Slashdot Mirror


Via One-ups Transmeta

An aonymous reader submitted that"Via just announced the Eden platform, which promises lower power consumption than Transmeta. If it follows the C3 line of CPUs, I'm guessing it will also deliver much better performance at a lower cost (the C3s gave significantly better performance than Transmeta, but at just under 10W, so a bit more power)."

31 of 112 comments (clear)

  1. Excellent.. by PopeAlien · · Score: 2

    Even if this doesn't pan out to be as good as it sounds, I love to see competition in this area. I would really like to have a fanless computer for my desktop, and a laptop that can last for more than a day on a full charge AND run some high-falutin' graphics & games..

    1. Re:Excellent.. by YKnot · · Score: 4, Offtopic

      There are more energy-hogs in a laptop than just the cpu. Display, harddisk and recently the graphics accelerators need a good share of the battery life. Decreasing the power consumption of the cpu alone won't get us very much closer to the one-day-per-charge laptop, although it's certainly a step in the right direction. Other interesting applications of less wasteful processors are clusters and servers, which otherwise need expensive cooling.

  2. If only Transmeta would release a different CPU by Drakino · · Score: 4, Insightful

    The low power chips are nice and all, but where is the CPU showing off Transmeta's true technology? All that code morphing stuff should enable a laptop to be made with a switch labeled "G4" or "x86".

    1. Re:If only Transmeta would release a different CPU by stripes · · Score: 3, Insightful
      All that code morphing stuff should enable a laptop to be made with a switch labeled "G4" or "x86"

      The TM CPUs have a lot of x86 like functionality wired into them. Sure they don't execute x86 code, but they do have x86 style MMUs, not PowerPC ones. They do set their condition flags based on when an x86 would, not a PowerPC. The 40 or so GP registers they have is plenty to try to emulate a CPU that only has 4 or so, but not so good for CPUs that really have 16 or 32 (you need to use some registers for the morphing code, some to hold state that may not come to pass, some...).

      The current Transmeta chips are not x86 CPUs, but they emulate x86 CPUs far far better then they can emulate any other CPU. This could be addressed by future TM CPUs, but only by adding things that mostly wouldn't be of use to them while being an x86, so if x86 is where 95+% of the market is, it may make sense to not even make the TM CPUs 6% more complex... let alone the morphing software 100% more complex.

    2. Re:If only Transmeta would release a different CPU by ghjm · · Score: 2

      Well, yes, that makes sense from a business perspective. Undoubtedly x86 is where the money is, for now and for a while to come. Transmeta is correct to play mostly in the x86 space.

      So what's the point of code morphing?

      I mean, if they're just going to be a low-power x86 clone, surely the resources (engineering, space on the die, whatever) taken up by code morphing could be put to better use as...nothing. Just make an x86-only design. AMD did, and theirs is better. Presumably because they focused all their effort on making a low-power x86, and not on code morphing.

      Seriously. Either develop some useful capability from the code morphing tech, or abandon it.

      -Graham

    3. Re:If only Transmeta would release a different CPU by stripes · · Score: 2
      Wouldn't it be great if you could run mac software at the flick of a switch? I agree, for the performance some more silicon may have to beadded. But if you don't optimize very much at the start you wouldn't need that much extra.

      Still makes the software design (code morpher, and BIOS) about 100% more complex. Maybe only about 50% more complex since they did a Java version. It would also help hilight it's slowness "Can't even keep up with a dead 300Mhz Mac", and "Jack of all trades, Master of none" are not headlines they need.

      As for running Mac software, it is great to be able to run it at the flick of a switch (or at least the opening of a lid), but it's great to actually be able to not see OSX draw each and every pixel. Really, if the existing TM hardware were asked to emulate a PowerPC it wouldn't be able to hold more then one CPU state in it's registers, so no real speculatave execution, making it's write guards useless, and it's fast shadow save and restore instructions useless as well. Cutting off whole sections of the functionality, and most of the opertunity to actually emulate a CPU better then CPUs not designed to do emulation.

      Let me repeat that part, because of the differences between it's target system and what it is trying to do pretty much everything put into the TM hardware to help it emulate a CPU is worthless. For speed you would be better off doing the emulator on a P4, or a recent Alpha.

      Going for the biggest market share is not always the point. It is like saying 90% of word for windows is not used, let's cut 90% of word for windows, making it 100% faster etc etc. However not the same 10% is used by everyone.

      But you are not doing a faster subset of the PowerPC here, you are doing a whole PowerPC that is slower then the existing embeded PPC designs, costs more, uses more power, and has fewer integrated parts. The only advantage to your proposed slow PPC is it is also a slow x86 and uses very little power for an x86.

