BDTI’s DSP Insider Archives
|
||
BDTI’s DSP Insider Archives |
|
|
| HOME << |
|
|
|
||
This month:
ParthusCeva Rolls Out New Name, New CoreLast month ParthusCeva, Inc. announced it would change its name to CEVA, Inc. Along with the name change, the company is shifting its focus to providing signal-processing application solutions based on its DSP cores. These moves come as part of a series of changes for the company, which was formed about a year ago by the merger of Parthus Technologies with the DSP core licensing division of DSP Group. Since the merger, the company has moved its headquarters from Dublin, Ireland to San Jose, California, and has hired a new chief executive officer, a new vice president of business development, and a new chief financial officer. The organizational changes have been accompanied by changes in the company’s lineup of DSP cores. Today, CEVA announced availability of its newest DSP core, the CEVA-X1620. (The CEVA-X1620 was previously announced as the Cedar1620—see the September 2002 DSP Insider for details.) Most of the DSP cores inherited by the company trace their lineage to the early days of DSP Group’s licensing business. Several of these cores are well established and supported by a mature development infrastructure. In contrast, the CEVA-X1620 is an all-new design, and the quality of its development infrastructure is not yet known. BDTI recently completed an analysis of the CEVA-X1620. Based on its projected BDTIsimMark2000™ score of 3620 at 450 MHz in a 0.13-micron process, the CEVA-X1620 is the fastest licensable DSP core benchmarked by BDTI to date. In comparison, the next-fastest DSP core benchmarked by BDTI is the StarCore SC1400, which has a projected BDTIsimMark2000 score of 3420 at 305 MHz in a 0.13-micron process. The CEVA-X1620 is faster than its competitors partly because it has a high projected clock rate. For example, the CEVA-X1620 projected clock rate is nearly 50% higher than that of the SC1400. The CEVA-X1620 is also projected to be faster than competing off-the-shelf DSPs. For example, BDTI’s analysis shows that the CEVA-X1620 is projected to be slightly faster than the 600 MHz Analog Devices ’BF53x (Blackfin) and about 2.5 times faster than the 300 MHz Texas Instruments ’C55x. (Benchmark scores for these and other processors are available at http://www.BDTI.com/bdtimark/BDTImark2000.htm.) Of course, speed is only one of many factors potential CEVA-X1620 licensees will consider. In particular, the CEVA-X1620 targets portable applications where energy efficiency is a paramount concern. BDTI has not yet evaluated the energy efficiency of the CEVA-X1620, so its standing on this metric remains unknown. Potential licensees are also likely to rank development infrastructure as a top consideration, and as mentioned earlier, the quality of the CEVA-X1620 development infrastructure is also unknown. The success of the CEVA-X1620 will likely depend on CEVA’s ability to deliver leading energy efficiency and development infrastructure as well as competitive processing speed.
The CEVA-X1620 is available now. For licensing information, contact
CEVA, whose new web site is http://www.ceva-dsp.com.
Analog Devices Announces New SHARCs for AudioToday Analog Devices announced four new members of its SHARC family. These new chips primarily target audio applications, and have a variety of audio-specific peripherals and I/O ports. The new chips include the ADSP-21266, ’21267, ’21364, and ’21365. The ’21266 and ’21267 primarily target car audio systems. The ’21266 is currently sampling at 150 MHz, and the ’21267 will begin sampling at 200 MHz in the first quarter of 2004. The ’21364 and ’21365 target high-performance car and professional audio systems, and will begin sampling in the first (’21365) and second (’21364) quarters of 2004. The ’21364 and ’21365 employ a slightly deeper pipeline than other family members, yielding a higher top clock speed of 300 MHz. The ’2136x chips are assembly-code compatible (but not binary compatible) with the other SHARC family members. Pricing for the four new parts will range from $9.95 for the ’21267 to $24.95 for the ’21364 in 10,000-unit quantities. This announcement continues Analog Devices’ reinvigoration of its floating-point lineup. Just last August the company announced 200 MHz members of the SHARC family (see DSP Insider, October 2003), which doubled the family’s performance compared to earlier chips. With the jump to 300 MHz, the new ADSP-213xx chips will be even faster—though the deeper pipeline means that they probably won’t get the full 50% speedup suggested by the increased clock rate, and users may need to re-optimize code written for other SHARC family members to achieve maximum performance. At $9.95, the ADSP-21267 also sets a new low price point for the family.
