10U High-Performance AI Training Server
KR9288 is a 10U high-performance AI computing platform designed for large language model (LLM) training and multimodal inference workloads. It features an 8-GPU NVLink module based on the latest NVIDIA Blackwell platform, now supporting NVIDIA HGX™ B300. It also includes support for OAM accelerators like Intel® Gaudi®.
This dual-processor system is configurable with the latest Intel or AMD processors to offer flexibility for diverse AI deployment needs. KR9288 enables extreme compute density and bandwidth for modern AI model development while maintaining stability with an advanced power and cooling design.
![]() |
Scalable GPU InfrastructureSupports 8x NVIDIA B300 architecture GPUs via NVLink and NVSwitch interconnect 2.3TB HBM3e memory size, 64TB/s GPU memory bandwidth for generative AI workloads Delivers 50% higher FP4 performance than previous generation |
|
|
![]() |
Reliable Power & Thermal Design12V/54V separated power delivery, N+N PDU level redundancy PSU redundancy (12V/54V: N+1 or N+N) ensures system uptime Front I/O design reduces failure points and improves signal stability |
|
|
![]() |
Robust Network and Storage ExpansionSupports 8x CX8 InfiniBand and 4x full height N-S network fabric Up to 8x U.2 NVMe drives + 2x M.2 NVMe; support for RAID card expansion |