DP Intel 8U AI Training Server with 8 Intel Gaudi 2 GPU

Anewtech-Systems Rackmount-Server Supermicro SYS-820GH-TNR2 GPU Server Supermicro Singapore Superserver Supermicro Servers
  • Dual Intel® Xeon® 3rd Gen Scalable Processors (Ice Lake) TDP up to 270W
  • Intel® C621A Chipset
  • 32x DIMM slots, up to 8TB Registered ECC DDR4 3200MHz SDRAM
  • 2x PCIe 4.0 x16 FHFL slot and 2x PCIe 4.0 x8 FHFL slot
  • 6x 400GbE QSFP-DD for scale-out and 1x 1GbE BaseT (dedicated IPMI port)
  • 4x 2.5”NVMe Hot-Swap Drive Bays
  • Up to 16x 2.5" SAS or SATA Hot-Swap Drive Bays (Additional SAOC is required)
  • Up to 8x 2.5" SATA Hot-Swap Drive Bays
  • 12x Removable Heavy-Duty FANs (Includes 2x Dummy PSU FAN)
  • 6x 3000W(54V+12V) Redundant Power Supplies, Titanium Level
Product Specification
Product SKUsSuperServer SYS-820GH-TNR2 (Silver)
MotherboardSuper X12DPG-OA6-GD2
CPUDual Socket P+ (LG-4189) 
3rd Gen Intel® Xeon® Scalable Processors
Core CountUp to 40C/80T; Up to 60MB Cache per CPU
NoteSupports up to 270W TDP CPUs (Air Cooled)
System Memory
MemoryMemory Capacity: 32 DIMM slots Up to 8TB: 32x 256 GB DRAM  
Memory Type: 3200/2933/2666MHz ECC DDR4 RDIMM/LRDIMM
Memory Voltage1.2 V
Error DetectionECC
On-Board Devices
ChipsetIntel® C621A
Network Connectivity6x 400GbE QSFP-DD1x 1GbE BaseT 
Note: IPMI LAN port on CPU processor tray, and QSFP-DD ports are for AI Scale-out purpose
Input / Output
USB2 USB 3.0 port(s) (2 USB 3.0)
Video1 VGA port(s)
System BIOS
Power ConfigurationsACPI Power Management 
Power-on mode for AC power recovery
HardwareTrusted Platform Module (TPM) 2.0
PC Health Monitoring
CPU7+1 Phase-switching voltage regulator 
Monitors for CPU Cores, Chipset Voltages, Memory
FANFans with tachometer monitoring 
Pulse Width Modulated (PWM) fan connectors 
Status monitor for speed control
TemperatureMonitoring for CPU and chassis environment 
Thermal Control for fan connectors
Form Factor8U Rackmount
Dimensions and Weight
Height14" (356mm)
Width17.6" (447mm)
Depth31.5" (800mm)
Package26.8" (H) x 40" (W) x 55.2" (D)
WeightGross Weight: 282 lbs (128 kg)
Available ColorSilver
Front Panel
ButtonsPower On/Off buttonReset button
Information LED 
Information LED 
Power On LED 
PSU Failed LED
Expansion Slots
PCI-Express (PCIe)2 PCIe 4.0 x16 FHFL slot(s)2 PCIe 4.0 x8 FHFL slot(s)
Drive Bays / Storage
Hot-swap20x 2.5" hot-swap NVMe/SATA/SAS drive bays 
(4x 2.5" NVMe dedicated; Up to 8x 2.5" SATA dedicated or Up to 16x 2.5" SAS/SATA supported by additional SAOC)
M.22 M.2 NVMe 
M-Key, 2280/22110
System Cooling
Fans12 Removable Heavy-Duty 84x203.5x82.4mm Fan(s)
Power Supply18000W Redundant Titanium Level power supplies
CertificationTitanium Level
Operating Environment
Environmental Spec.Operating Temperature: 10°C ~ 35°C (50°F ~ 95°F)  
Non-operating Temperature: -30°C to 60°C (-22°F to 140°F)  
Operating Relative Humidity: 8% to 80% (non-condensing)  
Non-operating Relative Humidity: 8% to 90% (non-condensing)
Supermicro Gaudi®2 AI Training Server (SYS-820GH-TNR2)

Building on the success of the original Supermicro Gaudi AI training system, the Gaudi2 AI Training Server prioritizes two key real-world considerations: integrating multi AI training system to analyze diverse AI models faster, while simultaneously multiple scalability function and price advantages.

Key Features

  • GPU: 8 Gaudi2 HL-225H mezzanine cards
  • CPU: Dual 3rd Gen Intel® Xeon® Scalable processors
  • Memory: 32 DIMMs - up to 8TB registered ECC DDR4-3200MHz SDRAM
  • Drives: up to 24 hot-swappable hard drives (SATA/NVMe/SAS)
  • Power: 6x 3000W High efficiency (54V+12V)
  • Networking: 24x 100GbE (48 x 56Gb) PAM4 SerDes Links by 6 QSFP-DDs
  • Expansion Slots: 2x PCI Gen4 switches
  • Workloads: AI Training and Inferencing


The Gaudi2 processor significantly increases training and inference performance, building on the high-efficiency, first-generation Gaudi architecture that delivers up to 40% better price performance in the AWS cloud with Amazon EC2 DL1 instances and on-premises with the Supermicro Gaudi®2 AI Training Server. 

Gaudi2 delivers training and inference performance on a whole new level with a leap from the first-gen Gaudi 16nm process technology to Gaudi2’s 7nm. It increases the number of AI-customized Tensor Processor Cores–from 8 to 24, adds support for FP8 and integrates a media processing engine for processing compressed media for offloading the host subsystem. Gaudi2’s in-package memory has tripled to 96 GB of HBM2e at 2.45 Terabytes-per-second bandwidth