The new chips will primarily compete with TI’s TMS320C67x 32-bit
floating-point family, and with Motorola’s audio-oriented 24-bit
fixed-point architecture, the DSP563xx. Compared to Motorola’s chips,
the new SHARC chips have several advantages. The SHARC’s 32-bit
floating-point data format offers better signal fidelity and easier
programming. In addition, BDTI’s analysis shows that the new SHARC
family members are considerably faster than the DSP563xx. But the
DSP563xx is available in less-expensive versions (down to $4.50 at 100
MHz), and it is more energy-efficient. Thus, the DSP563xx is likely
to win sockets where cost and energy efficiency are critical.
Compared to TI’s 225 MHz ’C67x, the 300 MHz SHARCs will be noticeably
faster.
BDTI Case Study
This Month: Reconfigurable Hardware for Signal ProcessingReconfigurable hardware encompasses a broad spectrum of architectures with run-time configurability. These range from FPGAs to instruction-set processors with reconfigurable data paths. In recent years, many companies have introduced new reconfigurable architectures targeting digital signal processing applications. These companies hope to combine the best attributes of ASICs (performance and efficiency) with the best attributes of traditional processors (ease of application development, flexibility to modify functionality in the field). In practice, however, these benefits are difficult to attain, and reconfigurable architectures can easily end up combining the worst attributes of other technologies—such as long, painful development cycles and poor cost-effectiveness. Vendors of reconfigurable hardware have increasingly emphasized the digital signal processing capabilities of their products. They have introduced new architectures, development tools, and IP building blocks targeting high-end signal processing applications. One reason for this emphasis is the daunting computational workload of some high-end signal processing applications. Reconfigurable hardware can enable system developers to attain high performance via highly parallel hardware structures that are tuned to the needs of key algorithms. This can be particularly effective for complex algorithms that do not map efficiently to traditional processors, and for performance-hungry applications with rapidly changing specifications. Comparing reconfigurable devices to traditional processors requires careful analysis because the underlying architectures and capabilities are so different. For example, reconfigurable devices often can perform an enormous number of operations in parallel, but it can be difficult to predict how much of this parallelism can actually be used in a given application. And mapping an application onto a reconfigurable architecture in a way that harnesses the chip’s full potential can be extremely challenging. BDTI has been active for several years in evaluating reconfigurable architectures for signal processing applications. In one typical consulting engagement BDTI evaluated a novel reconfigurable processor for use in telecom applications. BDTI compared the performance of this reconfigurable processor to that of a popular DSP processor using both the BDTI Benchmarks™ and custom benchmarks derived from the target application. And earlier this year, BDTI published the first independent analysis comparing high-end FPGAs to DSPs. This report, FPGAs for DSP, provides insightful qualitative and quantitative analysis—the latter based on innovative benchmarks that show the true application performance of the two types of chips.
Learn how BDTI’s reconfigurable architecture and digital signal
processing expertise can help you, or find out more about BDTI’s FPGAs
for DSP report, by contacting Jeremy Giddings at BDTI
(giddings@BDTI.com).
Impulse Response, by Jeff Bier
Floating-Point DSPs Catch the Audio WaveAs we approach year-end, it seems fitting to look back at developments and trends in DSP processors. And it’s an opportune time to do so, since my colleagues and I just completed the latest iteration of our exhaustive (and exhausting) study of leading DSPs. It has been three years since we published the previous edition, and in reviewing the new version I am struck by one of the key changes in our industry. Back in 2000, telecom was the DSP killer app. Most DSP architectural innovations were happening in 16-bit fixed-point processors, the mainstays of telecom. At that time, we wondered about the future of floating-point DSPs. They were rarely used in telecom, and were coming under competitive pressure from CPUs in their traditional markets of military, medical, and imaging equipment. The two key floating-point architectures, Analog Devices’ SHARC and Texas Instruments’ ’C67x, seemed to be stagnating. But the telecom industry has had a rough couple of years. Although telecom is still a formidable force driving DSPs, it is starting to share the spotlight with a significant new market: consumer audio and video. This market includes DVD players, home theater equipment, digital cameras, MP3 players, and car audio. Programmable processors are an attractive choice for these applications, because they enable OEMs to differentiate their products with features like proprietary effects processing. Most portable products use fixed-point processors to minimize cost and power, but car and home audio systems often tolerate somewhat higher cost and power in exchange for better-sounding audio. This is where floating-point DSPs have found a new sweet spot. Floating-point DSPs have been used in professional audio products for years, primarily because their 32-bit floating-point data format offers better signal fidelity than is afforded by most fixed-point processors. The floating-point format also makes it easier to implement proprietary signal processing algorithms. These advantages make floating-point DSPs a natural choice for the burgeoning consumer audio market, thus breathing fresh life into the two aging architectures. TI now offers ’C67x chips with audio-oriented I/O ports, and this month, ADI announced four new members of its SHARC family that pair aggressively increased clock speeds (up to 300 MHz) with specialized, audio-oriented on-chip peripherals.
