Tuesday, November 30, 2010

Interview: AMAX Showcases GPU Solutions & Personal Supercomputers at SC10

In this podcast, I discuss the latest HPC developments at AMAX with Matt Thauberger, the company's Corporate Technology Advisor, and Julia Shih, Strategic Alliance Manager. Download the MP3 or Subscribe on iTunes.

AMAX made a number of announcements around SC10 including:

  • ServMax Xn-5801G High-Density GPGPU Cluster Solution is a 1792-GPU core server with 8 x86 sockets in a 5U enclosure.

  • StorMax Ultra-Dense JBOD Expansion Storage with dual SAS 2.0 module fill-over design and 4 SAS 6Gb/sec ports. Features dual redundant power supply and is an ideal building block to scale out NAS or scale up NAS.

  • ServMax GPGPU Personal Supercomputer is powered by the massively parallel CUDA architecture and supports four NVIDIA Tesla C2050 or C2070 GPUs transforming a workstation into the equivalent of a small cluster at 5% the power consumption and 10% the cost. It delivers an astounding 4.12 Teraflops of performance in a single chassis.

  • ServMax GPGPU Supercomputing Server with 2.06 Teraflops of computing horsepower is the fastest 1U server available. Powered by the massively parallel CUDA architecture and Tesla GPU computing processors, it can seamlessly fit into existing HPC environments and is ideal for financial services applications.

Sunday, November 28, 2010

Interview: Allinea Software Gears up for Exascale

In this podcast from SC10, we discuss recent developments from Allinea Software with Mike Fish, CEO, and Jacques Philouze, the company's VP of marketing. Download the MP3 or Subscribe on iTunes.

Allinea had a number of announcements at SC10 including:

  • Release of Allinea DDT version 2.6.1. Supports NVIDIA Toolkits 3.1 and 3.2. This latest version adds support for multiple GPU devices within the same node, in addition to the existing capabilities for debugging clusters of GPU systems. It also supports the latest hardware, with some of the newer features such as device function call and stepping into and out of nested functions on the Fermi hardware now possible.

  • An expanded collaboration with Oak Ridge National Laboratory, who have selected Allinea DDT as the debugger of choice for Hybrid GPU software - where few industry-strength tools exist.

Wednesday, November 17, 2010

Interview: LSI Rolls Out High-Density Storage

In this podcast, LSI's Steve Hochberg describes the company's newly announced high-density Engenio 2600-HD storage system, which consists of two LSITM 6Gb/s SAS-based controllers integrated into the new Engenio DE6600 high-density SAS drive enclosure. The system is capable of sustaining up to 4GB/s of throughput and housing up to 60 SAS drives in a 4U space.

The Lustre and General Parallel File System implementations found in today’s HPC market demand scalability in throughput and capacity without sacrificing data access," said Addison Snell, CEO, Intersect360 Research. "At the same time, HPC organizations need systems that can maximize the available data center floor space. The Engenio 2600-HD is designed for high performance and scalability in a dense architecture and is well positioned to meet the needs of a wide range of HPC customers."

You can check out LSI technologies at SC10 booth #2946. Download the MP3 or Subscribe on iTunes.

Monday, November 15, 2010

Interview: HP's Marc Hamilton on Record GPU Efficiency at SC10

In this podcast, Marc Hamilton, HP's Vice President of HPC gives us a summary of the company's announcements and TOP500 results, including the #4 Tsubame 2 system at Tokyo Tech. Download the MP3 or Subscribe on iTunes.

Sunday, November 14, 2010

Interview: Supermicro Fires Up Many-Core Blades at SC10


In this podcast, I interview Tau Leng, GM of Supermicro. Of note is Supermicro's announcement of a blade enclosure that has up to 2,880 x86 cores in a single 42U rack.

"When powered by new 12-core AMD Opteron 6100 Series (Socket G34) processors, the TwinBlade enables up to 2,880 processing cores, 240 SATA drives and 15TB of DDR3 memory per 42U rack. Featuring dual 40Gb/s InfiniBand, FCoE or 10GbE switches and dual 1/10GbE switches in one 7U enclosure, TwinBlade provides the highest performing I/O throughput and scalability in the industry and is a superb solution for high-performance computing (HPC), datacenter, enterprise and cloud computing deployments. With successful installations at CERN, PRACE, and other Fortune 100 organizations, the SuperBlade® has already become the preferred solution for a wide range of mission-critical applications."


The company has a number of other announcements this week around GPUs and scalable storage, so be sure to stop by their booth this week at booth #3829. Download the MP3 or Subscribe on iTunes.

Friday, November 12, 2010

Interview: Xyratex Ramps Up for Full-Scale Lustre Initiative

In this podcast, Mike Stolz and Steve Paulhus from Xyratex discuss their recent acquisition of ClusterStor and how the merged company plans to serve the Lustre community.

Xyratex has been around for a long time as an OEM company, and their storage products are re-labeled by nearly all the big-name HPC vendors out there.

Download the MP3 or Subscribe on iTunes.

Wednesday, November 10, 2010

Interview: SC10 Session for Women n HPC

In this podcast, I talk with the Oak Ridge team behind the SC10 Birds of a Feather session: Developing, Recruiting, and Retaining in Women in HPC. The idea for the session came from ORNL's Rebecca Hartman-Baker, and she recruited Judy Hill and Hai-Ah Nam to help bring it all together. As part of the discussion, each team member explains how they got into HPC and the challenges they faced along the way.

I think this is going to be one session not to miss at SC10. Women are still very much a minority in the HPC community and I think this kind of program is important. Things may be starting to change though. Anne Baker at ORNL has started up a Facebook page for women in HPC and projects like DotDiva.org are in place for getting young girls interested in computing. Download the Podcast MP3 or Subscribe on iTunes.

Tuesday, November 2, 2010

Interview: Pervasive Datarush Sets TeraCUPS Record on Big Data


According to their web site, Pervasive DataRush has a simple mission: Simplify How You Process and Analyze Big Data. The company's product is a parallel dataflow platform that eliminates performance bottlenecks in big data preparation and analytics.

Pervasive Datarush set a remarkable performance record recently on an SGI Altix system with 384 cores. These types of benchmarks are measured in CUPS (cell updates per second) and the company was able to achieve nearly one TeraCUPS of performance on the Smith-Waterman algorithm, which is popular in bioinformatics.

In this podcast, I catch up with Davin Potts to talk about how Pervasive Datarush leverages the parallel processing capabilities of multicore processors and SMP systems to deliver extreme performance on big data. Download the podcast or subscribe on iTunes.