AI コンピューティングサーバー
8-10GPU 液冷 AI サーバー
A cold-plate liquid-cooled platform for high power density, low PUE and scale-out clusters: 8-10 triple-width GPUs (up to 600W each) with full CPU/GPU cold-plate coverage, CDU and facility water-loop integration, and cluster-level coolant monitoring.
8-10×GPU8-10× GPU · 600W4U/5U ラックマウント8-10 triple-width GPUs · full cold-plate coverage · 12× PCIe 5.0
主な用途:モデル学習大規模モデル推論データセンター
仕様
| フォームファクタ | 4U/5U rack-mount |
|---|---|
| プロセッサ | Dual-socket Intel Xeon or AMD EPYC |
| GPU 拡張 | 8-10 triple-width GPUs, up to 600W each, for high-power inference and training |
| PCIe 拡張 | 12× PCIe 5.0 slots for high-speed network and storage expansion |
| 冷却 | Full CPU/GPU cold-plate coverage; CDU and facility water-loop ready; cluster-level coolant monitoring |
| 製品ポジション | Power density and efficiency first: more compute per rack, less fan noise and heat load |
* 構成は参考値です。最終仕様はプロジェクトBOM、納入仕様書、実測結果に基づきます。
特殊機能
CPU/GPU cold-plate liquid cooling対応
CDU & facility water-loop ready対応
Cluster-level coolant monitoring対応
製品の特長
GPU topology, PCIe 5.0, power and airflow engineering
BIOS/BMC, driver, CUDA and firmware compatibility validation
Burn-in, NCCL / GPUDirect RDMA and multi-node stress testing
Tiered system/cache/data drives from our enterprise SSD and DDR5 RDIMM lines
Design, integration testing, on-site deployment and O&M in one delivery