• Worldwide delivery & Support
  • up to 6 years Warranty
  • 24/7 Hotline +49 (0) 4181 235770

GPU Computing with NVidia CUDA and Supermicro

27/09/2017 -
  • Datacenter & Cloud
  • |
  • High Availability
  • |
  • HPC & GPU
  • |
  • Server & Storage
  • |
  • Uncategorized
  • |

GPU Computing with NVidia CUDA and Supermicro enables new alternatives in the HPC or High Performance Computing area. In recent years, it has become clear that GPU-based computing systems offer maximum performance and energy efficiency. This makes them the first choice for technical and scientific applications. This optimization of the computing power is achieved by building a hybrid system of CPU and GPU and using NVidia’s CUDA or the programming language OpenCL (for AMD) the computing load is then distributed by the software to the parallel cores of the GPU and made usable there.


NVidia CUDA was introduced by Nvidia in 2006 and offers the possibility to extend the processor with the graphics processor using the CUDA API. This gives you the possibility of HPC (High Performance Computing). CUDA works with all NVIDIA GeForce®, Quadro® and Tesla® GPUs. In the business sector, NVIDIA Tesla® series is the first choice for maximum performance. NVidia’s latest graphics processor solution, the NVIDIA® Tesla® P100, offers the most innovative data center solution to solve application cases in a high-performance and energy-efficient way. It achieves maximum energy efficiency and performance. Almost every HPC application and deep learning framework can be optimized to accelerate them to a level never reached before. By using multiple GPUs, you can usually lose significant amounts of system performance, but NVidia’s NVLink high-speed connectivity technology offers a way to negate the previous performance loss by increasing multi-GPU performance by a factor of five.


Supermicro uses NVidia CUDA in its latest HPC server portfolio to ensure maximum performance and energy efficiency for HPC applications in the SYS-1028GQ-TRT and SYS-4028GR-TXR (T). Both systems allow you to reduce the maximum cost of ownership and optimize your maintenance costs. The Supermicro SYS-1028GQ-TRT can support up to 4 NVIDIA® Tesla® K80 dual GPUs. It also supports Intel Xeon Phi? The system also supports up to 2 Intel® Xeon® E5-2600 v3/v4 CPUs, up to 1TB ECC DDR4-2133 MHz in 16 DIMM slots, 2x 2x 2.5 “hot-swap SATA drive bays and 2 2.5” internal memory slots. In addition, support for dual 10GBase-T ports and intelligent redundant 2000W (1+1) Titanium Level Titanium level high efficiency power supplies is provided. The SYS-1028GQ-TXR (T) has the form factor 1U. The Supermicro SYS-4028GR-TXR (T) is the extension of the SYS-1028GQ-TXR (T). It offers support for the Intel® Xeon® E5-2600 v4 CPUS. It extends the 4 NVIDIA® Tesla® K80 dual GPUs to up to 8 NVIDIA® Tesla® P100 accelerators. The server board offers 2 versions, the TXR Super X10DGQ-O+-CPU or the TXRT Super X10DGQ-OT+-CPU. It also supports up to 8 NVMe SSDs and supports up to 4 2000W of redundant energy efficient power supplies. The SYS-4028GR-TXR (T) has the form factor 4U.

About the Author



Do you have any questions about my article? Get in touch! Call +44 203 318 5994 or send me a mail to jpaegelow@happyware.com