Overview
Innovations
-
Highly Optimized GPU-to-GPU Interconnect ArchitecturesChoose from a broad range of Supermicro GPU optimized SuperServers supporting NVIDIA® NVLink™, Dual-Root Complex or Single Root Complex PCI-E interconnect architectures for target applications and the best AI / Deep Learning performance. -
Accelerated Deployment and Peace of MindRisk-free investments for guaranteed quality and system performance with our Qualified GPU and Supermicro Server Platform List.
Product Features
Form Factor
1U, 2U, 4U and 10U Rackmount
Memory
Up to 24 DIMM slots, up to 6TB DDR4 memory, Intel® Optane™ DC persistent memory
Power
Redundant Titanium (96%) and Platinum Level Power Supplies
Management
Open Industry Standard IPMI, Redfish APIs, Rack Scale Management
Processors
Dual or single Intel® Xeon® Scalable processors up to 205W TDP
Drives
Up to 16x U.2 NVMe drives with 2.5" and 3.5" SAS3/SATA3 options
Input/Output
Up to 4 GPUs in 1U, 6 GPUs in 2U, 20 GPUs in 4U and 16 GPUs in 10U
Services
Worldwide Global Services
Product Lines
1U GPU Line
 |
Up to 4 GPUs |
 |
NVIDIA® NVLink™ or Dual PCI-E Root Complex |
 |
2 Processors |
 |
16 / 12 / 8 / 6 DIMM Slots |
 |
Up to 2000W Redundant Titanium Level PSU |
2U GPU Line
|
Up to 6 GPUs |
|
Dual PCI-E Root Complex |
|
2 Processors |
|
16 DIMM Slots |
|
Up to 2000W Redundant Platinum Level PSU |
4U / Tower GPU Line
|
Up to 20 GPUs |
|
NVIDIA® NVLink™, Single or Dual PCI-E Root Complex |
|
2 Processors |
|
24 / 16 DIMM Slots |
|
Up to 2200W Redundant Titanium Level PSU |
10U GPU Line
|
16 NVIDIA® Tesla V100 GPUs |
|
NVIDIA® NVLink™ and NVSwitch |
|
2 Processors |
|
24 DIMM Slots |
|
6x 3000W Redundant Titanium Level PSU |
Resources
Supermicro SuperMinute: 10U GPU
As deep learning models continue to increase in size and complexity, the industry is looking for more and more computational capability. To meet the needs of data scientists, we are introducing the most powerful GPU system in the industry for artificial intelligence and high-performance computing.
Play the video
SuperServers Optimized For NVIDIA® Tesla® GPUs
SuperServers Optimized For NVIDIA® Tesla® GPUs With the Fastest GPU to GPU Interconnections - Most Powerful Solution for Deep Learning Training.
View the flyer
High Performance GPU Accelerated Virtual Desktop Infrastructure Solutions with Supermicro Ultra SuperServers
Supermicro has developed a line of fully validated server solutions featuring high performance GPUs from NVIDIA, along with NVIDIA virtual GPU (vGPU) software, to address the rapidly growing Virtual Desktop Infrastructure (VDI) market.
View the white paper
1U 4 GPU Server White Paper
1U 4 GPU Solution Extends Supermicro’s Leadership in Supercomputing Servers
View the white paper
GPU/Coprocessor Server Solutions Flyer
The Supermicro GPU/Coprocessor SuperServer® systems deliver the maximum processing acceleration for the most compute intensive workloads in the smallest physical dimensions.
View the flyer
Supermicro at GTC 2018
Supermicro at GTC 2018 displays the latest GPU-optimized systems that address market demand for 10x growth in deep learning, AI, and big data analytic applications with best-in-class features including NVIDIA® Tesla® V100 32GB with NVLink and maximum GPU density.
Play the video
4U 8 GPU
The 4028GR-TXRT is Supermicro’s most powerful GPU Server delivering supercompute level performance for Deep Learning applications. It includes 2 Intel® Xeon® E5 v4 CPUs and 8 Pascal Generation Tesla P100 GPUs, delivering 170 TeraFLOPs of performance in a 4U system with no thermal limitations.
Play the video
Models
SKU
Generation
Features
