Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!


Forgot your password?

Intel Shows Off 'Banias' Chip for Mobile Devices 180

deano writes "Intel has unveiled the first prototypes of their latest mobile "Banias" processors. The article states Banias systems with the Intel Odem Chipset will come out early 2003 and feature 802.11b. The article also speaks of the new Itanium with a 6Mb cache!"
This discussion has been archived. No new comments can be posted.

Intel Shows Off 'Banias' Chip for Mobile Devices

Comments Filter:

  • Why Intel is setting themselves up for this, I have no idea. ;)

  • 6 MB cache? (Score:2, Interesting)

    by Tony ( 765 )
    6 MB cache? The UltraSparc III has an 8 MB cache. Intel is still playing catch-up.

    Of course, it wan't that long ago I was excited to have 8MB *RAM*.
    • by larry bagina ( 561269 ) on Friday April 19, 2002 @01:09PM (#3374449) Journal

      6 MB cache? The UltraSparc III has an 8 MB cache. Intel is still playing catch-up.

      No, they're playing cache-up.
      • The really interesting part is this: (quoted from intel.com [intel.com])
        Unified 2MB or 4MB on-cartridge L3 cache. Runs at full processor frequency and is organized as 4-way set-associative with 64-byte cache line size. Fully pipelined and optimized to provide fast access to data at a bandwidth of 12.8GB/sec using a 128-bit wide cache bus.
        Now that's cool.
      • bananas and ketchup? yuk!
      • I thought that's what you called what the bully vegetables did to the little tomatos [when they beat them to a paste].
    • Re:6 MB cache? (Score:4, Informative)

      by Glonk ( 103787 ) on Friday April 19, 2002 @01:20PM (#3374543) Homepage
      6 MB cache? The UltraSparc III has an 8 MB cache. Intel is still playing catch-up.

      The UltraSparc III has off-die L3 cache. The Intel chip would be on-die.

      Off-die L3 cache isn't too hard to do, and it's significantly slower.
      • No, according to another post which provides a quote from Intel it would be "on-cartridge" not "on-die." That could be a big difference, thinking a cartridge is like an old Pentium II slot1 cartiridge which is just a mini PCH board and hence the cache may as well be on the motherboard.
    • by Anonymous Coward
      > The UltraSparc III has an 8 MB cache.

      And that "spontaneous reboot" feature...
    • 8MB RAM? Hell yeah! I remember *octupling* my RAM to that amount, the full capacity ;) My computer now is much the same way, running with one eighth of the total possible RAM, but I don't see myself going up to 2 gigs any time soon
  • by PissingInTheWind ( 573929 ) on Friday April 19, 2002 @01:05PM (#3374421)
    The poster seems to be all excited saying:
    The article also speaks of the new Itanium with a 6Mb cache!"

    Well, it's important to see that a fast and useful cache technology is much more than a big cache. You need good access times, good hit ratio, etc. Sometimes you can even get better performance by disabling the cache.

    It's application-specific, don't assume bigger is better.

    • Well, at least in this case, bigger might not always be better, but it certainly doesn't hurt! :)
      • by Anonymous Coward
        bigger might not always be better, but it certainly doesn't hurt!

        Oh really? Ask the goatse.cx [goatse.cx] guy if that's true...

      • Sure it can hurt. If increasing the size of the cache causes your access time to increase, then it can hurt very badly, because you're overall latency could increase if the benefit from a lower miss rate if offset by the higher hit time. Although it's true that the latency of second and third level caches affect performance much less than the first level.

        Cache size is one of the most misleading processor benchmarks, more misleading than frequency, yet big caches command huge price premiums.
  • ... that the Intel Roadmap has been displaced -1.5 yrs. The 6 MB cache Itanic will be at least 18 months late. So sad - the chip has so much potential despite it's bad press. I've had a Sitka 450 2MB cache server for 3 years and its got to be the best Intel product ever made.
    • I find a couple things interesting in the page your sig points to.

      1) Windows and Linux can coexist on the same computer. For additional information, refer to your Linux documentation.

      Hey, at least they admit that Windows CAN coexist with other OS's. I was shocked to even see this.

      2) The Fdisk tool included with Linux can be used to delete the partitions. (There are other utilities that work just as well, such as Fdisk from MS-DOS 5.0 and later

      So, do they really think that DOS fdisk is as powerful as the Linux equivalent? They can't really be serious. You have to jump through major hoops to get it to delete an NTFS partition most times.

      3) Also, Linux recognizes more than forty different partition types, such as: (see the page for the list)

      Another surprise from MS. Of course, they do not mention the limited partition type support of Windows.

