Flexible Universal GPU Server for Gen-AI
KR4268V2 is a powerful 4U AI server with the latest PCIe Gen 5.0 architecture, supporting 8 FHFL PCIe GPU cards and up to 2 additional double-width PCIe GPUs. Its flexible architecture and topologies support a variety of advanced AI applications including generative AI, AI cloud, IVA, video processing, and more.
 |
Advanced PerformancePCIe 5.0 architecture supports up to 600W power consumption, high-throughput data transmission up to 64GB/s bandwidth channel Built-in 8TB system memory and 300TB local high-speed storage for improved data access speed Supports NVIDIA L40S, H100, and A100 GPUs |
|
|
|
 |
Enhanced ArchitectureLatest CXL1.1 interconnect technology supports storage-level memory space expansion Supports latest 400G NDR infiniband network and 400G OCP 3.0 |
|
|
|
 |
Flexible ConfigurationsModular design concept supports flexible component-level configuration 4 CPU-GPU topologies for various AI application scenarios |
Available Models
|
|
|
|---|---|---|
| Form Factor | 4U rack server | |
| GPU | 8x FHFL double-width PCIe 5.0 GPU cards Supports additional 2x FL double-width PCIe GPU, 1x FL single-width PCIe GPU Supports L40S, H100, A100 |
|
| CPU | 2x 4th Gen Intel® Xeon® Scalable | 2x AMD EPYC™ 9004 |
| Memory | 32x DDR5 4800MT/s RDIMMs | 24x DDR5 4800MT/s RDIMMs |
| Storage | 24x 2.5” SSD, up to 16x NVMe U.2 2x NVMe/SATA M.2 |
|
| PCIe | 11x PCIe 5.0 x16 slots | |
| Network Interface | 1x OCP 3.0, supports NCSI | |
| Cooling | Air cooling | |
| Management | ASPEED 2600 | |
| Security | TPM 2.0 (Trusted Platform Module) | |
| Power Supply | 4x 12V 3000W, 2+2 | |
| Dimensions | 447mm (W) x 174.5mm (H) x 850mm (D) | |
