Follow Slashdot blog updates by subscribing to our blog RSS feed

 


Forgot your password?
Close
typodupeerror
×
Intel

Intel Allowed to Buy Digital Signal Processor Co. 53

Posted by Roblimo from the big-get-even-bigger dept.
vivekb writes "C|Net reports that the federal anti-trust review period for Intel's bid to buy DSP Communications expired yesterday. This gives clearance for Intel's $1.6 billion offer to proceed. Watch for Intel DSPs to arrive shortly." Texas Instruments has dominated the market for digital signal processor chips for years. It looks like this is about to change big-time.
This discussion has been archived. No new comments can be posted.

Intel Allowed to Buy Digital Signal Processor Co. More

Intel Allowed to Buy Digital Signal Processor Co.

Comments Filter:

  • DSPs (Score:1)

    by vectro ( 54263 )
    I thought that most DSPs were put in modems, and included on chips made by either Rockwell or 3Com. Or are they talking about non-modem dsps?

    Oh, and first post.

  • Why... (Score:3)

    by punkass ( 70637 ) on Thursday November 11, 1999 @10:52PM (#1540248)
    ...do big companies like Intel feel they have something to offer in neighboring markerts? Isn't there such a thing as effeciency due to being small and fast? Big companies that do this don't necessarily make a better product, they merely weed out competitors more easily. Microsoft, Worldcom, IBM, Sun...it seems like when they buy shit, they either kill in that markert by their sheer size, or the product flops because they can't keep up with smaller, specialized shops.

    Argh, enough with the mergers already...one-stop shopping is cool for supermarkert, but that's about it...
  • Actually.. they put em in sound cards now.. In fact.. the Soundblaster Live! Value has a DSP with the horsepower of a pentium 166. :)

    Uh, third post? :)

    -Warren

  • DSPs are in modems, cell-phones, hard-drives, sound cards, and tons of other places. I wonder which market Intel is going to aim for. The article seems to state they're interested in wireless, which is currently a big business for DSP. I can't imagine that's all they're interested in though.... Hrmm......

    --Joe
    --
  • I know someone who worked for a large chemical company and had an idea of how the company could expand their business. He brought it to the attention of the company and they looked into the feasability of the project. After careful consideration they decided that they could only make a couple million on the project.

    The guy simply quit his job and started his own company to do it instead. Now he heads up a multimillion dollar corporation.

    Sounds like a real fairy tale right? The moral is that a million bucks means more to an individual that a large corporation. These big companies have to throw a lot of cash into a project to get it going because they have more overhead, and they often don't make enough profit to make it worthwile.

    -pos


    The truth is more important than the facts.
  • The reason why is that they need to grow somehow. The rule is in business, that you grow or you die. And besides, growth makes the shareholders happy :-)

    It's often easier for said big companies to grow by invading new markets than by somehow trying to stimulate extra demand in the markets you already are in. Though of course, they often try both, as that would maximize profits even more.

  • I understand that getting a finger in every pie is the wave of business today, and that the most popular method of doing so is by absorbing other companies, but is Intel really doing the smart thing here? I mean, their core business, CPUs, is, for the first time in a long while, actually being threatened by serious competition, and this is occuring at a crucial time as the focus on 64-bit processing is broadening, an arena already occupied by others that Intel will have to break into. On the 32-bit front, AMD's Athlon has been very well received by the reviewers and the OEMs are now expanding their sales of Athlon-based systems while Intel is having an apparant problem with meeting the demand for Pentium IIIs as the holiday season approaches, according to this article [go.com] at ABCNews.com. Perhaps they have gotten too secure in their niche market that they have dominated for so long to recognize when it is beginning to be seriously threatened.

    Deosyne

  • Article on DSP Comm and its founder - Davidi Gilo (Score:3)

    by Gino ( 32932 ) on Thursday November 11, 1999 @11:31PM (#1540254) Homepage
    Forbes [forbes.com] recently featured this article [forbes.com] about DSP communications and its founder Davidi Gilo. DSP Communications is on Forbes' list of 300 most promising (small) global companies.

    A nice read with some background on the company, how it came to be and noting the big breakthrough in 1994 - capturing a large piece of the digital cellular phone market in Japan. Their biggest focus is now on the American cellular market with US contracts that should contribute more than 24% of their estimated revenue for 1999.

