Cluster Solutions

Why use a 1000 node CPU cluster when you can solve your problem in less time, on a 16-node, hardware accelerated cluster?

Acceleware Cluster Solutions

* NVIDIA S1070 with up to 4 GPUs

Acceleware Cluster Solutions are used by Fortune 500 companies and research institutions around the world to provide massive compute power in a price range, footprint size and power envelope that can only be obtained by utlizing GPUs. Acceleware provides software and integration toolkits to power NVIDIA® Tesla™ S1070 GPUs to accelerate your data center. The Tesla S1070 contains four-GPUs in a 1U form factor and when combined with the Acceleware software libraries, the S1070 offers a powerful and scalable, enterprise ready data processing solution. If you work with extremely large simulations or data sets need to process them quickly and economically, then the Acceleware Cluster Solution is the answer.

Sample GPU Cluster Technical Specifications

Form Factor: Standard 19" 1U rack-mount chassis
Number of GPUs: 4 GPUs per node, up to 16 nodes
Number of Streaming Processor Cores: 960 (240 per processor)
Peak FLOPS Single Precision: 23.73 to 4.14 TFlops
Peak FLOPS Double Precision: 311 to 345 GFlops
Total Dedicated Memory: 16 GB per node
Floating Point Precision: IEEE 74 single-precision floating point
Power: 550W typ. (900W max.), 100-240 VAC Autosensing per node
Host Adapter Card: PCI Express x16 or x8, Small Form Factor, Passive (10W)
 

Features & Benefits

  Features   Benefits
     
 
Run immense simulations at unprecedented speeds
  
  • Run jobs up to 2.8 GCells (FDTD) using 96 GB of high speed Accelerator RAM.
  • Attain speeds of up to 14 GCells/s (FDTD) leveraging all 64 GPUs.
       
       
 
Reduce your design cycle and improve your products
  
  • Outcompute your competition by running larger jobs at accelerated speeds.
     
  • Add complexity to your simulations where computational limits prevented you from doing so before.
       
       
 
Scalability
  
  • Solution is also available in 4, 8 and 12 node, upgradeable configurations.
     
  • Let Acceleware manage the evolution of Accelerator hardware for you.
       
       
 
Process Data up to 50X Faster*
  
  • Process more data with a footprint and power consumption profile over 70% smaller than CPU based servers alone.
  • Provides outstanding return on IT investment.
  • Additional computing power provides capacity for extensive optimization of product designs.
       
       
 
16 GB of High BW GPU RAM per node
  
  • Run larger simulations.
     
  • Large memory size ensures GPU streaming processors are fully utilized, maximizing throughput.
     
  • Unmatched performance in high density form factors enables breakthrough levels of capability and productivity.
       
       
 
Supports Complex Simulations
  
  • Higher speeds enable increased complexity in simulations that was previously not practical.
     
  • Reliability tested and proven scientific models and algorithms.
       
       
 
Pre-Configured or Certified Hosts Available
  
  • Pre-configured hosts allow you to use your simulation package out of the box.
  • Certified hosts support the addition of GPU acceleration to existing server configurations.
       
       
  1U Form Factor  
  • 3.73 to 4.14 TFlops (Single Precision) or 311 to 345 GFlops (Double Precision) in 1U!
   
  • PCIe connection to host (two x16).
     
  • Power supply is self contained.
     
  • Highest performance of parallel applications.
       
       
  Compatible with Existing Software  
  • Use your familiar simulation packages.
  * Application dependant    

Performance

Computational ClusterThis graph shows the computational performance for an example EDA task and compares Acceleware's hardware accelerated clusters with a traditional CPU cluster. The task is representative of work required in antenna and RF circuit design, image pixel sensor and photonic device design, and lithography.