SC11 generated some exciting technologies to watch for in the coming year. This was an eventful show, so we've put together some key highlights:

• Directives-based programming for NVIDIA Tesla GPUs and OpenACC 
  GPU acceleration is no longer limited to specialized codes! New directives-based programming models allow for easy acceleration of existing codes. Direct the compiler with flags to paralellizable sections of code for immediate speedups without CUDA. NVIDIA, PGI, and others have established a new MPI-like standard, OpenACC, to help programmers leverage directives-based programming. Expect more modest speedups than native CUDA coding, but with very little coding effort.
  Press release and standard information at: http://www.openacc-standard.org/
• 2X in 4 Weeks, Guaranteed 
  NVIDIA and PGI announced this new program. Using PGI's directives-based Accelerator compilers, programmers are guaranteed a 2X speedup in 4 weeks or less. 
  Still not faster? PGI and NVIDIA will provide hands-on assistance to get you there. Microway is pleased to offer its customers access to this program. 
  See this post: http://blogs.nvidia.com/2011/11/gpu-acceleration-made-easy-get-2x-in-4-weeks/
• Simcluster Program
  Tesla GPU-based fixed configurations, Simclusters, are now available. These include the latest Tesla M2090 GPUs, Intel CPUs, and InfiniBand in optimized hardware configurations. Furthermore, these cluster configurations in 4, 8, or 16 nodes arrive with full Microway software integration. Pick the appropriately sized configuration for your needs, or work with Microway sales staff on a custom cluster.
  Email sales@microway.com for a quote.
• AMD Opteron 6200 Series CPUs 
  Opteron 6200 series CPUs made quite a splash at SC11. Many leading supercomputing installations, including the massive Jaguar cluster at Oak Ridge National Lab, announced upgrades to the 16 core Opteron 6200 series platform. ORNL, AMD, and Cray announced Jaguar will be reworked with the latest Opterons and with GPUs to become Titan, a new leader on the Top500 list. 
  Learn more about how you can scale your performance at http://www.microway.com/opteron/
• Future Xeon E5 Series CPUs 
  Intel's Xeon CPU products offer big performance gains in 2012. With 8 cores per CPU and PCI-E 3.0 support, Xeon E5 will provide stunning performance for HPC applications. Additional details and benchmarks are available today under NDA. Microway's CPU + GPU accelerated 3D VMD demo ran on a Xeon E5 + NVIDIA Tesla platform.
  Please contact Microway sales for more details.
• Intel MIC x86 Accelerators 
  Scattered around the show floor was information on Intel's upcoming MIC coprocessor product. Intel also delivered a press demo of the chip running applications at 1 TFLOP/sec (double precision). Intel's SC11 press briefing and release claims the production chip will have greater than 50 cores. SC11 demos were run off a PCI-E connected card.
  HPCWire covered the press demo, and their article can be found here: http://bit.ly/tlhsMJ
• Bright Computing Cloud Bursting
  Bright Computing demonstrated their cloud capabilities. Using Bright Cluster Manager, you can burst into public clouds, such as Amazon EC2, with only a few mouse clicks, without the need for expert knowledge of Linux or cloud computing.
  More about this feature: http://www.brightcomputing.com/Linux-Cluster-Cloud-Bursting.php

Here's a preview of new Microway hardware platforms available to our customers:

• Navion QuadPuter 2U GPU Server
  Microway's new Navion QuadPuter 2U GPU Server provides 4 CPUs + 2 GPUs in a single Opteron server. Expect up to 64 x86 cores from the new AMD Opteron 6200 series CPUs and 2 GPUs with TFLOPS of floating point performance. This platform is ideal for a user who wants as many CPU and GPU cores as possible in a single system.
  For more information visit our QuadPuter Landing Page