    ...by the pricking of my thumbs,

  • yup the days of lots of different sound cards may be on their way out .... look for one standard Intel sound DSP integrated into their core logic ....
  • The move by Intel into the DSP market will be good for them. The market demand for DSPs is huge and the potential for development is also great.

    DSP's are going to shape the future of communications - Just look at the TI advert on CNN that runs for most of the day. It shows technology that I want and that is only possible by using DSP chips. TI may end up being forced out of the market and I think that the clearence given by the federal antitrust agencies should have forced Intel to create a seperate company for this rather than permit DSP Communication to be brought into the fold.

    What could they be working on? I think it likely what wireless communications are going to have big development cash thrown at them. Already manufacturing the memory and StrongARM chips and now the DSPs - it would be stupid of them not to.

  • Re:Bad timing at Intel (Score:3)

    by NovaX ( 37364 ) on Friday November 12, 1999 @12:01AM (#1540258)
    The CPU market for Intel is not really that threatened by AMD. Even if the Athlon is better by all accounts, Intel wont loose enough market share to really hurt them. AMD generally cannot fill orders, does not have a trusted brand name, and unlike Intel is not a standard choice (or the standard) for resellers.

    So the giant is loosing some money as those 'in the know' buy from its competitors. How many people are buying machines and don't know, or don't care? Intel's still the king, and that wont change on a dime. Considering their exploits of the P6 architecture, they've spent all of their R&D on a new design, and hopefully a P7 design too. What will really hurt Intel is if both, especially IA-64, don't float well on the market and leave IBM, Motorola, Sun, Compaq, and AMD to eat Intel's market share.

    Intel expaning outward is a good move. Look at Soney and IBM. If one section looses market, they're so widespread that most likely another will gain, and profits will even out. And that's still profits, so lifes good. If IBM kills the PowerPC line, they have hard drives, software, etc. all gaining in market share. So lifes good.
  • It is clear that Intel is slipping away from their original core business and maybe even starting to lose the race. Frankly I don't see how they will manage to regain the market domination they used to have.

    What else to do? Diversify is the key. Intel already makes the ARM processor (licensed from the UK's ARM holdings [arm.com]) and Intel will pretty soon have a big share in DSPC. What do these two companies (ARM and DSPC) have in common? Both are targeting the cellular market - and both have a good share in it already.

    Instead of having all its eggs in one basket, I see this as a strategic move from Intel to diversify its equity. Getting a foot in the door you may say...

    ...by the pricking of my thumbs,

  • Intel isn't. But they have people making sure that they stay legal at all levels of the company, so although they may skirt the law, they generally do not blatantly violate it, unlike Micro$oft. A monopoly is not in its self ilegal, using a monopoly to lock out competitors is.

    But Intel is looking prety desperate with regard to the Athlon and Rambus issues. Screwing up bigtime, and scrambling to scare motherboard makers into not supporting AMD.

  • Re:DSPs (Score:2)

    by azi ( 60438 )
    Modems are only one place where DSPs is used. Actually computers itselfs is only one fragment of all DSP usage. DSPs are widely used by electronics industry.

    Besides of the computers DSPs are used in (for example):

    - Motor Control Systems
    - Measurement Equipment
    - Data Communications Systems
    - Scientific Research
    - Industry Automation Systems
    and so on.

    DSP is quite new invention on computer related products if you compare it to the time line of the DSP.


    "Just the information.."

    -AZi
  • Intel is getting some attention from the federal antitrust people, but so far they've managed to avoid big consequences a la Microsoft. The FTC had a case against them, but they settled [cnet.com] with a consent decree in March. Just in time not to catch heat for trying to muscle Athlon motherboards off the market, although supposedly the FTC continues to keep an eye on them. They recently won an appeal [infoworld.com] of an injunction against them in Intergraph's antitrust suit. But the actual suit hasn't come to trial yet.

    Bottom line is they'll probably wriggle out of any serious consequences unless they do something really boneheaded, like falsify evidence at the trial. :)

  • With DSPC, Intel plans to develop and sell digital signal processors, chips that capture the digital impulses and translate them into audio for cell phones and other devices.
    So Intel is moving into the D/A converter market?

    Why don't tech reporters have any clue when it comes to DSPs?

