Cirrascale Adopts NVIDIA HGX H100 Servers to Enhance AI and HPC Efficiency

0
74

Cirrascale

Cirrascale Cloud Providers, a international cloud providers supplier and operator of its AI Innovation Cloud, might be integrating the newest improve to NVIDIA HGX AI supercomputing platform into its specialised multi-GPU deep studying cloud. The improve options NVIDIA HGX H100 servers, which might present a major enchancment in efficiency, scalability, and safety.

The servers are powered by the NVIDIA Hopper structure and are geared up with the NVIDIA H100 Tensor Core GPU, which might allow a ten-fold enhance in efficiency in comparison with earlier fashions. The combination of this know-how would permit Cirrascale to supply a variety of providers, together with small enterprise workloads, high-performance computing, and trillion-parameter synthetic intelligence (AI).

The multi-GPU deep studying cloud could be securely accelerated with the assistance of the NVIDIA H100 Tensor Core GPU. In keeping with Mike LaPan, Vice President of Cirrascale Cloud Providers, the mixing of the NVIDIA H100 Tensor Core GPU gives breakthrough innovation that’s important in at the moment’s AI panorama. By using the NVIDIA Hopper structure, the NVIDIA H100 GPU can pace up massive language fashions by 30 instances in comparison with the earlier era.

NVIDIA HGX H100 Clusters

Mike LaPan, Vice President of Cirrascale Cloud Services
In keeping with Mike LaPan, Vice President of Cirrascale Cloud Providers, the mixing of the NVIDIA H100 Tensor Core GPU gives breakthrough innovation that’s important in at the moment’s AI panorama.

Cirrascale gives large-scale NVIDIA HGX H100 clusters that make the most of the NVIDIA Quantum-2 InfiniBand networking know-how, permitting customers to expertise unparalleled utility efficiency throughout quite a few computer systems.

Moreover, the NVIDIA HGX H100 server structure, offered by Cirrascale, can construct among the strongest scale-up servers on this planet, combining eight NVIDIA H100 GPUs with high-speed interconnect know-how developed by NVLink and NVSwitch. An eight-way NVIDIA HGX H100 gives greater than 32 petaFLOPS of FP8 deep studying compute efficiency, made doable by using the facility of the multi-precision Tensor Cores discovered within the H100.

Dave Salvator, Director of Accelerated Computing Applied sciences at NVIDIA, acknowledged that the mixing of AI and high-performance computing can revolutionize organizations and society by fixing among the most difficult points on this planet. NVIDIA’s collaboration with cloud builders comparable to Cirrascale can be serving to to stretch the boundaries of AI and HPC investigations, leading to a brand new surge of discoveries which might be to the collective benefit of society.

Supply By https://hostingjournalist.com/cirrascale-adopts-nvidia-hgx-h100-servers-to-boost-ai-and-hpc-performance/