For optimum acceleration of extremely parallel functions like synthetic intelligence (AI), deep studying, autonomous car techniques, vitality and engineering/science, Supermicro’s new 4U system with next-generation NVIDIA NVLink is optimized for total efficiency. The SuperServer 4028GR-TXRT helps eight NVIDIA Tesla V100 SXM2 GPU accelerators with most GPU-to-GPU bandwidth for necessary HPC clusters and hyper-scale workloads. Incorporating the newest NVIDIA NVLink GPU interconnect know-how with over 5 instances the bandwidth of PCI-E 3.0, this technique features an unbiased GPU and CPU thermal zoning design, which ensures uncompromised efficiency and stability below probably the most demanding workloads.
Equally, the efficiency optimized 4U SuperServer 4028GR-TRT2 system can help as much as 10 PCI-E Tesla V100 accelerators with Supermicro’s progressive and GPU optimized single root complicated PCI-E design, which dramatically improves GPU peer-to-peer communication efficiency. For even larger density, the SuperServer 1028GQ-TRT helps as much as 4 PCI-E Tesla V100 GPU accelerators in solely 1U of rack area. Ultimate for media, leisure, medical imaging, and rendering functions, the highly effective 7049GP-TRT workstation helps as much as 4 NVIDIA Tesla V100 GPU accelerators.
“Supermicro designs probably the most application-optimized GPU techniques and provides the widest choice of GPU-optimized servers and workstations within the trade,” stated Charles Liang, President and CEO of Supermicro. “Our high efficiency computing options enable deep studying, engineering and scientific fields to scale out their compute clusters to speed up their most demanding workloads and obtain quickest time-to-results with most efficiency per watt, per sq. foot and per greenback. With our newest improvements incorporating the new NVIDIA V100 PCI-E and V100 SXM2 GPUs in performance-optimized 1U and 4U techniques with next-generation NVLink, our clients can speed up their functions and improvements to assist remedy the world’s most complicated and difficult issues.”
“Supermicro’s new high-density servers are optimized to totally leverage the new NVIDIA Tesla V100 knowledge heart GPUs to supply enterprise and HPC clients with a wholly new degree of computing effectivity,” stated Ian Buck, vice chairman and normal supervisor of the Accelerated Computing Group at NVIDIA. “The new SuperServers ship dramatically increased throughput for compute-intensive knowledge analytics, deep studying and scientific functions whereas minimizing power consumption.”
With the convergence of Massive Knowledge Analytics, the newest NVIDIA GPU architectures, and improved Machine Studying algorithms, Deep Studying functions require the processing power of a number of GPUs that should talk effectively and successfully to develop the GPU network. Supermicro’s single-root GPU system permits a number of GPUs to speak effectively to attenuate latency and maximize throughput as measured by the NCCL P2PBandwidthTest.
For complete info on Supermicro NVIDIA GPU system product traces, please go to https://www.supermicro.com/products/nfo/gpu.cfm.