  • Texas Instruments has dominated the market for digital signal processor chips for years. It looks like this is about to change big-time.

    I wouldn't put money on that. DSPC's product line covers only one segment of the DSP market: CDMA and TDMA wireless products. Qualcomm wrote the book on CDMA and has its own line of chipsets to handle the protocol which sells very well. And AFAIK, the TDMA market isn't anywhere near that of CDMA.

    The people who design DSP-based products select chips for their systems based on things they can measure. Texas Instruments has a huge head start with the TMS320Cx0 line in terms of available development tools and the excellent track record of its deployed units. Once you've got a library of software for one DSP, switching to another requires almost as much effort as rewriting it, so even if Intel was going into the general-purpose DSP business, you're not going to see engineers switching on trendy whims.

    I wouldn't start screaming "monopoly" or "antitrust violation," either. Aside from the x86 line, the only other processors Intel makes are the i960, StrongARM (under license) and MCS microcontrollers. None of those come anywhere near being suitable for high-performance DSP applications, nor are any of them leaders in their segment. Somebody at Intel may be realizing that the x86 cash cow may eventually stop giving milk and that if the intend to survive beyond the PC era, they're going to have to diversify.

    Texas Instruments is the Intel of the DSP industry, except that they make a better product and don't throw their weight around. (Funny how the companies that make better products and are on top of their markets don't have to resort to intimidation to beat their competition.) I've done work on the i960 and TMS320Cx0, and where the former was a bug-ridden disaster from square one, the latter was a joy to work with. (It might sound like I'm comparing apples and oranges here because the i960 isn't a DSP chip, but both are processors and one showed much better design than the other.) If Intel wants to compete in the DSP arena, they've got a lot of self-improvement and catching up to do.

    And to make this a complete slashdot post: All of these chips would be lousy in Beowulf clusters. :-)

  • A move in this direction is probably necessary to Intel. First of all to keep its stock price up they need to exhibit growth. In the consumer space Intel can only grow as the market for computers grow, once that maeket saturates their growth would largely fall off. Investors and the street doesn't like to see this happen. They get all warm and fuzzy inside when their is quarter over quarter growth.

    I'm not sure if Intel has tried entering the DSP market before or not, if they have I don't recall anything about it. One other potential growth market is embedded devices since just about everything except cheap toasters has a microcontroller in it. Historically Intel has not done real well in this market. Motorola dominates at the high end while Microchip dominates the low end.

    The second reason is that general purpose microprocessors are becoming increasingly more DSP like. Most DSP algorithms break down to doing lots of multiply and accumulate instructions at their heart. If you can do these in a small number of cycles and pipeline the Hell out of it then you can do things like a fast fourier transform, the fast discrete cosine transform, FIR or IIR filters fast which is the heart of a lot of digital signal processing algorithms. These and other mathematical operations form a great deal of what makes a DSP a DSP. Recently interest in doing this fast in general purpose microprocessors has been happening for both graphics and audio algorithms.
  • Economies of scopeare the gains from being in different markets that share (in this case) technology. Muchof what intel does is alsouseful for DSP's, making it less expensive overall for intel to make CPU's and DSP's than for intel to makeone, and someone else the other.

  • ... and announced plans to develop a more powerful wireless platform. http://www.dotelectronics.com/news/index.shtml But I'll be interested to see if Intel can take advantage of their knowledge-base to bring anything innovative to DSPs. I'd guess that they aren't planning on doing much more than focussing as much as possible on CDMA/ 3rd Gen mobile systems. (To the person who commented that Qualcomm wrote CDMA; that's true, but the standards have moved on since then. I don't remember that any of the major international mobile phone manufacturers went with anything other than TI, Lucent or Analog Devices, but I could be wrong.)
  • Damn!
    Ithinkyouneedtobuyakeyboadwithaspacekeythatworks

  • Intel DSPs (Score:1)

    by Anonymous Coward
    There's no reason why TI's lead in the DSP arena should change. TI is way ahead of Intel in terms of technology, marketting, and development tools and support. They're effectively throwing their whole weight at DSPs these days (and they are the third biggest chip manufacturer in the world). More likely than not, either Intel will grab a small niche market, or they will do about as well as they did in the graphics business.
  • Note: These are just a guess; IANAL...; yada.. yada.. yada..

