Startup Aims To Harness Specialized Chips To Make Supercomputing Affordable - InformationWeek

InformationWeek is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them.Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.


Startup Aims To Harness Specialized Chips To Make Supercomputing Affordable

PeakStream is working on using graphics chips and other specialized processors to boost performance for scientific and technical applications up to 20-fold.

PeakStream, a Silicon Valley startup, this week released software to boost performance for scientific and financial software up to 20-fold using graphics chips and other specialized processors.

The technology is designed to feed the appetite for supercomputing technology in industries including oil and gas exploration, investment banking, manufacturing, and pharmaceuticals.

PeakStream, a 35-person company founded by Pat Hanrahan, a Stanford University computer science professor, and Matthew Papakipos, a former technical executive at Nvidia, is backed by $17 million in venture capital from Kleiner Perkins Caufield & Byers, Sequoia Capital, and Foundation Capital. This week, the company released an evaluation version of its software, which lets standard C and C++ programs enlist the aid of graphics chips to execute complex mathematical functions at blazing speeds. Specifically, PeakStream's development tools and run-time environment can execute portions of high-performance computing applications on graphics chips from FireGL, Radeon, and ATI Technologies (which is in the process of being acquired by Advanced Micro Devices in a $5.4 billion deal slated to close by year's end).

The idea is to let customers, which include Hess and Science Applications International Corp., take advantage of specialized silicon without slowing down their software development work, says PeakStream CEO Neil Knox.

"There's a tremendous need for more and more computing power," says Knox, a former executive at Sun Microsystems, and a veteran of Data General and Burroughs.

Eventually, the company plans to release versions of its software for IBM's Cell chip, as well as for CPUs from Intel and AMD that incorporate multiple processing cores, which makes programming more difficult.

The $9 billion technical computing market is growing at 20% annually as companies apply supercomputing power to scientific and engineering apps to speed products to market and gain new insights into data. A massive oil strike by Chevron and others in the Gulf of Mexico this month was enabled in part by computerized seismic imaging technology. Auto and aerospace manufacturers including Boeing, Ford Motor, and General Motors are using supercomputers for product design and simulated testing. To capitalize on the growth, Microsoft is entering the high-performance computing market with a specially tuned version of Windows and development tools for scientists.

At the same time, interest in using GPUs, field programmable gate arrays, and other specialized chips to speed up software performance is increasing as the gains designers can realize from increasing chips' clock speeds has diminished. Teams at IBM, Cray, Stanford, and the Tokyo Institute of Technology have been experimenting with programming graphics processors and other specialized chips to speed up the performance of supercomputing applications. And IBM this month said it will build a supercomputer capable of 1,000 trillion computations per second—three times as fast than any current system—using 16,000 of its Cell processors, originally designed for Sony's PlayStation video game.

PeakStream will initially target companies in oil and gas exploration, defense applications such as signal analysis, options pricing applications in finance, and academic research. It won't target auto and aerospace engineering markets that are served by well-established independent software vendors.

PeakStream's technology can hide the complexity of GPUs and multicore chips, says chief technology officer Papakipos. Programmers use a PeakStream API that lets C and C++ apps call a virtual machine, which lives in a run-time environment installed on each machine in a high-performance computing cluster. The API specifies what math functions to execute, and the virtual machine calls a library of pre-compiled numerical software. PeakStream's compiler then schedules those jobs to the ATI chips.

The first version of PeakStream's software will run on Linux, and the company plans a version for Windows, including a plug-in for programming in the C# language with Microsoft's Visual Studio development tools. Papakipos says PeakStream is also working on a version of its software for Fortran.

We welcome your comments on this topic on our social media channels, or [contact us directly] with questions about the site.
Comment  | 
Print  | 
More Insights
InformationWeek Is Getting an Upgrade!

Find out more about our plans to improve the look, functionality, and performance of the InformationWeek site in the coming months.

Remote Work Tops SF, NYC for Most High-Paying Job Openings
Jessica Davis, Senior Editor, Enterprise Apps,  7/20/2021
Blockchain Gets Real Across Industries
Lisa Morgan, Freelance Writer,  7/22/2021
Seeking a Competitive Edge vs. Chasing Savings in the Cloud
Joao-Pierre S. Ruth, Senior Writer,  7/19/2021
White Papers
Register for InformationWeek Newsletters
2021 State of ITOps and SecOps Report
2021 State of ITOps and SecOps Report
This new report from InformationWeek explores what we've learned over the past year, critical trends around ITOps and SecOps, and where leaders are focusing their time and efforts to support a growing digital economy. Download it today!
Current Issue
Monitoring Critical Cloud Workloads Report
In this report, our experts will discuss how to advance your ability to monitor critical workloads as they move about the various cloud platforms in your company.
Flash Poll