      Well, I just had to comment.
      • That's the point. My .sig is heavily laden in sarcasm. I took one look at the MS page and laughed. It turns out that they have a pile of "how to remove Linux" pages, often with wrong information and hardly mentioning the possibility of a dual boot machine. It cracks me up, and I have to admit, I love my .sig.

        Thanks for the feedback.
      • Please tell me how you would get fdisk.exe to remove an NTFS partition? I've always had to use delpart.

  • Meaning of Banias (Score:5, Informative)

    by Anonymous Coward on Friday April 19, 2002 @01:09PM (#3374447)
    Posted AC to leave the concept of karma with at least a little dignity. Taken from a March 2001 ZDNet article [com.com]:

    Banias, formerly Caesarea Philippi, is the Arabic name for the Hellenistic city of Paneas whose name derives from Pan, the Greek god of herds and shepherds. His cult was observed in a large cave at the foot of Mount Hermon, where a source of the River Jordan emerges.

    Pepperdine University has conducted digs in the area that have unearthed parts of a palace from Herod Agrippa II. Modern-day Banias is located in Israel, where the Intel design team for the new chip is based. The company typically code-names its chips after geographical features.

    • Welcome to the Cult of Intel. I am your leader Banias. [giggles from the back of the room]
    • The company typically code-names its chips after geographical features.
      Like Chicago and Memphis and... oh wait. . . nevermind.
    • The company typically code-names its chips after geographical features.

      is this the code name or the release name? i thought the word 'Pentium' came to be by way of researchers determining that it wasn't a real word in any language and therefore could be copyrighted (same with 'Itanium' i suppose as well)...

    • "Pan, the Greek god of herds and shepherds"
      Is it just me or is this a subtle dig by Intel at the consumer market? If banias is the shepherd, who are the sheep?

  • Really? (Score:4, Funny)

    by Indras ( 515472 ) on Friday April 19, 2002 @01:09PM (#3374448)
    I heard you could make a battery out of a potato or a lemon, but a processor out of a banana?
  • Perhaps all those countless hours of interfacing one's computer to a network can all be replaced by the Banana and other technological advances. The largest hinderance that holds people back with bringing laptops to LAN parties is the performance. Along with the introduction of the Raedon Mobile into the laptop community, chips like these configured strictly for mobility purposes will perhaps change the way LAN parties and the such are done. Bravo!
  • "The article states Banias systems with the Intel Odem Chipset will come out early 2003..."

    And in 2004/5 the news headlines on C|NET will talk about about:

    Intel's New Slip Up - Banana Processor Fails to Satisfy

    Why do we predict failure for products with silly names? Is Intel running out of accounting codes or something? ... Maybe their next processer will be called "Placenta [dictionary.com]" .

  • we have no Banias, we have no banias today.

  • Odem chipset? (Score:1, Offtopic)

    by Mr. Neutron ( 3115 )
    Isn't Odem the guy in the Bible who was smited by God for masturbating, or something? Perhaps Odem would be a better name for a pr0n^H^H^H^H "graphics and multimedia" chipset.
  • Are wireless NOCs something that may be coming down the pipeline? More and more webmasters are making the transition to dedicated servers but are still held back by the price tag... this could be the kick in the butt needed to help this sluggish and saturated market come back to life again.
  • Meaning of Banias (Score:5, Informative)

    by Dicky ( 1327 ) <slash3@vmlinuz.org> on Friday April 19, 2002 @01:13PM (#3374488) Homepage
    For reference, Banias is the name of a river in northern Israel - one of the three sources of the river Jordan. It's a pretty area, and great for gentle hiking. There are nice pictures of it here [jewishtour.com], here [zentraveler.com], here [ucsd.edu] and through Google image search [google.com]
  • McKinley, which Otellini said will go into production "midyear," is built on a 180-nanometer process and has about 220 million transistors. Madison is built on a 130-nanometer process and has about 500 million transistors, he said.

    They went to nanometer terminiology instead of micrometer ...

    And DAMN ... 0.5 BILLION transistors ... impressive.

    • by Anonymous Coward
      Yeah, that is what 6MB of cache will give you though.

      6,000,000 bytes
      * 8 bits to a byte
      * 6 transistors to a bit
      = 288 million transistors on its own

      Add on the 1MB L2 (48 million transistors) and there you have the majority of the transistors in Madison (Avenue) processor. And the L2 cache is most likely has more than 6 transistors to a bit of storage, and so on - 450 million of the 500 million transistors are most likely for cache or cache control in the end.