    Whenever a large company like Intel decides to buy a company that makes a few niche products, I find that it is not the product that the company wants, but rights to a patent (Intellectual Property) which the smaller company either has access to via an existing contract which in turn gives the "larger" company temporary access till the patent runs out, or full ownership. So... the question then becomes, what patents does this smaller company owns that Intel wants?

    [in some cases, it may be a CYA reaction where they have already used/made a device that uses that patent, and are trying to get access to it before they release the product]

  • Almost any program with a lot of arithmetic, in reasonably regular patterns, can benefit from DSPs. This includes graphics, audio, compression and decompression, rendering movies, medical imaging, cryptographic key cracking, radar and sonar, various other things, and the traditional supercomputer problems like weather prediction and nuclear weapon simulation.

    The MMX extensions to the Pentium processor was a move in the direction of on-chip DSP. The G4 processor is a PowerPC core with an on-chip VLIW DSP. You could stretch the analogy and say that off-chip math coprocessors like the 80387 and the 68881/2 are a similar idea.

    DSPs are useful in modems because of the kind of math used for picking signals out of noise, and for decoding bits from the extracted signal. I took a course on detection and estimation once, and the math is fascinating, but it's definitely not the only thing DSPs can do.

  • Texas Instruments has dominated the market for digital signal processor chips for years. It looks like this is about to change big-time.

    and

    Texas Instruments is the Intel of the DSP industry, except that they make a better product and don't throw their weight around.

    Texas used to be tops at DSP, but AFAIK Motorola has passed them some time ago. Their 56000 series was a huge success and in fact the G4 is an audacious attempt to merge the DSP and MPU lines at the high end - it can run circles around the top TI DSP.

    Fe people remember that the G4 was originally launched at NetWorld+InterOp. The demo they had was emulating over 30 V.34 modems on a single G4... not bad at all.

  • Right now Intel has aweful margins on CPUs. They're only really making money at the high end. However a smaller percentage of the market is buying high end. Part of this is due to the changing nature of Joe, the average computer user, as well as the speed increase due to 3D cards and other smarter add ons. Intel has to find applications that will make people want to buy the high end processors. At this point in time, promising applications involve video compression (and transmission over the net), new methods of interacting with the computer, new displays, and anything other than just a core processor. All of these require DSPs. Most real time applications work better on DSP type technologies anyway. In the end, Intel is concerned about margins and keeping a reason for a high end computer to exist. To do this Intel must become a content provider, and applications provider, and, from their perspective, it's better to do it on their own terms.
  • I'd agree that TI is the Intel of DSPs. I've programmed for the 32010, 15, 50, and now the 320C6200 and 320C6211. The latter is capable of gigaoperations/sec at a price way under $100, which gives it a very wide base of applicability for small designs in many fields.

    Like Intel, the instruction architectures leave a great deal to be desired, but the speed per price is hard to argue with. And the difficulty of programming them efficiently keeps people like me employed :).

  • My favorite DSP company (and product-line) is Analog Devices [analogdevices.com]. They are relatively cheap, fast, and well architected.

    I am glad people have stayed away from buying these guys.

    -AP


  • I worked at TI for 2 years and I can tell you they know which goose laid to golden egg. They don't maintain 50% marketshare by accident just because Intel hasn't had a crack at the market. They invented DSP and they intend to rule the market.
    They aim to be the biggest semiconductor company in the world and they can get there by maintaining and expanding DSP marketshare. Anybody who challenges them (and that includes Lucent here in Orlando) has got to want it bad. I've often been puzzled by the lack of press TI gets...it's no where near the level of their technical expertise.
  • Yes, thank you for mentioning Analog Devices. Yes, Virginia, there are other DSP manufacturers than TI. The top three are TI, Analog Devices [analog.com], and Motorola, in that order. My company [c2-tech.com] uses Analog Devices, and so do a lot of the sound board OEMs.

    Another player in the field (Intel) only pushes the others toward excellence and is a Good Thing (tm). IMHO.

    BTW, Intel doesn't make any analog-to-digital (A/D) or digital-to-analog (D/A) hardware, so systems OEMs integrating the Intel DSPs would have to go elsewhere for those. A/D's & D/A's are required for most DSP apps. So: monopoly? bah!

