Visit additional Tabor Communication Publications
June 08, 2007
Google + Stream Computing = ?
As we reported on Wednesday morning, PeakStream Inc. has been bought by Google under undisclosed terms. Now InforWorld is reporting an inside source at Google has confirmed that PeakStream's product line will no longer be commercially available.
PeakStream, a company that has developed a stream computing platform for multicore architectures, has been featured in this publication several times since it introduced its product in September 2006. Their platform is designed to allow programmers to develop applications for various multicore architectures (x86, GPUs, Cell) in a hardware-independent manner. The target applications are all in technical computing, including energy exploration and modeling, financial analytics, and image processing/signal processing.
In January, co-founder and chief technology officer Matthew Papakipos wrote an article for HPCwire about the convergence of GPU and CPU architectures. In it Papakipos outlined his belief that CPUs and GPUs are morphing toward a Cell processor-like model, which encapsulates some of the functionality of both architectures. He also suggested that the stream programming model would be the path of least resistance from serial programming to the world of multicore software.
So what does this have to do with Google and its web-based universe? Google has not provided any specific rationale for the acquisition. In a cryptic statement, first reported by the Register, Google stated:
"We believe the PeakStream team's broad technical expertise can help build products and features that will benefit our users. We look forward to providing them with additional resources as they continue developing high performance applications for modern multi-core systems."
While that's not much help, I think it's essentially true. Google may be planning to direct the PeakStream technology into some future projects it's cooking up. This might entail the development of sophisticated image, video and audio search engines. The acquisition of Neven Vision last summer confirms Google's interest in this area. Currently, looking for objects on the Web relies on text keywords that are associated with the item. For example, a picture of a zebra is labeled with the word "zebra" in the image caption, surrounding text, or the image's hypertext metadata. In many cases, it would be more useful to do searches based on human-like pattern recognition rather than text. With this capability, a scene of an African savanna with a herd of zebra would be recognized as a picture of zebras, even if the image text labels were "Africa" and "savanna." The same applies to audio and video recognition.
The reason true multimedia searching is not available today is the lack of sophistication in the recognition algorithms and the lack computational resources (imagine thousands of Google users simultaneously pattern matching on Paris Hilton photos). As most computing geeks know, pattern recognition is much more compute-intensive than text searching. But Moore's Law is conveniently providing just the kind of performance required by such workloads -- multicore performance. And the stream computing technology offered by PeakStream seems well-suited to develop these types of pattern recognition software. Now Google just needs to build some GPU server farms. Hmm... there's a nice big Google datacenter on the Columbia River being built. I wonder what's going on inside?
When AMD Slips, Cray Falls
When you build supercomputers for a living, profitability can be an ethereal experience. The folks at Cray Inc., the supercomputing icon, know this better than anyone. This week Cray lowered its revenue outlook for 2007, predicting that it will probably not post a profit this year. The company is anticipating that AMD's quad-core Opteron processors, which Cray is planning to use in its XT4 deployments, will not be available for volume shipping until late in the fourth quarter of the year -- too late to book the revenue for 2007.
More specifically, it's AMD's quad-core Budapest chips that will be delayed. These are the ones Cray will use in their XT4 line. The quad-core Barcelona chips are still expected to ship in the third quarter of the year. In fact, servers using Barcelona processors were previewed at the Computex conference in Taipei just this past week. AMD has not given any indication on the nature or timing of the delay for the Budapest processors.
The Barcelona chips are designed for 2-, 4-, and 8-way server boxes, while Budapest is designed for single-way configurations. Since Cray systems use a massively parallel processing (MPP) architecture, the single-way Budapest is the processor of choice for them. The Barcelona chip, which is destined for vanilla servers, represents AMD's mainstream line of quad-core Opterons, and presumably the ones they are most interested in delivering as soon as possible. As much as AMD would love to accommodate Cray, commodity chipmakers are in the volume business.
Cray expressed "disappointment" with the development, but one can imagine their reaction ran a little deeper than that. Cray's commitment to Opteron technology for the remainder of the decade means that the company has hitched its wagon to AMD and their ability to execute. The supercomputer maker's dependency on AMD has become more risky of late as Intel's resurgence has caused problems for AMD.
Over the past six months, Intel has been stealing market share away AMD. According to research firm iSuppli, as of the first quarter of 2007, Intel had 80.2 percent of the microprocessor market, up 4.5 percentage points from 75.7 percent in the previous quarter. At the same time, AMD's share of the microprocessor market slid from 15.7 to 11.1 percent.