  • by SamTheButcher ( 574069 ) on Friday April 19, 2002 @01:18PM (#3374526) Journal
    ...I read "ba~nos", the spanish word for "bathroom".

    Where intel chips belong, of course.

  • Looks like transmeta is in trouble. The description of the chip in the article (shuts down unused parts completely) makes it sound like a low-power application chip. That's exactly what Transmeta is trying to market. An intel offering in that arena can only hurt Transmeta...

  • by Ed Avis ( 5917 ) <ed@membled.com> on Friday April 19, 2002 @01:21PM (#3374551) Homepage
    It sounds like this chip will become 'unexpectedly' popular for desktop systems as well as portables. Things like wireless Ethernet will be useful on the desktop too, and if the chip runs cool then it won't need a noisy fan. (And reduced power consumption means you can use a fan-less PSU as well.)

    I wouldn't be surprised to see Intel making a desktop version of this chip as the Celeron replacement, depending on what AMD come up with.
    • The wireless ethernet is part of the chipset, so basically, it'd be like onboard ethernet, but why on earth would you want a friggin mobile cpu in a desktop? You don't have to worry about power consumption, space limitations, excessive heat, etc.

      There are already alternatives for wireless on desktops as well, I believe linksys makes a PCI card for wireless now, you can also buy a pcmcia thingy for a desktop and slip a pcmcia wireless card in that.

      The chipset probably won't have AGP or PCI support, or limited, so why on earth would you want it in a desktop?

      The article does mention that:
      To cut energy consumption, Banias automatically shuts off its different subcomponents when not in use. Although it's designed for notebooks, the chip will also appear in thin "blade" servers.

      I'm curious what the cost of turning off the subcomponents is to performance? Anyone have a clue?

      • Yes, a desktop does have to worry about power consumption and *definitely* space limitations. We're always being told how computers are getting smaller and smaller, yet the desktop PC hasn't shrunk at all recently, if anything it is bigger than ten years ago (minitower cases, larger CRTs).

        You don't really need AGP, it's not essential for a video card (unless you play certain games). Not having PCI support is more contentious, but if the stuff you need is on the motherboard you could build a desktop PC without PCI slots. I mean what does the average desktop have beyond video, sound and Ethernet?

        I'm thinking of something like the IBM PS/2 E [tripod.com], which was essentially a laptop in a desktop case. It had four PCMCIA slots, an LCD screen and trackpoint keyboard. Video and IDE interfaces were on the motherboard. The machine is very small and, when the HD spins down, completely silent. With its 50MHz 486SLC2 processor it wasn't a speed demon even at the time (1992), but quietness, reliability and a small footprint are sometimes more important than raw speed.

        • I'll revise my statement a bit =] Computers could definitely be made smaller and more power efficient (for us power hungry USians, we just don't really care about power consumption too much, we leave our home pcs on 24/7 anyway, eh?) But, smaller pcs cost a little bit more, and the parts (such as a laptop hard drive) are a little more susceptible to damage. If you're that worried about space, go buy a new iMac.

          PCI video is nice, but AGP is better. I'd argue for more than just games, but that's all I use that is heavy in the video department so I can't speak for it too much.

          Also, is PCMCIA still limited to 11mpbs throughput? I haven't looked up on this lately, but that's another to consider in lieu of lack of PCI =]

          Another odd note.. I just bought a small box and put a flex atx in it, so maybe I care about space too =]
      • I'm curious what the cost of turning off the subcomponents is to performance? Anyone have a clue?

        Obviously I don't know the implementation specifics, but I still have a clue.

        The normal method of "turning off subcomponents" is to simply gate the clock to those parts that aren't being used that clock. E.g. when running pure integer code, the floating-point unit will not be clocked, saving the power consumed by the local clock drivers, flops, and other circuits by preventing the transistors from switching. If the floating-point adder is used but the multiplier isn't, then the clock is gated to the multiplier. This has essentially zero impact on performance, since the only things it affects are things you weren't using anyway.

        As transistors get smaller, leakage current -- which results in power loss even when the transistors aren't switching -- becomes a bigger factor. A more ambitious power-saving method would be to drop the high voltage rail for the subcomponent, eliminating leakage current. This is a lot harder (especially without SOI), and I don't know if anyone does it. Anyway, this would have a performance penalty as you wouldn't be able to turn it on and off on a cycle-by-cycle basis. It would take several (possibly many) cycles for the voltage rail to stabilize.

        My money is on it just being some basic clock-gating.
      • The chipset probably won't have AGP or PCI support, or limited, so why on earth would you want it in a desktop?