    Hardware Guy by day, Linux slacker by nite,

    QuantumHack

  • Actually, the i960CF is market as a DSP chip, with full FP unit benchmarks and some libs. It just gets his ass kicked by TIs that cost about 1/10 :)
  • Sorry about that. It's an old gateway, and just *try* to get an upgrade or new part in a state university . . .
  • I hope that this will put some fire into TI to do a better job with their development tools. Primarily I'm pissed about the prices for their compilers and evaluation environments. Their market share is all big customers. All the small shops that I know about are using other products.
  • Oh come on. You're really just spewing a bunch of anti-AMD FUD. It's really pathetic.

    The CPU market for Intel is not really that threatened by AMD. Even if the Athlon is better by all accounts, Intel wont loose enough market share to really hurt them.

    Now, I don't know much about how sales are going for either AMD or Intel at the moment but I'm guessing that Neither Do You. What magic ball are you looking in that could tell you that Intel won't be hurt at all? Do you have any evidence (as in numbers) to back up this claim?

    AMD generally cannot fill orders At one time, this was true, but it is not anymore. All of AMD's demand problems are a thing of the past. When their Dresden fab comes online, it will be even less of a problem.

    , does not have a trusted brand name,

    Trusted by who? The general consumer? Well, according to you later on, they don't seem to care at all. The power user? I think they would care less about brand name and more about the fact that the Athlon is fastest x86 chip around. OEMs? I would claim that the relationship AMD as with OEMs has only improved over the past few years as they consistently produce fast and inexpensive chips.

    and unlike Intel is not a standard choice (or the standard) for resellers.

    Who cares about standard choice? An AMD chip/motherboard is going to work just the same as an Intel chip/motherboard. The only thing that might make a difference is optimizations, and I've seen a lot more support for 3Dnow then for Intel's SSE.

    So the giant is loosing some money as those 'in the know' buy from its competitors. How many people are buying machines and don't know, or don't care?

    Exactly! AMD chips cost less! People care about that. They know or can find out from a saleperson that an AMD chip will run just as fast or faster than an Intel one for cheaper. If they don't care about what chip they are, they'll go with AMD.

    Intel's still the king, and that wont change on a dime.

    So is Microsoft. Are you paying homage to them too?

    Considering their exploits of the P6 architecture, they've spent all of their R&D on a new design, and hopefully a P7 design too.

    The fact that they have dragged the P6 core this long is *not* an argument for Intel being a strong company. The fact that all of theirs chips are based on a 5-year old core is a *bad* thing for them. Recenct reports show that their P7 design was scrapped. Intel doesn't have any new technology coming in. AMD does.

    Who the heck gave this guy a +3? This comment has no substance but bashes a good company that brings new technology to people at a reasonable price. Can you even imagine what the CPU world would be like with no AMD? We'd all be screwed.

  • I never called AMDa "bad company" nor would I. Your taking many liberties on what I was saying.

    1. I never said Intel wont lose money. I said that it wont hurt them, and that to me means a major drain in capital. Intel has tried to undercut AMD at times so AMD will lose more profits, and Intel's hope is leave the market. Unlike AMD, Intel has deep pockets which means if they lose profits, but so does AMD - AMD will have a rougher time than Intel. That holds true in any market, as long as the bigger business is not losing a significant amount to empty their pockets.

    Intel x86 chips are still standard in Dell laptops and desktops, as well as many other large resellers. Smaller chains and stores sell a good deal of AMD based machines, but also many Intel based ones too. Intel has pressured motherboard manufactures in attempts to hurt AMD even more.

    2. AMD's brand name is not as well known or well acreditted as Intel's. Intel lived of the Pentium brandname for years, which has for the most part been in good light. Before that the public knew of Intel because they have always been the major supplier of PC processors, which the Intel Inside campaign strengthened. Businesses and even the non-computer savy public would sooner trust Intel than AMD - its all through name brands. I've read numerous articles claiming that businesses will be weary towards multi-proccessor AMD machines, as Intel has proven itself in SMP over the years. Read Creating the Digital Future by Albert Yu. Oh, and I never said something new cannot be better, it just has a hard time replacing the old.

    3. Being a standard choice for resellers means Intel has significant market. That does not mean AMD and chipset/motherboards will not be compatable (though I remember the software warnings of incompatabilities before the K6).

