Slashdot Mirror
Transmeta to Release Processor in January?
Scipius writes "German tech-mag c't reports that Transmeta's new processor will likely be released on the 19th of January 2000. It also reveals the apparent code name: Crusoe." The article's in German, of course. But we'll take a juicy Transmeta rumor - and that's all this is - in any language. Babelfish time!
If Transmeta is a fabless chip manufacturer, then this will just be an announcement of what they plan to build. Not that I'm accusing them of vaprware, but early chip announcements tend to be more like bad sci-fi than actual news. Think of Intel and their gHz processor - cooled by a desk sized freezer beneath the unit.
Either way, it will be more interesting than the Lucent (might not be a) router announcement.
Well, yes, we'd kind of known that for some time. :-)
The new-news hidden in this article (apart from the codename itself) is that Crusoe is to be aimed at low-power devices like laptops, which is quite a different market to the mega-workstation many people here wanted it to be (perhaps due to dislike for Intel).
'Course, low-electrical-power doesn't mean low-computing-power. Look at the ARM series, for one. Or c't could be wrong, though they usually aren't. Guess we'll just have to wait and see - to use a phrase already worn out in Transmeta discussion...
--
This comment was brought to you by And Clover.
If it's a processor then some software is going to run on it? This could explained, as mentioned in an article above, why they hired Linus.
:-)
Think about it. A *new* architecture, with some really fancy new characteristics but no software? I don't think so. On the other hand, if they involve the opensource community . . . BINGO! a real competitor to Wintel.
Who else would you choose as youre link to the community but Linus. He's the head of development. He can make sure everything runs on this "Crusoe".
I'm not directly tied into kernel development, but you sometimes wonder why some patches don't get included . . .
But in the end, if it's a processor, Linux will run on it. How else can you have a top-secret processor? Who can they really trust as their OS of choice? Windoze? MacOS? BE? . . .
You never know, kernel 2.4 might as well be ready to run on "Crusoe"
cl
--
Reply . . . let's get it over with
--
Reply . . . let's get it over with.
The fact is that if the Transmeta CPU architecture is anything like what's in the Transmeta patents, and if they can at least come up with a few engineering samples, if will mean a radical shift in our ideas about processor design. As it stands, the instruction set is what defines a CPU--CISC, RISC, x86, HP-UX, etc., are all involved in defining the processors which use these instruction sets, but Transmeta changes this. The Transmeta idea as expressed in their patents would create a category above this--no longer is it 'an x86 processor', it's 'a processor running x86 instructions'. This is a radical idea, and a radical paradigm shift--we should all hope it comes to pass. If it's a great and practicable design, it shouldn't be too difficult for Transmeta to partner up with anyone from AMD to Motorola. This sort of radical advancement--again, if it comes to pass--makes me wonder what the heck Intel and all their capital were doing designing the inflexible Itanium, which executes its native (and sure to be poorly supported except for network/server apps for at least a year or two till prices come down remarkably) instructions with Alpha-killing speed but chokes on anything else including the x86 with which they were supposed to be compatible to some reasonable degree. Just 2cents from a guy who plans on supporting anything but the Itanium (mmmm, legacy games under 64-bit AMD....)
"The more corrupt the state, the more numerous the laws."--Tacitus, *The Annals*
In the short term, it could turn out that the product isn't the fusion-powered anti-gravity time-travel device that all the secrecy has led me to expect.
In the long term, even a fantastic product could end up going nowhere. I'm thinking particularly of AMD's woes. Not only has Intel (allegedly) managed to convince some major motherboard manufacturers not to ship their Athlon boards, now they've gotten a major OEM (Gateway) to drop all AMD processors from their product line. And Intel's anti-trust case inexplicably disappeared into thin air.
Even if Transmeta has the coolest CPU ever, do they stand a chance against Chipzilla? Here's hoping...
11.0010010000111111011010101000100010000101101000
This is why Transmeta is absolutely cool. Notwithstanding people that can read German (and I'm sure you're plentiful), we've got a one paragraph article that babelfishes really poorly about some bizarrely unsubstantiated rumors and it's going to be a VERY popular slashdot thread, because, well, because it's Transmeta. (Circular logic... cool).
As for real content, I'm surprised by even the rumor that the supposed chip would be a notebook chip. Why a notebook? Linus has said recently that Linux is likely to develop towards embedded applications (it really does perform well there). How let down would we be if Transmeta's first chips were low powered, linux-powered embedded app chips? Really think TV/Network Computers or the like...
