By Vivien Marx

Convey Computer, a provider of "hybrid-core" computing systems that integrate multi-core x86 processors with field-programmable gate arrays, is targeting the life sciences as one of the first markets for its HC-1 system.

The company, based in Richardson, Texas, said this week that Lawrence Berkeley National Laboratory, Stanford University's Center of Computational Earth and Environmental Science, and Oak Ridge National Laboratory will use the HC-1 for projects in several fields, such as climate research, seismic processing, and -omics analysis.

Those institutes join the first customer for the system, the University of California, San Diego, which received a beta version of the HC-1 during the summer with the goal of accelerating the performance of the InsPecT/MS-Alignment proteomics software package.

The firm, which was founded in 2006, began shipping its HC-1 hybrid-core system last month, Bruce Toal, president and CEO of the firm, told BioInform this week.

The HC-1 system is, according to the company, an integration of "patent-pending computer architecture and compiler technology" with off-the-shelf hardware: namely Intel Xeon processors and field programmable gate arrays by Xilinx, which is an investor in the company.

The company claims that the system combines the low cost and programming ease of x86 processors with the performance gains of application-specific hardware. Toal said that the HC-1 offers tighter integration than other systems that marry FPGA cards with commodity processors, as well as better "productivity of the software development environment."

The entry price for the HC-1 is "around $32,000," Toal said. The basic system includes a dual-socket computer with one Intel Xeon 2.13 GHz dual-core processor and a "coprocessor" based on the Xilinx Virtex 5 FPGA.

"Life sciences is a very interesting area for us," he said. Second-generation sequencing and genomics overall "will expand the universe of people needing to analyze genes and do it cost-effectively with a lot more data."

The company has one staff member who devotes his time solely to applications in the life sciences, Toal said.

While LBL, Stanford, and ORNL are among the first large institutes to publicly say they are using HC-1, Toal said that the company has other customers "who are not so anxious to go public."

The hybrid-core approach puts application-specific hardware to work on compute-intensive algorithms, achieving better performance than commodity processors by 5-fold to 25-fold, according to company documents.

At the Supercomputing 2009 conference this week, FPGA software firm Mitrionics demonstrated an accelerated Smith-Waterman algorithm run on HC-1 as part of a new partnership between the two firms.

The accelerated algorithm, written with the Mitrion C programming language, ran at 64 giga cell updates per second, which is more than 10 times faster than the non-accelerated Smith-Waterman on the same system, Robert Wall, Mitrionics' director of sales, told BioInform via e-mail.

He noted that this is still "an early Smith-Waterman demo, and not a fully optimized or production-ready version."

Wall believes Convey's HC-1 will help users "dramatically increase processing performance simply by recompiling their software to make use of the acceleration capabilities of the system."