    4. AMD obtained a significant market share by undercutting Intel. People do care, but AMD lost significant profits combatting the Celeron processor. Intel made significant money by creating chips cheaper than its rivals on other platforms. I agree people do care, but that doesn't mean everyone immediately flocks for the cheapest brand.

    5. I mentioned the exploits in showing that Intel has not spent money on R&D in the x86 processor business compared to AMD. Intel has merely made tweaks to a processor and than resold it, while AMD has reserched, tested, and brought to market new chip designs. Why would anyone be happy that Intel has exploited its userbase by refusing to improve its technology if being idle creates such massive profits? One could easily say the same with Microsoft Windows, considering how it used Windows 98, and soon Windows Millenium. It created Windows 95, and will exploit that codebase with tweaks for 5 or more years. Its the same.

    If your right that Intel scrapped their P7 design, than Intel's in trouble. Plain and simple. All they have left is to use the hype of the IA-64 technology to sell their new high end chip, and perhaps begin to pay attention to the personal computer market after AMD wins it. AMD's competing for survival, Intel for deeper profits. Guess who has, and will continue, to work in the intrests of consumers?
  • "Can you even imagine what the CPU world would be like with no AMD? We'd all be screwed."
    Nonsense. Microsoft would have had plenty of time to develop a lean, efficient follow-up to Windows 3.1 designed to take advantage of the soon to be released follow-up to the 486 (rumours of any need to call it something besides the 586 are just silly)which I hear will sell for less than $1,000 and that guy in Finland with the funny name (Snoopy or something like that)is really wringing every last drop of performance out of those 386's.

  • "Texas Instruments has dominated the market for digital signal processor chips for years. It looks like this is about to change big-time."
    Thought this would get a bunch of people grabbing up Intel shares first thing Friday morning but instead the price took a beating.

  • I'm worried about getting drivers for whatever cool thing Intel makes with their DSPs. I hope that they will keep with their past practice of releasing most everything really necessary about their stuff to developers, but Intel could do something different for DSPs. If they make some awesome sound card or other thing that uses an Intel DSP and I don't have drivers for it, I might just get mad.

    What will Intel do with a DSP? Same thing it did with the 8086. Make a functional core, then pile loads and loads of junk on top of it while keeping it compatible with release 0.001pre1. Then, once they think that there has been enough junk added on in the core, they halt all core development and start coming up with little add-ons that slightly increase the usability and performance of the DSP chip. Then they will up the clock speed of the dang thing without changing any of its internals and charge us more for the one that just so happens to run faster. We will have DSPs with a clock rate of 1 gigahertz or more, and people will finally realize that when you are working with audio, 1 gigahertz is a bit excessive. They will keep upping the speed anyway. And of course, someone will make an arcade emulator, an mp3 player, a text editor, and a web browser for the poor thing that wasn't supposed to do anything other than in the mp3 player category. And people will have fun. They will discover neat effects that are supposed to make their sound sound better, and when they hear the exact same thing, they will praise it to the ends of the earth. Then they find something that actually changes how it sounds, and they think it makes it sound worse. Then the crackers will find a way to exploit some malicious instruction, and Intel will give each one of the things a globally unique id. Then web advertisers or whatever will be able to figure out when you last replaced your sound card and in whose computer the old one went. They won't care, obviously. Oh, I almost forgot. What cool stuff can I do with one of these anyway? If it takes you longer than 2 seconds to answer that, I just wasted my time writing all of that LAME stuff.

    Gotta go. Ken

  • I'm not a tech person, but I did understand a good portion of what I read about the so-called Velocity Engine for the G4. The vector processing stuff is what gives that processor its power. So a vector engine is a DSP? Sounds reasonable. If someone could get a definition (reply to some other comment; this one is too deep) in here, maybe that could help relative newbies like me get a clue what is going on, particularly (1) What a DSP is, exactly, and (2) Why should I care?. Then I will be happy, and some other people will be happy to learn a little interesting tidbit of knowledge

    Ken

    PS - Will an Intel DSP make my modem faster? Seriously, will they be any better then their competition? Am I asking too many questions? Am I right to say yes to my own question?

