XingBan Ark · built in-house
One server carries the AI load of an entire care home.
Compute produces tokens, storage retains data, orchestration runs agents. A single Ark concentrates scheduling, storage and role agents in one server, with no additional machines required.
- 8
- Logical servers from one Ark server, minimum.
- 60%
- Reduction in total cost of ownership (Contract Annex A condition).
- 384TB
- Maximum on-premise storage across the range.
Three forces, one machine
Compute produces tokens, storage retains data, orchestration runs agents.
Care homes usually cost a server on compute alone. The form factor is a single machine, with dual-machine redundancy where high availability is required; video, health records and model files stay on site, and storage and scheduling make up the larger part of the bill.
- Compute
- Runs model inference and fine-tuning.
- Storage
- Compliant retention of video, health records and model files.
- Orchestration
- Eight or more role agents on one machine, each isolated.
The naming rule is embedded in the model number: tier, generation, VRAM, memory and storage. X1 covers entry inference, X2 standard training and X3 high-density compute.
- X1 · 4U
- On-premise inference for small and mid-size homes and single sites; at least 8–12 logical servers.
- X2 · 5U
- Regional groups, private fine-tuning and district compute centres; at least 16–24.
- X3 · 6U
- City hubs, hundred-billion-parameter training and fully domestic stacks; at least 32–64.
- Whole range
- Dual 10GbE / dual 25G RDMA / 4×100G IB; NVMe throughout.
One server replaces several devices together with the room space, power and maintenance they require, reducing TCO by more than 60%. XingBan OS and the models ship pre-deployed, so implementation days are not charged separately.
- At least 8 logical servers (1 scheduling + 1 distributed storage + 6 role agents).
- TCO reduced by more than 60%.
- Up to 384TB of on-premise storage.
- Full-stack support for domestic CPU, GPU and operating systems.
- Ships with XingBan OS; models pre-deployed at the factory.
- Six built-in scenario agent templates: care, medical, security, logistics, management and service desk.
Sizing
Bed count determines the model; a POC settles any doubt.
A single mid-site needs one entry-level X1 configuration. Regional groups and chains move to X2. A regional compute hub requires X3. Where sizing is unclear, the POC conclusion governs.
Book a demo
Title
Thirty minutes to observe nine digital workers operating in a real care setting. An advisor will call to confirm the time and focus.