While users wait for Barcelona, Intel's quad-core Clovertown processor has been shipping for six months. AMD is quick to point out that Intel's version is not a "true" quad-core, relying on a dual-core, dual-module setup to get four cores into a single socket. The quad-core Opterons are also being advertised as having a 50 percent advantage in floating point performance and 20 percent in integer performance over the Intel's quad-core processors at the same frequency. But while Intel's quads have been in the field crunching on real applications, AMD is relegated to talking about its offering in the future tense.
And as I've noted before, any performance advantages enjoyed by the quad-core Opterons are likely to be short-lived. Intel is planning to deliver its first 45nm Penryn processors late this year, which should result in a significant performance boost compared to Intel's current 65nm offerings. If AMD gets in a mode of continually having to play catch-up with its rival, its competitiveness in the x86 market will continue to erode.
So what's a company like Cray supposed to do? Unlike mainstream server makers, who can build generic boxes based on either Intel or AMD x86 chips, for now Cray is stuck with AMD. The XT3 and XT4 supercomputer lines are designed around AMD's HyperTransport interconnect, associated Torrenza socket standard (which allows for co-processor accelerator plug-ins), and on-chip memory management. These features give the Opteron the kind of scalability needed for MPP architectures like Cray's. When Intel gets its CSI interconnect and on-chip memory controllers established on its x86 chips, maybe Cray will take a fresh look at AMD's nemesis.
As always, comments about HPCwire are welcomed and encouraged. Write to me, Michael Feldman, at firstname.lastname@example.org.
Posted by Michael Feldman - June 07, 2007 @ 9:00 PM, Pacific Daylight Time
Michael Feldman is the editor of HPCwire.
No Recent Blog Comments
The Xeon Phi coprocessor might be the new kid on the high performance block, but out of all first-rate kickers of the Intel tires, the Texas Advanced Computing Center (TACC) got the first real jab with its new top ten Stampede system.We talk with the center's Karl Schultz about the challenges of programming for Phi--but more specifically, the optimization...
Although Horst Simon was named Deputy Director of Lawrence Berkeley National Laboratory, he maintains his strong ties to the scientific computing community as an editor of the TOP500 list and as an invited speaker at conferences.
Supercomputing veteran, Bo Ewald, has been neck-deep in bleeding edge system development since his twelve-year stint at Cray Research back in the mid-1980s, which was followed by his tenure at large organizations like SGI and startups, including Scale Eight Corporation and Linux Networx. He has put his weight behind quantum company....
May 16, 2013 |
When it comes to cloud, long distances mean unacceptably high latencies. Researchers from the University of Bonn in Germany examined those latency issues of doing CFD modeling in the cloud by utilizing a common CFD and its utilization in HPC instance types including both CPU and GPU cores of Amazon EC2.
May 15, 2013 |
Supercomputers at the Department of Energy’s National Energy Research Scientific Computing Center (NERSC) have worked on important computational problems such as collapse of the atomic state, the optimization of chemical catalysts, and now modeling popping bubbles.
May 10, 2013 |
Program provides cash awards up to $10,000 for the best open-source end-user applications deployed on 100G network.
May 09, 2013 |
The Japanese government has revealed its plans to best its previous K Computer efforts with what they hope will be the first exascale system...
May 08, 2013 |
For engineers looking to leverage high-performance computing, the accessibility of a cloud-based approach is a powerful draw, but there are costs that may not be readily apparent.
05/10/2013 | Cleversafe, Cray, DDN, NetApp, & Panasas | From Wall Street to Hollywood, drug discovery to homeland security, companies and organizations of all sizes and stripes are coming face to face with the challenges – and opportunities – afforded by Big Data. Before anyone can utilize these extraordinary data repositories, however, they must first harness and manage their data stores, and do so utilizing technologies that underscore affordability, security, and scalability.
04/15/2013 | Bull | “50% of HPC users say their largest jobs scale to 120 cores or less.” How about yours? Are your codes ready to take advantage of today’s and tomorrow’s ultra-parallel HPC systems? Download this White Paper by Analysts Intersect360 Research to see what Bull and Intel’s Center for Excellence in Parallel Programming can do for your codes.
In this demonstration of SGI DMF ZeroWatt disk solution, Dr. Eng Lim Goh, SGI CTO, discusses a function of SGI DMF software to reduce costs and power consumption in an exascale (Big Data) storage datacenter.
The Cray CS300-AC cluster supercomputer offers energy efficient, air-cooled design based on modular, industry-standard platforms featuring the latest processor and network technologies and a wide range of datacenter cooling requirements.