  • BTW, what do you think of the C6211? :-)

    --Joe
    --
  • I like it, though it has a very steep learning curve -- lots of strange architecture details to absorb. The C6 instruction set is very nice compared to past models, much more complete and fairly orthogonal. It's quite easy to program the part in C or using the scalar assembly optimizer tool. You can achieve half theoretical speed or better fairly easily that way. It's a real pain to hand schedule eight parallel streams -- and incremental changes simply don't exist then -- but the part *really* screams when you do.

    It requires a slightly different mode of thought due to things like branches affecting all parallel streams, and the instruction latencies for loads, branches, and multiplies.

    If you stay within the internal 4k data and 4k code the '211 is fairly self contained, and memory is fast, which makes it's low price very attractive. Going to external memory on the '211 is a slow and painful process unless you get a cache hit.

    The tools are still a little new, but I haven't hit any show stopper bugs, just some bad ones, like a misassembled opcode.

  • To try to save myself from sounding stupid:

    There could certainly be an interest for developing DSP products for PCs, in such areas as sound cards and graphics boards (although I'm not completely sure about the usefulness of the second one. A DSP that is already very connected with personal computers is the so-called Velocity Engine on the G4. (I'm getting this information from someone else, so I may be wrong.) It employs advanced vector processing to make a very fast and efficient SIMD (single instruction multiple data) CPU design. I also mentioned graphics cards. A digital signal need not be a sound. It may be possible (I haven't researched this) to implement very fast image transformations using DSPs. Given a suitable implementation, a DSP could accomplish many things a CPU of a computer could; the Velocity Engine proves this.

    The text editor / web browser stuff was supposed to be funny but obviously wasn't. What was funny was a web browser for a TI-8(something) calculator. You had to hook it up to a laptop computer to get the Internet connection, so why not just use the laptop computer?

    I have heard a lot about Fourier transforms, but I don't really understand them (that is where I really am confused). I know that they are used for COFDM (see my Digital Television Transmission Standards article from last Sunday) to create thousands of individual carriers without having thousands of different oscillators and the like. Could you provide a link to somewhere that explains this technology (since you seem to know about this)? If it's a good site, it's a good chance that Nerds would be interested in it.

    I have gotten tired of telling Anonymous Cowards to get a Slashdot user account. You can do that easily from users.pl [slashdot.org] on the Slashdot main site. Then you can do all sorts of cool stuff like customize your main page and make your comments start at Score 1. I've seen plenty of good AC comments down at the bottom of the comments because they had low scores.

    See 'ya later, digital friends: Kenneth

  • C6211 (Score:1)

    by Mr Z ( 6791 )

    You mention: "If you stay within the internal 4k data and 4k code the '211 is fairly self contained . . ." Don't forget that there's the 64K L2 on there, which can be configured partly as SRAM and partly as cache. Cache misses to L2 aren't all that expensive.

    Overall, though, looks like you've got a good impression of the part. The reason I ask is that I work with the folks who put the C6211 together. :-) I'm always interested in hearing "reports from the field." Perhaps send me an email sometime?

    --Joe
    --
  • I'm not aware of a specific criterion that distinguishes a DSP from a normal processor. DSP as an intellectual field preceded DSP chips. People noticed that there were a few mathematical operations that were very useful in signal processing: discrete-time filters, fast fourier transforms, discrete-time convolutions, etc. It seemed logical to start designing chips that were carefully optimized to do these things as fast as possible, as opposed to more general processors that would be optimized for, say, running an operating system (e.g. the 68000).

    To the best of my knowledge, the business of DSP chips started with Texas Instruments' TMS320C10, a quirky little 16-bit processor with a 4-deep hardware return stack, built around a multiply-accumulator (MAC). The MAC meant that the chip could be very efficient for performing vector dot products, which are the inner loop in all the DSP operations. There were many (fairly ugly) hardware hacks embedded in the 320C10 to make sure that dot products were as fast as possible, at the expense of anything like elegance or comprehensibility. When my boss handed me the 320C10 data sheet in 1983 and asked if I wanted to work with the thing, I cringed.

    Since then, Texas Instruments is still quite prominent in the DSP chip business, but they've been joined by others like Analog Devices, National Semiconductor, and Motorola. DSPs now come in a range of prices and capabilities, and you can get both fixed-point and floating-point DSPs. As with the rest of the semiconductor industry, DSPs have been steadily increasing in speed and complexity.

Slashdot Top Deals

He has not acquired a fortune; the fortune has acquired him. -- Bion

Close