Advanced Computing in the Age of AI | Friday, March 29, 2024

AMAX to Offer New NVIDIA Tesla GPU Accelerators in ClusterMax and ServMax Platforms 

AMAX today announced that it will offer the new NVIDIA Tesla K80 family of GPU accelerators in its ClusterMax GPU cluster and ServMax GPU server platforms. Featuring two high-performance GPU processors in a single board, the Tesla K80 dual-GPU accelerator delivers 10 times higher performance than today’s fastest CPU, outperforming CPUs and competing accelerators on hundreds of complex analytics and large, computationally intensive scientific computing applications.

The ClusterMax and ServMax platform solutions featuring the new NVIDIA Tesla K80 dual-GPU accelerators are ideal for universities, national research labs, government, and enterprise high-performance computing environments where massive computational power is required. To meet these demands, AMAX offers up the ClusterMax SuperG, a complete turnkey GPU-enabled cluster that has been pre-configured, ready to deploy on-site and configurable with up to 108 NVIDIA Tesla K80 dual-GPU accelerators in a standard 42U cabinet, delivering over 604+ Teraflops of peak single precision performance and 201+ Teraflops of peak double precision performance per standard 42U cabinet.

“With NVIDIA’s high-performance, energy-efficient Tesla K80 dual-GPU accelerators and AMAX’s dynamic cluster and server solution designs, we are able to deliver breakthrough performance and compute efficiency for double-precision focused applications,” said Julia Shih, Director of Strategic Alliances at AMAX. “With multi-GPU Hyper-Q, system monitoring, memory protection, NVIDIA GPUDirect with RDMA technologies embedded in the NVIDIA GPU compute architecture, the performance increases we’ve seen in data analytics, satellite imaging, and weather modeling applications have simply been astounding.”

The latest AMAX solutions designed for the Tesla K80 accelerators include:

ClusterMax SuperG GPU Cluster:

  • Up to 539,136 GPU cores, 604+ Teraflops of single precision and 201+ Teraflops of double precision performance and 2,592GB ultra-fast GDDR5 GPU memory per 42U cluster
  • Features FDR (56Gb/s) Infiniband interconnect and NVIDIA GPUDirect technology for high-speed cluster communication
  • Dual Intel Xeon E5-2600 v3 series processors with up to 512GB high speed DDR4 2133MHz memory per node
  • Integrated Cluster Management to manage and monitor cluster & GPU resources

 

Preconfigured AMAX ClusterMax SimCluster:

  • Off-the-shelf and ready-to-deploy Tesla K80 dual-GPU cluster solution with both hardware and software, delivering 67+, 134+, 268+ Teraflops of single precision, and 22+, 44+, 44+ Teraflops of double precision performance per cluster
  • Pre-configured with 4x, 8x, or 16x Tesla K80 dual-GPU cluster nodes, with 12x, 24x, 48x GPU accelerators per cluster
  • Dual Intel Xeon E5-2600 v3 series processors with up to 512GB high speed DDR4 2133MHz memory per node
  • Integrated Cluster Management to manage and monitor cluster & GPU resources

 

ServMax XG-2401Gk8 GPU Server:

  • 2U rack server with support for 4x Tesla K80 dual-GPU accelerators ( > 22+ Teraflops SP performance, 7+ Teraflops DP performance, 19,968 CUDA cores, 96GB of dedicated GPU memory)
  • Dual Intel Xeon E5-2600 v3 series processors with up to 512GB high speed DDR4 2133MHz memory
  • Options for on-board FDR Infiniband interconnectivity

 

ServMax  XG-4802Gk8 GPU Server:

  • 4U rack server with support for 8x Tesla K80 dual-GPU accelerators ( > 44+ Teraflops SP performance, 14+ Teraflops DP performance, 19,968 CUDA cores, 96GB of dedicated GPU memory)
  • Dual Intel Xeon E5-2600 v3 series processors with up to 1TB high speed DDR4 2133MHz memory
  • Options for on-board 10GbE interconnectivity

 

EnterpriseAI