S4215
A new generation of rack AI server products, using NVIDIA NVlink interconnection technology, support 8 A800 in a 4U space, support NVLink600GB/s bandwidth interconnection between two adjacent cards, release powerful computing power, and further improve the communication performance between cards. Achieve better video memory and performance scaling to dramatically accelerate AI model training.

Processor

2x AMD EPYC™ 7003/7002 Series Processors  ,TDP 280W

Memory

32 x DIMM slots DDR4 3200MHz RECC DDR4 , up to 8TB

GPU

8 x  80GB HGX-8 A100 or 40GB HGX-8 A100

Backplane

6 x SAS3/SATA3/NVMe4

Disk drive

bays

6 x 2.5“SATA/SAS/NVMe hot-swap drive bays(2x NVMe dedicated) Optional rear 4 x hot-swap 2.5 NVMe

Storage

controller

offers a variety of controller options for internal drives

PCI-E slots

Front 1 * PCI-E 4.0 16 +1 * PCI-E 4.0 8 ,from CPU

Rear 8 * PCI-E 4.0 16 from PLX

1 x PCI-E 4.0 16 for AIOM/OCP 3.0

Ports

Front 2*USB 3.0 、 1* VGA  、 1*BMC LAN

Network

interfaces

2 x 1Gb/s RJ45

1 x AIOM

1 x 10/100/1000 Gb/s for IPMI

IPMI

IPMI 2.0, KVM/media over LAN (Aspeed® AST2500)

PCI-E slots

3000W 2+2 Redundant Power Supplies

Or 2200W 3+1 Redundant Power Supplies

Dimensions

W 446 x H 174 x L 900 mm