Visit Microway at the High Performance Computing & Communications Conference in Newport, RI. March 24-26, 2009.

Higher Density + Higher Efficiency =
Highest Performance
HPC Times  February-March 2009

Intel "Nehalem" Xeon Processors = 40% power savings
2U Twin2 Server Achieve Higher Performance using Less Power with the next-generation Intel quad-core "Nehalem" Xeon 8P Twin2 server. This 2U chassis incorporates a 1200W redundant hot-swap power supply and 4 cold-swap servers, each with two Intel Xeon 5500 CPUs, up to 96 GB DRAM per server, three SAS/SATA hot-swap hard drives. Choose gigabit ethernet plus QDR or DDR InfiniBand interconnect. Use this platform for your next high-efficiency cluster.

Liquid Rack Cooling = 40% energy savings
Liebert XDK-W Rack February's 2009 IEEE Spectrum Magazine cover story features high efficiency, large scale data centers being developed by Google, Microsoft and others.

The article reviews how these leaders are optimizing PUE (Power Utilization Efficiency, 1.0 is perfect) by eliminating raised floors, air ducts and separate air conditioners. Their solution employs efficient rack cooling to achieve higher densities and higher efficiency. But you don't have to order tens of thousands of servers to match their density and efficiency. Microway offers the same density and efficiency and a big overall cost savings with up to 80 high density compute servers with just ONE 40U Liebert® XDK-W Rack.

How much of your annual budget is spent cooling and moving air? How much is spent on rent? The big guys are headed from a PUE of 2.0 to a PUE of 1.2. You can, too!
Call us now for more information.

AMD "Shanghai" Opteron Processors/Memory =
30-50% shorter run-times

Opteron Memory Speeds A Microway compute cluster with quad-core AMD "Shanghai" Opteron CPUs can slash run times as much as 50% compared to previous generation Opterons. Memory bound applications will benefit from Shanghai's 10 GB/sec bandwidth per CPU using DDR2-800 memory. Up to 32GB DDR2-800 memory may be connected to each CPU.

AMD is offering an Opteron socket upgrade program. Call us to see if your cluster can benefit.

NVIDIA® Tesla GPU Computing = up to 90% speedup

Different GPU applications require specific hardware configurations. Microway supports them all:

Next-Generation NumberSmasher® InfiniBand GPU Cluster design achieves a density of 2 Tesla GPUs and 2 CPUs per 1U. Each high-density 24U rack contains 36 Quad-Core CPUs, 36 Tesla 1 TFLOP GPUs and 40 Gbit/sec QDR ConnectX InfiniBand.

Custom configurations with any number of CPUs and GPUs are available to fit your budget and application.

Microway GPU Cluster
NumberSmasher GPU Cluster

New Designed for low-cost GPU supercomputing, Microway's 1U NumberSmasher GPU node offers one or two quad-core CPUs and one GPU with gigabit ethernet and optional InfiniBand connectivity.

Microway 1U GPU Node
1U GPU Node

Microway's Fully-integrated WhisperStation-PSC is available with one or two quad-core CPUs and one to four Tesla GPUs, for a Desktop Personal Supercomputer. Plus your choice of Linux, Windows or Dual-Boot.

Learn more about Microway's Tesla solutions.


Seamless GPU Computing with MATLAB®
Jacket Middleware Marries MATLAB® and CUDA.
Accelereyes Microway now offers 'Jacket' software for direct control of GPUs from MATLAB without CUDA programming. In fact, no code changes are necessary to achieve more than 10x speedup in most applications - just substitute 'gsingle' for 'single' float data definitions and when functions access these new variables, Jacket automatically performs GPU computations. Jacket also includes an OpenGL visualization library for MATLAB.
Sign up for a 15 day free trial.

For more information, please contact:
Eliot Eshelman at 508-732-5534