Skip to main content

2U Hyper-U A+ Server

2U Hyper-U with 12 hot-swap 3.5" drive bays and up to 8 PCIe 5.0 slots

Key Features:

  • Single AMD EPYC™ 9004/9005* Series Processor (*AMD EPYC™ 9005 series drop-in support requires board revision 2.x) Support up to DDR5 24 DIMM slots (2DPC) (Please check System Memory Section for detail)
  • Optional PCIe slot configurations up to 8 PCIe 5.0 x8 or 4 PCIe 5.0 x16 slots or mix and match
  • Flexible networking options with 1 AIOM networking slot (OCP NIC 3.0 compatible)
  • 12 Hot-swap 3.5"/2.5" NVMe/SATA/SAS drive bays, 2 M.2 PCIe 3.0 NVMe slots
  • Support up to 4 GPUs
  • Dual 1200W Redundant Titanium Level power supplies (Power supply full redundancy based on configuration and application load)
Supermicro 2U Hyper-U system (AS -2015HS-TNR) for VDURA Solution

AS -2015HS-TNR

90-Bay Top-Load JBOD

4U • 90 x 3.5" SAS3 • 12Gb/s Single Expander • 2000W 1+1 Redundant

Key Features:

  • Single-Path Storage Enclosure
  • Single canister controllers with field serviceable tray
  • Support 90 3.5"/2.5" Top Loading SATA 6Gb/s / SAS3 12Gb/s Hot-Swappable HDDs
  • 6x SAS3.0 IO ports (MiniSAS HD/SFF-8644)
  • Dedicated IPMI RJ45 management port
  • Redfish OOB management protocol support
  • SCSI Enclosure Services SES-4 support
  • Hot-swappable Tool-less Modular Design for Easy Service and Easy Maintenance
  • Redundant Titanium 2000W power supply
Supermicro 4U 90-Bay Top-Loading JBOD (CSE-947HE1C-R2K05JBOD) for VDURA Solution

CSE-947HE1C-R2K05JBOD

VDURA Data Platform on Supermicro

V5000 Configuration options
V5000 configuration options, left to right: all flash, 50 percent flash, 98 percent HDD.

The VDURA Data Platform is modern data storage infrastructure software that runs on Supermicro's industry-standard server and storage platforms. VDURA holds the first AI Data Storage Reference Design with AMD and is a technology partner with NVIDIA. Configure from 100% flash for low-latency inference to mixed-fleet NVMe + HDD for cost-optimized training and retention, all under a single namespace with intelligent data tiering and end-to-end encryption. Start at a single rack. Scale linearly to exabytes.

Configurations range from all-flash NVMe for inference and checkpoint-heavy workloads, to hybrid NVMe + HDD for cost-optimized training and long-tail data, all managed under one global namespace with intelligent data tiering and end-to-end encryption.