Acer's Nvidia Blackwell workstation for tasks with local AI models

Acer has released the compact Veriton GN100 workstation. This new product offers a unique opportunity to run large language models. The workstation is based on the Nvidia GB10 Blackwell Superchip.

Veriton GN100 features

The Veriton GN100 is designed for researchers and developers seeking to minimize their reliance on the cloud.

In terms of performance, users can expect 1 PFLOPS in FP4 tasks. The platform consists of a Tensor Cores subsystem and 20 Arm processor cores. Additionally, 128 GB of unified memory is available, and up to 4 TB of NVMe SSD storage can be installed.

The company also offers a unique opportunity to combine two workstations using a ConnectX-7 SmartNIC network adapter. This makes it possible to run large AI models without the need for a server rack.

Along with the Veriton GN100, Nvidia software tools will also be available in the package, with Ollama and PyTorch support “out of the box.” The price of the device will be around $3,999.