When an AI model outgrows a single workstation, replacing the entire system is not your only option. The Acer Veriton GN100 supports connecting two units through NVIDIA ConnectX-7 networking, giving developers a way to expand their local AI setup.
For teams exploring larger language models, this creates a practical upgrade path: start with one workstation, then add a second when your workload requires more capacity.
How Two GN100 Systems Work Together
A compatible QSFP cable connects the systems through their ConnectX-7 ports. After network and software configuration, a supported distributed application can divide an AI workload between them. Acer confirms this connectivity for the GN100, while NVIDIA provides a two-system setup playbook for the underlying DGX Spark platform.
Each machine remains a separate computer with its own processor, memory, storage, and power supply. The connection lets them exchange data; it does not automatically turn them into one conventional PC.
What Changes When You Add a Second System
| Configuration | One VGN100-UD13 | Two VGN100-UD13 Units |
|---|---|---|
| GB10 Superchips | 1 | 2 |
| Unified memory | 128GB | 256GB total across two separate nodes |
| SSD storage | 2TB | Two separate 2TB SSDs |
| Workload setup | Single-system software | Distributed software and network configuration |
The totals above are calculated from two identical units. Memory and storage do not automatically merge into a single shared pool. Your application must support using resources across both machines.
Can You Run Larger AI Models
Acer advertises support for models up to 405 billion parameters with two connected GN100 systems. Treat this as a supported configuration target, rather than a guarantee for every model. Quantization, context length, runtime overhead, and software compatibility determine whether a particular workload fits and performs acceptably.
Adding a second unit is especially worth considering when memory capacity limits your model choice. It does not guarantee twice the speed: communication between machines adds overhead, and results depend on how well the software divides the workload.
What You Need to Get Started
- Two Acer Veriton GN100 workstations, each with its own supplied power adapter.
- A QSFP cable confirmed compatible with the GN100 ConnectX-7 ports and intended link speed.
- Compatible system software, drivers, and administrative access on both units.
- An AI framework or inference runtime that supports execution across multiple nodes.
A direct two-system connection does not require a high-speed network switch. Use NVIDIA's Connect Two Sparks playbook as a platform reference for network and SSH setup, and check Acer's instructions for your GN100 hardware. Test communication before launching the distributed workload.
Do the Workstations Have to Be Stacked
No. Physical placement and AI clustering are separate choices. Acer describes the GN100 as stackable in a two-unit arrangement, but the systems can also sit side by side. Keep ventilation openings clear and choose a cable length suitable for your layout.
Start With One and Expand When You Need More
A second GN100 can make sense for developers whose target workload needs more memory than one unit provides and supports distributed execution. If your model already fits on one system, test its performance first to see whether another unit would address your actual bottleneck.
Explore the Acer Veriton GN100 VGN100-UD13 at AAAWave for current pricing and availability, or browse our AI computer collection to compare options for your local AI workspace.