        You're talking about the chipset - not the CPU. It's very possible, if not likely that they will produce more then one chipset. How many chipsets does the P4 now have?

        • It was more of a mixed comment. The person I replied made it sound like the wireless was part of the cpu. The story said that there was a system being built on a specific chipset that would have the wireless on it. I think that is cool, but that specific one, designed for a laptop, probably isn't going to have AGP or PCI support, so you would want a different chipset if you were going to put it in your desktop. Make sense?
  • New Poll!! (Score:5, Funny)

    by dbretton ( 242493 ) on Friday April 19, 2002 @01:25PM (#3374579) Homepage
    With the new "Banias" line of Intel portable processors, which of the following should be Intel's new company mascot?

    1. This guy [thinkquest.org]
    2. This guy [pbskids.org]
    3. This guy [gasolineal...tiques.com]
    4. This guy [ufl.edu]
    5. This gal [danheller.com]
    6. CowboyNeal! [drunkmonkey.org]

  • Would you buy a 200mhz Pentium with 6mb of cache? Are you willing to rewrite your entire code base to take advantage of Itanium's architecture? With Itanium chips, those are your options. Wait for Sledgehammer or Whatever Intel cooks up at Ronler Acres to clone x86-64 (AA-64). 6mb sounds exciting until you address the costs associated moving to the Itanium platform.

    But wait, with 6mb of cache, you might not have to optimize your existing code for the Itanium. You could just rely on BFI (Brute Force & Ignorance) to solve your troubles.
    • Are you willing to rewrite your entire code base to take advantage of Itanium's architecture?

      Ever heard of a high-level language that isn't architecture bound? How about an optimizing compiler? I don't have to rewrite code for any architecture. Even the performance geeks usually have C code that does the same thing as the optimized assembly.
      • Except as I understand it, even Intel's own compiler does not do a very good job of optimizing code for IA-64 making optimizations of code by hand in machine language nessecary. If your data set fits in 6mb of cache, and you need the performance, then moveing to IA-64 is a good option. The 6mb stipulation implies that you won't have to do much hand codeing because the data set is so small. If you have a larger data set, use Alpha, because it's performance is so much better, it's compiler is mature, and Intel is expected to incorperate Alpha's technology into Itanium in the future, by which time the compiler should be mature, if Itanium survives the onslaught from AMD's x86-64 Hammer, Apple's G5, Sparc x, and even Alpha, which is king now, and may, if ever relased in it's next generation, continue to pose a threat.

        Nice run on sentance if I do say so myself... Hope it is clearer for you :)
  • Seinfeld? (Score:2, Interesting)

    Isn't Bania that other comedian from seinfeld who was kind of annoying and always stole Jerry's bits?
  • Otellini also said that Intel's hyper-threading technology, which lets a single processor act in some ways like two, will debut in desktop Pentium 4 systems in 2003. Hyper-threading currently is available only on Intel's top-end Xeon version of the Pentium 4 and is enabled in servers but not workstations. The technology allows two different applications to use different components of a microprocessor simultaneously.
    Not to be availablke as retail, huh&nbsp?
    This actually reminds me of the 486SX which was a 486DX with its math copro deactivated...
    • Thankfully, my Linux box would be able to take advantage of the hyperthreading. Unfortunatelly, for the masses who still use Win95/98/ME, they don't take advantage of multiple processors (DOS based, go fig.), and I know that for WinNT based systems, you need a license for each processor. That has gotta hurt the wallet something fierce. Thank you, GNU.
      • Win95/98/ME are dead, you can stop talking about old OSs now unless we want to start comparing XP to Slackware 3. I mean who in 2003 is going to buy a new hyperthreaded p4 machine and put a 7 year old OS on it? How many of you are running slack 3 on your new AMD/P4 2 gig machines?

        Second regular single user liscence of NT/2k/XP allows for 2 processors so it will work (to some degree) without paying extra.

  • When will Intel start having *cool* processor code names? Like "Death Eagle" or "Killer Robot" or "Massive Rampaging Gorilla"? Sure, "Thunderbird" sounds cool, but "Death Eagle" is way cooler! I bet they could really have a marketing coup by picking better names. Or maybe just marketing them under that name to kids or something, and leaving erudite poosae names for business marketing... or something.

    Just think... the Intel Death Eagle 4! They could market it as 2 million KILLERhertz! Yeah!
  • So... does this processor demand 2 dinners at "Mendi" in exchange for an Armani suit?

    "Mendi is the best, Jerry... the BEST."

    For those of you not familiar with Seinfeld... Kenny Bania was a terrible commedian (hack) that Jerry couldn't stand. Although later, he became Bania's mentor writing the now imoratal "Ovaltine" routine...

