Jetson Orin Nano — data sheet
|Jetson Orin Nano (4GB)||Jetson Orin Nano (8GB)|
|GPU||512 Ampere cores, 16 Tensor cores, clocked at 625 MHz||1024 Ampere cores, 32 Tensor cores, clocked at 625 MHz|
|CPU||Six 1.5GHz Cortex-A78AE cores with L2 (1.5MB) and L3 (4MB)||Six 1.5GHz Cortex-A78AE cores with L2 (1.5MB) and L3 (4MB)|
|Memory||4GB of LPDDR5 with 64-bit interface||8 GB of LPDDR5 with 128-bit interface|
|compatible connections||Three USB 3.2 Gen 2, three USB 2.0, one Gigabit Ethernet, PCIe 3, GPIO connectors, UART and the like||Three USB 3.2 Gen 2, three USB 2.0, one Gigabit Ethernet, PCIe 3, GPIO connectors, UART and the like|
|video support||A 4K display at 30 fps on DisplayPort 1.2 or HDMI 1.4||A 4K display at 30 fps on DisplayPort 1.2 or HDMI 1.4|
|standard consumption||5W to 10W||7W to 15W|
|Performance||20 TOPs||40 TOPs|
|dimensions||69.6 x 45 mm||69.6 x 45 mm|