Also, if I read the babeled German correctly, they're going to announce the Concept on Jan. 19th. It still could be sometime before we see product (whatever it may be). This should come as no surprise, since TransMeta clearly hasn't employed hundreds of chip-builders lately (someone would have noticed that, I think).
I'm waiting to be awed by whatever they eventually produce, but for now, it's enough to be in awe of the amazing hype and free-publicity. Amazing, isn't it that doing the exact opposite of Microsoft (by spending NOTHING on advertising) is garnering TransMeta (and thus Linus, and thus Linux) a decent amount of press?
Keep it up TransMeta!
Isn't this going to be capable of becoming a JAVA OS.
If it stores the machine instruction set in a local cache and runs it from there.. Well that seems to be exactly what JAVA would need to have a Java-Chip.
Considering it's low power consumption it would make an excellent candidate for the jini - appliance environment as well as the PC. Add to that the potential JAVA chip concept and you have the Jini Project from Sun sitting in your lap.
So Transmeta are finally going to be ready to *say* something. The funny thing is that their patents are quite revealing about what they're up to - a speeded up version of the self-modifying FPGA technology that has occasionally spawned 'new era' claims. I'm not saying that their chip is just an FPGA, but that the effect is meant to be much the same: a metamicrocode that can be optimised in near-real time by a JIT-like (or is dynamic compiler a better term than JIT?) compiler and scheduler.
Please though, don't beleive all the speed hype. Remember, it was a year ago or so when 1GHz sounded astonishing, but now it's almost boring for those chiller guys. The thing is going to be *flexible* not *necessarily* fast.
Curiosity killed the cat, but who ever saw a cat reading a patent application?
.sig thingy
As has been pointed out elsewhere (comp.arch I think), the patents that are available on the net reflect what Transmeta was doing a couple of years ago, when the patents were filed. So don't be sure that they are central to their current plans.
Had anyone else heard the rumor about the Transmeta chip being low power consumpion before? I sure hadn't, and to me, it doesn't mesh well with the idea that it can run multiple instruction sets.
Surely this would require a large amount of memory, and isn't (fast) memory something of a killer for low powered devices?
If Transmeta can produce something that emulates other architectures, and uses a comparable amount of power to the low power versions of those architectures, it has to be one of the most impressive breakthoughs ever.
I do worry though - you know what they say -
Okay, I made up the quote, but I think it is slightly accurate at least, esp. in the early generations of a design.
What else.... Oh yeah.
If they are really going to announce this in January (or at Comdex), I don't think we will see it in use anywhere for a couple of year. If Tranmeta had contracts with fab plants somewhere, someone would have said something by now.
I doubt very much if you can go down to your local chip maker, and say "We want you to switch your plant to making our funcky new designs - forget about this multi-billion dollar contract you have", so they can't just get manufacturing facilities like that. It takes a long time to build a fab plant, too, and it's not like you can just convert a derelic factory to a state of the art chip fabrication plant.
--Donate food by clicking: www.thehungersite.com
The Linux connection (strongly hinted at by Jim Collas, former President of Gateway's Amiga division, when I spoke to him about the various AmigaNG rumors) is, of course, Linus himself. But the coolness factor is this: here we have a CPU, running emulated instructions as fast as some version of the Pentium II/III, and doing it cheaper and with much less power. Only the first one is new work -- it's easy to find CPUs that run faster than PII/IIIs using less power (PowerPC, for example). The thing is, the x86 code runs at 1/2 or so of the native Transmeta CPU core speed, perhaps based on the limits of their dynamic recompiler, the loss of abstraction in binary code, etc. What about native code?
The modern trick in all this, same thing Sun's doing with MAJC, is to make the idea of a VLIW processor legit by never tying a system to native binaries. You run x86 or Java Byte Code or whatever through a translator, and when the machine architecture, and thus instruction set, change, you build a new translator, everything's hunky-dorey code wise, and you get to forget entirely about hardware legacy. But Linux and other open source stuff doesn't have a binary legacy problem, period. So there's no problem in coding Linux native, and if you did this, Linux would run twice as fast, relatively, as any closed source OS on this platform. The need for Linus on this would be building a mixed code manager for Linux, so that the kernel could cooperate actively with the dynamic translator and run your choice of x86, native, or other code modules (full-speed Java, etc).
This is the last layer of HW abstration. Some day, all CPUs may be built this way.
-Dave Haynie