    "Why do they call it 'Ovaltine'? I mean, the jar is round... the glass is round... They should call it 'Roundtine'."

    "This is GOLD, Jerry!"

    Ok... so I need help. Damn tivo's ability to record every Sienfeld shown!

  • So who else first saw that as "Intel Shows Off 'Bananas' Chip for Mobile Devices"
  • "Yes, we have no Banias. We have no Banias today".
  • 4/19/2002

    "McKinley, which Otellini said will go into production "midyear," is built on a 180-nanometer process"
    "Intel did not demonstrate Madison working in an actual computer or disclose when the chip would be delivered"


    "AMD's demonstration featured "Hammer" running both a 64-bit Linux and 32-bit Microsoft® Windows® operating system. The AMD "Hammer" processors were manufactured on 0.13 micron, Silicon on Insulator (SOI) technology, which together enable higher performance and lower power consumption."

    So Intel's 64-bit processor built on a .13 micron process still doesn't work, yet AMD showed theirs almost 2 months ago!
  • Ooh, a 6MB cache. We've got a crusty old Alpha with an 8MB cache. So what? Someday Intel might buy into a fast worstation chip and make good use of it. They did buy the Alpha, but I doubt they'll make good use of it. That's because the only way I believe Intel could make good use of the Alpha would be to reunite the Alpha team and continue its development. It really appears that Intel will never design a fast workstation chip.

    -Paul Komarek
    • With todays chip architectures, you can't simply compare 6MB to 8MB and say 8MB is better. Everything needs to be in balance to get max performance. Cache size interacts with cache latency which interacts with main memory latency which interacts with how many caches you have which interacts with how many cache ways you have which interacts with replacement policy which interacts with branch prediction algorithms which interacts with code and data prefetch algorithms which interacts with the compiler's code optimizations which interacts with the compilers data layout optimizations... etc, etc. And then, let's talk aps-- how big is the data footprint? code footprint? tight loops or lots of branchy non-loop code traces? Float intensive? Lock intensive? look-up intensive? decision logic intensive?. Go spend a couple of years learning the basics of benchmarking, then come back and share your wisdom. This is not a defense of Intel... my criticism would be the same no matter who's products you were comparing so simplisticly.
      • I agree things aren't that simple. I was making fun of the front page comment about 6MB. The Alpha performance I refer to don't come from reading magazines; they come from the last three years I've spent getting intimate with a handful of Alphas and x86 machines. What I don't know about is the Merced. But that's because nobody I know at Carnegie Mellon University has one. From what I have read, the only researchers who think the Merced is worth the money are the researchers who got one for free.

        At any rate, I wasn't attempting to share Wisdom. I was attempting to share Scorn for the garbage put out by Intel over the last 10 years. Thanks for the free refreshing on cpu microarchitecture. ;-)

        -Paul Komarek
  • It was hand carried through Intels's Hudson, MA. Fab in 16 days. Nothing was to stand in its way.
  • by Sebastopol ( 189276 ) on Friday April 19, 2002 @03:13PM (#3375280) Homepage
    sure is lots of "+5 funny" goin' to people for saying "huh huh... banana.. huh huh... huh huh ... odius ... huh huh... "

    is everybody on this thread six years old or what?

  • So I'm wondering if this chip is worth waiting for in a laptop. The main power usage in a laptop should be the screen, discs like a HD or CD drive, and then a CPU, right?

    And right now tha Banias is so far off in the future it won't change my upcoming purchase of an ultralight notebook.

  • To cut energy consumption, Banias automatically shuts off its different subcomponents when not in use. Although it's designed for notebooks, the chip will also appear in thin "blade" servers.

    That was my thought, too, when I heard about a chip that had a notebook-motivated balance between performance and power consumption.

    That it would find some acceptance in the server room, where power dissipation issues loom large.

    And then I thought: why stop there?

    Why not use these in the next generation desktops, too? So that people won't need those noisy fans and big honkin power supplies? Most people are just running email, Word and a browser.

    I mean, as anyone looked seriously at the huge gap that exists between how much of the desktop CPU power is actually needed versus how much is available if the processor runs flat out?

    If there were some multimedia hardware implementations in these things, I doubt there would be much need for anyone to go out and buy a 3.4 GHz Pentium 4 or whatever McKinley/Madison.

    Those minority applications that really need CPU power should just use a rack of these things.

  • In India, in the national language (Hindi), the shopkeepers are called banias (ba-nee-yah).
  • (or something like that)

Do not underestimate the value of print statements for debugging.