To paraphrase Mark Twain, reports of the demise of floating-point DSPs
appear to have been somewhat exaggerated. We wouldn’t be surprised if
some of the most interesting DSP developments in 2004 are in the realm
of floating-point DSPs.
Buyer’s Guide to DSP Processors, 2004 EditionThe sixth edition of BDTI’s renowned Buyer’s Guide to DSP Processors is set for publication this month. As in previous editions, Buyer’s Guide 2004 includes:
In the 2004 edition you will find new features and analysis including:
The first copy is $2,695, with additional copies substantially less. This year, for the first time, Buyer’s Guide 2004 will also be available in an electronic version (PDF) under the terms of an enterprise license. Contact BDTI at info@BDTI.com for a quotation on multiple-copy purchases or for information on enterprise licenses.
For more information, including order forms, visit our Web site at
http://www.BDTI.com.
BDTI Seminars Now On-Line!BDTI’s highly regarded seminars on digital signal processing technology are now available on your desktop! For years, BDTI’s highly rated seminars have been available only on-site for corporate clients and at industry events. Now BDTI is making these unique presentations available in streaming media format over the Internet. Two seminars are currently available on-line: Microprocessors versus DSPs A half-day seminar that will enable viewers to understand:
A single-user license for this half-day seminar, valid for twenty-one days, is $195. Processors for Communication and Multimedia Applications A full-day seminar that will enable viewers to understand:
A single-user license for this full-day seminar, valid for twenty-one days, is $395. Introductory Offer: Until December 31, 2003, purchase a BDTI on-line seminar and complete a short questionnaire, and BDTI will send you a credit for the amount of your purchase good towards purchase of any published BDTI report. The credit may be used for the purchase of any BDTI report through March 30, 2004. One credit will be allowed per person. For detailed information on both seminars, including outlines and free previews, follow the links from BDTI’s home page at http://www.BDTI.com.
Multi-user and Corporate Licenses: Multi-user and corporate licenses
are available for companies who wish to license the seminars for
multiple users. Contact BDTI at info@BDTI.com for details.
BDTI Updates Benchmark ScoresBDTI has updated its BDTImark2000™ and BDTIsimMark2000™ scores. Processors with updated scores include the CEVA CEVA-X1620 and Intrinsity FastMATH. For these scores, go to http://www.BDTI.com/bdtimark/BDTImark2000.htm.
The BDTImark2000™ and BDTIsimMark2000™ are summary measures of
DSP speed distilled from a suite of DSP benchmarks developed and
independently verified by BDTI.
New Year, New Look: BDTI and EE Times Launch Inside DSPBeginning in January 2004, look for Inside DSP, a new series of quarterly supplements to EE Times written by BDTI. In each issue BDTI will deliver its characteristic insightful and expert view of a key application area for signal processing technology. Inside DSP supplements in 2004 will focus on digital audio, digital video, handsets and handhelds, and automotive applications. BDTI and EE Times will also cooperate in publishing a monthly electronic Inside DSP newsletter, as well as a Web site, www.InsideDSP.com. Subscribers to this newsletter will automatically receive the new monthly newsletter, which will replace this newsletter.
The first Inside DSP supplement will appear with the January 5, 2004
edition of EE Times.
About BDTIBDTI is an independent source for digital signal processing technology analysis and optimized software development services. From rigorous technical analyses of processors for DSP, such as the Inside series of processor analyses, to highly regarded technology seminars, BDTI is the trusted independent source for reliable information on digital signal processing technology. As a software developer, BDTI is known for highly optimized implementations of signal processing algorithms and applications and for solutions to complex problems of integration, code size, and performance.
For more information, visit our Web site at http://www.BDTI.com.
This is the last issue of BDTI’s DSP Insider. Subscribers will receive the new Inside DSP newsletter beginning in January 2004. Previous issues of BDTI’s DSP Insider are archived and will continue to be available at http://www.BDTI.com. Both BDTI’s DSP Insider and the new BDTI-CMP Inside DSP are free monthly electronic newsletters published by Berkeley Design Technology, Inc. If this newsletter was forwarded to you and you would like to receive the new Inside DSP newsletter regularly, please register at http://www.BDTI.com/dspinsider.htm.
If you do not wish to receive the new BDTI-CMP Inside DSP
newsletter, send an email message to dspinsider@BDTI.com with the
words “Remove me” in the subject line.
BDTI’s DSP Insider © 2003 Berkeley Design Technology, Inc. |