Compare GPU systems
The GPU-server range side by side — from a 2-GPU inference box to an 8-GPU HGX cluster node. Pick by model-fit and budget; every system ships Make-in-India with a GST invoice and pan-India onsite support.
| Best for | Dev & inference | Team training | 70B training | Cluster node / 405B |
|---|---|---|---|---|
| GPUs | 2× L40S 48GB | 4× L40S 48GB | 4× H100 SXM | 8× H100 SXM |
| GPU memory | 96 GB | 192 GB | 320 GB | 640 GB |
| Model fit | 7–13B | up to 34B | 70B | 70B–405B |
| CPU | 2× Xeon Silver | 2× Xeon Gold | 2× EPYC 9004 | 2× EPYC 9004 |
| System memory | 512 GB | 1 TB | 2 TB | 2 TB |
| Storage | 2× 1.92 TB NVMe | 4× 3.84 TB NVMe | 8× 7.68 TB NVMe | 8× 7.68 TB NVMe |
| Networking | 2× 25GbE | 2× 100GbE | 400G InfiniBand | 400G IB ×2 |
| Cooling | Air | Air | Direct liquid | Direct liquid |
| Form factor | 2U | 4U | 4U | 8U |
| Max power | ~1.9 kW | ~2.6 kW | ~5.6 kW | ~6.8 kW |
| Buy mode | Buy now | Buy now | Configure | Made to order |
Need a different mix? Configure your own → or talk to a solution architect for a sizing recommendation.