Supermicro's First to Market NEBS Level 3 Certified 1U Server 

Delivers 2,560 NVIDIA GPU Cores for 5G Edge AI & VR Innovation

Supermicro has recently introduced the industries-first 1U NEBS (Network Equipment Building System) as a Level 3 certified system with up to 4x NVIDIA V100 or V100S Tensor Core GPUs. This newly innovative server system allows for the latest AI, VR/AR, transcoding, and other high compute data workloads fitted with low latency requirements in the telecom environment.

Charles Liang, CEO of Supermicro, has stated this this first-to-market 1U NEBS Level 3 V100 GPU accelerated server will be a key enabler for the transition to 5G. With industry adoption of the most advanced applications and workloads in IoT and AI/AR.

Supermicro's V100 1U GPU System Relevance To The Market 

With the growing market development from 4G to 5G, there will be more complex and demanding workloads/applications within the Edge. This new system allows for extensive temperate control of 55 °C, and meets shock, vibration, flame resistance and other particular strict testings required for the full NEBS level 3 compliance, with NEBS 3 being the strictest level.

The system’s foundation is based on the 2^nd Gen Intel Xeon Scalable processors that provides built-in AI acceleration and Intel Optane persistent memory to enable NVIDIA GPUs to their fullest potential.

The hardware and architectural design significantly benefit the telecommunications market by giving them the ability to upgrade their data centre with higher-performance GPUs to accelerate voice and data traffic. In addition, this provides a new edge application and service by using a server that meets high reliability standards.

The Difference between NVIDIA V100 and V100S Tensor Core GPUs

The V100S delivers an increasing 17.1% higher single and double-precision performance when compared to the V100 with the same PCIe format. Overall, having a 16.1% better Tensor performance as NVIDIA has clocked the memory on the V100s model a little more quicker.

Relative to memory, the V100 provides 16TB and the V100S 32TB. The V100S maintains the 4,096-bit memory with 26% more memory bandwidth when compared to the V100 model. The upgrade to the V100S also has no effect on the graphic card’s TDP (thermal design power), with both models rated 250W.

The NVIDIA V100S is an ideal choice for workloads that require higher compute via higher core clock speeds over the NVIDIA V100. However, both are still the most advanced data centre GPU ever built, ideal for virtualisation, advanced design and HPC.