- Acer Veriton GN100 uses Nvidia GB10 Superchip with a memory of 128 GB and a storage of 4 to
- The compact design of less than 1.5 kg always offers 1 pflops of FP4 AI performance
- Two linked systems are on the scale of AI models with up to 405 billion parameters
Acer announced the workstation Veriton GN100 AI Mini, a compact office built on the GB10 Grace Blackwell of Nvidia.
It is designed to locally execute large AI models, with the possibility of setting up workloads by connecting two systems together.
The Veriton GN100, revealed at IFA 2025 in Berlin, offers up to 1 pflops of FP4 AI performance thanks to a combination of 20 ARM processor nuclei, Cuda hearts and fifth generation tensor nuclei.
Up to 405 billion parameters
The device is delivered with 128 GB of unified LPDDR5X memory and 4 TB of NVME Auto-encrypative storage.
Despite its 150 mm x 150 mm 150 mm imprint and its weight less than 1.5 kg, the Veriton GN100 is intended to provide a server class calculation power in a mini PC form factor.
The system includes the AI software battery of Nvidia and the basic DGX operating system, with management of commonly used development tools such as Pytorch, Jupyter and Olllama.
Acer says that this will allow developers, researchers and students to prototyper, refine and test large languages directly on the system, reducing dependence on remote cloud infrastructure and helping to reduce operating costs.
With an NVIDIA Connectx-7 Smartnic installed, two Veriton GN100 units can be linked to the workload scale, allowing the treatment of AI models with up to 405 billion parameters.
This will allow you to work with training sets and even more important experiences without requiring immediate access to the data center equipment.
Connectivity includes four USB 3.2 type C, HDMI 2.1b, Ethernet, Wi-Fi 7 and Bluetooth 5.1.
Security features include a Kensington lock and local execution of AI models to help protect sensitive data.
Acer has set the price of the Veriton GN100 at $ 3,999 in North America and € 3,999 in EMEA, with availability varying depending on the region.
Acer is only the last PC manufacturer to launch an NVIDIA DGX Spark clone, according to the traces of Asus, HP and Dell and MSI.