      Plus pretty much only Apple can make a computer run Mac OS (this is a lot less true with OS X around), and I doubt they would try to provide a very slow laptop that can run PC software faster then Mac software. They would either see it as a bridge away from the Mac, or as an amusing diversion almost nobody would buy. As cool as the iBook is, I wouldn't pay more then $500 for a 300Mhz one given what they currently have on the market for $875! No, I won't even buy it for $500, I wouldn't hesitate at $200 though.





      P.S. yes the 300Mhz number is pretty made up, it is around half what people say the x86 feels like. The real number could be a lot lower, but I doubt it will be a lot higher. In fact if Apple is right, it is a lot lower even if they make it feel like a 500Mhz x86 -- but I think they are wrong about the G3 being much faster per Mhz. They are right about the G4 for some tasks though.

    4. Re:If only Transmeta would release a different CPU by stripes · · Score: 2
      So what's the point of code morphing?

      Well for one thing it dances around a lot of Intel's and AMD's patents.

      I mean, if they're just going to be a low-power x86 clone, surely the resources (engineering, space on the die, whatever) taken up by code morphing could be put to better use as...nothing. Just make an x86-only design. AMD did, and theirs is better.

      Maybe, but AMD also spent a lot more on research. Money TM didn't have. AMD's also came out a while after TM's CPU.

      Personally I think TM blew it with their first CPU that didn't run 16bit code very quickly at all, and lost a ton of time doing the second CPU which did run 16 and 32 bit code fairly well. If they had payed more attention to 16 bit code they may have brought out the TM CPUs quickly enough that they didn't seem so slow. Remember everyone else's CPUs were getting faster. Especally right then, AMD and Intel had just finished the race for 1Ghz, six to eight months earlyer and the TM CPU would have be every bit as fast as the notebook CPUs that were out.

      Of corse that's just a thery, and I have no real evidence.

      Seriously. Either develop some useful capability from the code morphing tech, or abandon it.

      Who says they havn't? It did give them a faster time to market, they had a number of bugs in their CPUs that they coded around in the morphing engine (instruction combos that should work, but didn't, the morpher was change to not pair them).

      The code morphing tech's useful capability may well be that TM actually has something on the market, and hasn't been sued by all the other x86 makers for patent infringment. That may not save TM from being an "also ran", but it doesn't make code morphing a bad idea.

      (Also just because TM aimed at low power doesn't mean that's the only place code morphing can be used -- it's not that diffrent from microcoded CPUs, or trace scheduling. With diffrent goals it may be useful to make fast CPUs, or just plain not)

      TM had a lot of obsticles, far more then most CPUs have. Definitly more then the one CPU I designed (in a class, not the real world). Being crushed by them doesn't mean all their choices were bad.

    5. Re:If only Transmeta would release a different CPU by Rasta+Prefect · · Score: 2


      This would require quite a bit more effort. Soldering a PPC onto an Intel motherboard (Even assuming that you could) does does not an Apple make.

      --
      Why?
  3. WinCE only for StrongARM? by class_A · · Score: 3, Interesting

    Taken from webpage:
    "...industry standard x86 architecture, the VIA Eden Embedded System Platform is fully compatible with Microsoft Windows XP and a full range of Embedded Windows, Windows CE..."

    I thought WinCE/PocketPC was now only built for the StrongARM processor, or am I missing something?

    Personally, I don't see low power as being Transmeta's primary selling point. I am much more interested in their code morphing software. I don't see where VIA's solution fits in. If you want a low power consumption PC type device, then are we still talking about an "embedded" device?

    1. Re:WinCE only for StrongARM? by Webmonger · · Score: 2

      Code morphing is cool and all, but what's it for if it isn't power savings? Isn't that their main selling point?

      And can't they think bigger than that? Wouldn't it be cool to have a machine that could run every platform from Windows XP to MAME to Commodore BASIC to PDP-11 Unix? Wouldn't that be a more fitting use of their tech? Sure, nobody would buy it, but wouldn't it be cooler?

    2. Re:WinCE only for StrongARM? by zulux · · Score: 2

      Actually CE was designed for a range of processors

      Microsoft is pusing all it's vendors to use StrongARM. The latest develement version of CE, is only StrongARM.

      NT was also able to run on Alpha et al, but we all know how long that lasted. NT running on an Alpha was kinga like a dancing bear - you diden't judge it by how well it was dancing, it's just was amazing that it was a bear doing the dancing.

      NT on an Alpha was just a waste of a good Alpha.

      --

      Moneyed corporations, non-working 'poor' and criminal prisoners are turning productive citizens into tax-slaves.

    3. Re:WinCE only for StrongARM? by michael.creasy · · Score: 2

      Yes, WinCE != PocketPC. PocketPC uses WinCE and made the decision to only support one architecture, but WindowsCE is built for ARM, MIPS, PowerPC, SH and x86 - you can find a full list of the support processors here

    4. Re:WinCE only for StrongARM? by stripes · · Score: 2
      Wouldn't it be cool to have a machine that could run every platform from Windows XP to MAME to Commodore BASIC to PDP-11 Unix?

