Choose your chassis

7U AMD Instinct MI325X eight-GPU server with dual AMD EPYC 9005 processors, 24 DDR5 DIMMs, ten NVMe bays and a one-GPU-to-one-NIC topology — a modular flagship for large AI models and HPC workloads.

2× AMD

up to 8× GPU

6 TB RAM max

10 drive bays

2U dual-socket GPU server on 5th Gen Intel Xeon Scalable with four dual-slot GPUs (NVLink-ready), 16 DDR5 DIMMs, six tri-mode NVMe/SATA/SAS bays and redundant 2600 W Titanium PSUs — independent CPU/GPU airflow tunnels keep it quiet under load.

2× Intel

up to 4× GPU

4 TB RAM max

6 drive bays

AMD EPYC 9004/9005 single-processor 2U GPU server with four dual-slot GPUs, 12 DDR5 DIMMs, 8 PCIe 5.0 slots and six tri-mode NVMe/SATA/SAS bays — a compact, power-efficient node for inference and visual computing.

1× AMD

up to 4× GPU

3 TB RAM max

6 drive bays

Intel Xeon 6 dual-processor 4U NVIDIA MGX server supporting eight dual-slot GPUs (H200, RTX PRO 6000 Blackwell or Gaudi 3), 32 DDR5 DIMMs and five PCIe 5.0 expansion slots — built for AI training, inference and HPC.

2× Intel

up to 8× GPU

4 TB RAM max

8 drive bays

AMD EPYC 9004/9005 dual-processor 4U server supporting eight dual-slot PCIe GPUs (H200, RTX PRO 6000 Blackwell), 24 DDR5 DIMMs, eight U.2 NVMe bays and 2+2 redundant 3200 W Titanium PSUs — an AI/HPC workhorse with independent CPU and GPU airflow tunnels.

2× AMD

up to 8× GPU

3 TB RAM max

8 drive bays

Density-optimized 2U AMD EPYC 9004 server packing up to ten dual-slot PCIe GPUs, 32 DDR5 DIMMs and eight U.2 NVMe bays into a short-depth chassis — maximum inference throughput per rack unit.

2× AMD

up to 10× GPU

8 TB RAM max

8 drive bays

Supermicro 4U dual AMD EPYC 9004 GPU server with eight dual-slot PCIe accelerators, 24 DDR5 DIMMs and mixed NVMe/SAS storage bays — a flexible platform for AI training, rendering and virtualization.

2× AMD

up to 8× GPU

6 TB RAM max

12 drive bays

Supermicro 8U flagship with the NVIDIA HGX H200 8-GPU SXM baseboard, dual 5th Gen Intel Xeon, 24 DDR5 DIMMs and 3+3 redundant 3000 W PSUs — NVLink-connected GPUs for large-model training at full tilt.

2× Intel

up to 8× GPU

8 TB RAM max

8 drive bays