Nvidia announced liquid-cooled GPU variants and Jetson AI edge devices at Computex 2022.

At the start of Computex 2022, Nvidia showcased a portfolio of upcoming products and industry blueprints designed to expand compute performance across multiple market segments. The announcements included designs for new "superchip" systems based on the Grace and Hopper architectures for massive modeling and learning tasks, a compact GPU targeted at data centers, and new "Jetson Orin" systems for AI-intensive workloads. Recognizing that high-performance computational hardware requires different configurations for specialized tasks, Nvidia introduced four new reference designs for data-center systems to help customers find the right solutions more easily.

The four reference designs vary in performance and hardware specifications and are built to be customizable, all based on iterations of the Grace and Hopper architectures and their associated giant-compute processors. Different Grace-chip-based blueprints target cloud-graphics and Nvidia Omniverse applications—the latter for 3D design simulation and collaboration—while a Grace Hopper-based design focuses on AI, learning, and inferencing tasks.

Nvidia also announced its first data-center graphics card, the Nvidia A100, a PCI Express GPU designed to fit inside a single PCI Express slot, unlike the two- and three-slot designs of mainstream and high-end consumer market cards. The "Ampere"-based A100 employs liquid cooling to provide thermally efficient operation in a smaller package than conventional air-cooled graphics cards, enabling higher compute densities in space-strapped server environments. A liquid-cooled H100 is expected to follow in 2023. Card partners including Asus, Asrock, and Gigabyte will source these GPUs, with the A100 targeted at customers who already have upgradeable servers or cannot justify purchasing new data-center systems. The A100 is expected to launch in the second half of the year.

Nvidia is expanding its Jetson product line with four new devices ranging in price from $399 to $1,599, designed to drive AI-intensive workloads while providing performance improvements over previous-generation products. The most affordable offering, the Jetson Orin NX 8GB, will feature six ARM Cortex-A78 CPUs and handle up to 70 trillion operations per second. The more powerful Jetson AGX Orin 64GB will provide significantly greater capability with a dozen ARM Cortex-A78 CPU cores and a peak bandwidth of 275 TOPS. These new high-performance components are expected to launch at varying points over the coming months.