      They buy them all the time. They buy "PC Compatibles". They can run Windows, Unix, C=64 emulators, MAME, and some fine PDP-11 emulators which run faster then the real PDPs.

      You really only need to buy CPUs designed to emulate other CPUs to try to emulate current-ish platforms. If you are interested in anything older then about 3 turns of the crank for Moore's "Law" a normal CPU will be fine (well unless you want to emulate historic supercomputers, then you may need 5 turns, or to emulate a mainframe then you may need something with lots of I/O...unless it is even older).

    5. Re:WinCE only for StrongARM? by thing12 · · Score: 2, Insightful
      I know, I think it ran on MIPS as well? ... Anyhow, I don't remember it ever running on x86...

      Actually it runs on 12 processor architectures, including x86 and MIPS.

      Pocket PC is kind of a separate beast from Windows CE. It's basically CE plus a bunch of extensions that make it fit the needs of PDA users better. It may very well only be available for StrongARM.

    6. Re:WinCE only for StrongARM? by Webmonger · · Score: 2

      See, this is the thing: I'm having a lot of trouble envioning circumstances where Code Morphing is actually good for something.

      You can't clone Macs, so it's not useful for budget Mac laptops. Maybe you could use it to run in some vmware-like state so you could run Mac apps on a PC, but that's rarely needed. Most mainstream apps for Mac are also available on PC. The other way around is not the case, but we can't do Mac clones.

      And you can't simulate a Geforce3, since bandwidth's the most important part.

      I guess if you're running Java or C#, it could provide a "native implementation". . .

  4. Take it from the source by pclinger · · Score: 2

    I'd wait before I heard an independent review from someone rather than going off of the hype from a company. I find that independent sources give you the real details on if something truly is a better product.

    --
    /. editors made it impossible to link to file:///c:/con/con in my sig. Please just type it in
  5. TMTA:RIP by Ars-Fartsica · · Score: 5, Informative
    Its fairly obvious at this point that there is little to no commercial support for the Transmeta product line.

    Power consumption and heat dissipation are issues to consumers and manufacturers, but clearly not enough to warrant employing a lower performance architecture at this point. Added to which, it appears that competitors were capable of rolling out competing technology far too quickly - Transmeta never hada chance to get support.

    At this point it seems that the smartest thing Transmeta can do is start shopping its assets around to possible suitors.

  6. So what is Transmeta for, anyway? by DaveWood · · Score: 4, Insightful

    I'm hoping the more clever watchers of the semiconductor industry can enlighten me on this. As far as I can tell, Transmeta has been an expensive and overhyped flop.

    They came out with low power consumption CPUs that, while cool, aren't THAT cool, really (to the point where Intel and AMD immediately responded with conventional laptop CPUs that were in the same spec ballpark), and weren't that fast, either. In fact, when you sit down with them, they're quite slow for the $$$. And that was they debuted - let alone now, in Q1 2002. Their design involved doing IA emulation right above the silicon, which sounds wacky to me; fine, advances in runtime optimization lately are quite interesting (hotspot) but it doesn't sprout wings and fly, and I can't see how we could ever expect it to.

    Then we have the fact that virtually no one sells transmeta-based products, and some significant percentage of the few companies that said they would, have since backed out of the deal (which screams trouble with the product).

    Maybe I'm just too cynical. Yes, everybody loves them because they're competing with Intel and they're a patron of Linux. Please, tell me why I'm wrong about this. I'd love to be convinced their killer app is right around the corner.

    If I'm right, though, they should call it a day, shut down now and return whatever money they have left to their investors...

    1. Re:So what is Transmeta for, anyway? by DaveWood · · Score: 2

      I wonder about this. Especially because someone raised intriguing speculation of adapting TM's code morphing technology to parallel processing in some kind of "original" way. AFAIK the problem with this is always bandwidth. When you start stacking up CPUs returns diminish pretty fast. Some problems adapt quite well to being broken apart into little pieces than can be executed simultaneously. Others are utterly resistant to it. For general purpose computing (i.e. desktop), I expect to continue to see very little parallelism barring a fundamental advance either in architecture or software design, so... for the next 5-10 years at least, I doubt it.

    2. Re:So what is Transmeta for, anyway? by DaveWood · · Score: 2

      Of course, if your point is that they could go after the _server_ market, which _does_ benefit from parallelism quite a bit, then hey, I agree in principle - but then we still have the problems of other CPUs on that market being faster/cheaper/practically lower power... and all of the established players... &c &c ;>

  7. Speaking of laptop power savings: LED backlights? by timothy · · Score: 3, Offtopic

    I'm hoping someone more knowledgeable than me can shed light (ha ha) on the possibilities for white-LED backlights in laptops.