• NumberSmasher 2U GPU Servers with 4 GPUs
  NumberSmasher 2U GPU provides 4 PCI-E x16 2.0 slots for up to 4 Tesla M-series GPUs in 2U. This robust platform also includes 2 Xeon CPUs, 80 PLUS Platinum redundant power supplies, 10x2.5" hot-swap drive bays.
  Contact us for a quote on this brand-new platform sales@microway.com
• NumberSmasher 1U GPU Servers with 3 GPUs
  This new platform adds a third Tesla M-series GPU to our 1U GPU server platform. Up to 2 Xeon CPUs, 96GB DDR3 memory, and 4x 2.5" drive bays, and 80 PLUS Platinum redundant power supplies provide an excellent single server or cluster building-block.
  Contact us for a quote on this brand-new platform sales@microway.com

Enhance your knowledge on the latest HPC technologies in time for 2012. Microway's tech team has collected the information for you on these new technologies:

• LRDIMM Memory Technology
  LRDIMMs (Load Reduced DIMMs) have been on our radar for some time. These memory modules add a memory buffer between the memory controller and the chip ranking of DDR3 RDIMMs, hiding the ranks from the system. LRDIMMs therefore provide two clear benefits to users:
  
  1. Higher capacity DIMMs with full memory performance (currently 32GB DDR3)
  2. Faster memory performance with complete DIMM socket population (often full 1333Mhz bandwidth even when fully populated)

  For more information on LRDIMMs, how they achieve this performance, and other advantages, we encourage you read this detailed whitepaper from Inphi. Inphi provides the buffer chips for Samsung, ELPIDA, and Micron LRDIMMs.

• AMD Opteron 6200 Series ("Bulldozer" architecture) Compiler Guide
  New AMD Opteron 6200 series CPUs have a different internal chip architecture from previous Opteron CPUs. The latest versions of popular compilers are best optimized for the new "Bulldozer" architecture. Use the Compiler Quick Reference Guide as a cheat sheet for the proper flags for your compiler. This will ensure optimal performance for your code on the new CPUs.

Microway's team has collected a few notable Open Source software developments from the past number of weeks:

• MVAPICH2 1.8a1 Released
  "The MVAPICH2 1.8a1 release is targeted for MVAPICH2 users to harness performance on InfiniBand (Mellanox) clusters with NVIDIA GPU adapters and CUDA support." For more information, see the release notes
• Ganglia GPU Monitoring
  Ganglia GPU monitoring is now available. For information on the packages, please visit the NVIDIA Developer Zone Page on the module.
• PyCUDA
  Python programmers can now access CUDA with the PyCUDA project. This project is ongoing, but it was prominently featured at the 2010 GPU Technology Conference. More about PyCUDA can be found on the PyCUDA homepage

Abstract: Most users know how to check the status of their CPUs, see how much memory is free or find out how much disk space is free. In contrast, keeping tabs on the health and status of GPUs has historically been more difficult. If you don’t know where to look, it can even be difficult to determine the type and capabilities of the GPUs in a system.

Thankfully, NVIDIA’s latest hardware and software tools have made good improvements in this respect. We explore the functions of nvidia-smi to learn more.

October and November flooding in Thailand has severely impacted hard drive production. Hard drive prices have doubled over the last month and we expect this industry-wide shortage to continue into at least the next quarter. Microway has been carefully monitoring the situation and working with our drive suppliers to keep our orders flowing. We've secured inventory for our orders, even at higher costs.

We endeavor to hold down the prices to you, the end-user, as much as possible. Also, our sales team will suggest alternate drives or SSDs if substitutions are appropriate to keep your costs consistent quote-to-quote. Still, there are situations where we must pass a portion of the cost differences in revised quotations. Our aim is to make this the least painful to you and the most transparent possible.

We thank you for your understanding throughout this drive supply crisis. Learn more from Seagate on this topic in their latest outlook. 

Microway is growing, and we're looking to hire for two positions. Click through to the full listings for more information:

Sales Engineer - Help HPC work better for scientists and researchers worldwide!  Well spoken techie will advise our clients on design and implementation of GPU and cluster solutions for their specific needs.

Integration/Tech Support - Integrate, test, and assist supporting the latest HPC clusters and WhisperStations  - work with bleeding edge technologies!  Our systems incorporate Intel and AMD processors, NVIDIA GPUs, storage, InfiniBand, graphics and Linux software. Knowledge of server hardware, parallel processing, parallel programming (use of MPI, OpenMP) and computer architecture is highly desirable.