    Certain high-end digital cameras (like the newest Nikon SLRs) have white LED backlights for their LCD displays. White LED prices are dropping (USD7.88 for a nice little waterproof, floating flashlight at Walmart :) !), and power consumption on white LEDs is ridiculously low. As I understand it, the backlight is the biggest draw in a lot of laptops, especially turned up bright.

    So why don't we see some low-power LED-light screens? I'd pay $200 more easily for my next laptop if it got (for instance) 50% more battery life.

    What's stopping those? Considering that there are now several approaches (AMD, Intel, Transmeta and now VIA) to saving power on laptop processors, what about the other powerhogs? :)

    timothy

    --
    jrnl: http://tinyurl.com/c2l8yr / foes: http://tinyurl.com/ckjno5
  8. Where does the "much better performance" come from by vlad_petric · · Score: 2, Interesting
    There is no indication of the actual performance of the Eden processor (i.e., results from real benchmarks ) on the page, only the halucinogenic Highest performance x86 embedded processor (trust the salesman!).

    Moreover, there are some other oddities in the description, like the Integrated 192KB internal L1/L2 cache (well ... what's the size of L1 ? )

    The Raven.

    --

    The Raven

  9. What's with this Eden name anyway? by zensonic · · Score: 2
    • Not allowed by humans?
    • Our processor knows good from evil?
    • Our processor makes you want to be somewhere exotic?

    Anybody know what/if the markting people had any ideas with the name?
    --
    Thomas S. Iversen
    1. Re:What's with this Eden name anyway? by zensonic · · Score: 2

      Think, man.

      Are you crazy? Tried that once (to think that is). Made me feel sick, so I never got the habbit :)

      Seriously, thanks for the answer, made sense. Merry christmas.

      --
      Thomas S. Iversen
  10. Transmeta isn't a Total Failure by nemesisj · · Score: 2

    Transmeta may end up being a business failure, but they will have achieved what they set out to do: delivering low power consuming chips to consumers. The chips just might just might not end up being supplied by Transmeta.

    All in all, we the consumers win. It's doubtful Intel or AMD would have ever considered low power chips had it not been for Transmeta.

  11. Re:Talk about heat generation by tzanger · · Score: 2

    So you're saying that ten transmeta processors put out as much heat as a single P3 card solution? What exactly is the problem?

  12. Re:Speaking of laptop power savings: LED backlight by stripes · · Score: 2
    Because they look crappy?

    They look fine on my PS100, and D30. Nice even light, and decently bright (the PS100 can be seen in fairly bright sunlight, the D30 can't though), and no strong color casts. They seem to work decently in low and high temperatures (the iPod backlight, or maybe LCD doesn't seem to work so well in the cold, and I think it has a LED backlight).

    I would believe "you can't make them big enough", or "can't make 'em big and cheep", or even "they have been using them for 14 months!", but I'm not buying "they look crappy".

  13. Re:x86 at 10W?? by LazyDawg · · Score: 2

    Even the transistors used to draw Mickey Mouse on the die? What are all those transistors used for? Is there supposed to be some one-to-n correlation between "transistor" and "computing element" in Intel marketing?

    --
    "Look at me, I invented the stove!" -- Ben Franklin
  14. Crusoe 6 Watts? by derwagner · · Score: 2, Informative

    Isn't Crusoe below 1 Watt when it's idle, and at 6 Watts at full load?
    Very stupid comparison.

  15. Re:Speaking of laptop power savings: LED backlight by Salamander · · Score: 3, Interesting
    So why don't we see some low-power LED-light screens?

    Very simply, because LEDs aren't powerful enough. They might seem pretty bright when viewed directly, but when you're putting that light through a lossy backlight assembly onto the relatively large area of a laptop screen, and hoping that the result is sufficient to counteract ambient glare, you get a different impression. Frontlights are even worse.

    Some vendors have tried replacing standard CCFLs with LEDs in PDA applications, where the screen size is smaller, and even there it has led to "customer acceptance issues". Translation: customers hated it. For the larger screens that laptops use, current-generation LED technology doesn't even merit serious consideration. With any luck, somebody will earn a Nobel prize figuring out how to make an ultra-bright LED that can compete with CCFL, but I wouldn't count on it.

    --
    Slashdot - News for Herds. Stuff that Splatters.
  16. Cynical? by fm6 · · Score: 2
    Maybe I'm just too cynical. Yes, everybody loves them because they're competing with Intel and they're a patron of Linux. Please, tell me why I'm wrong about this. I'd love to be convinced their killer app is right around the corner.
    Two or so years ago, I would have called that opinion cynical. But now you're just describing how the market has actually responded to the TM chips. There's a painful absence of anything useful based on this technlogy. Just another idea that looked good on paper.