Slashdot Mirror
Transmeta set to Introduce Crusoe Processor
senthilp sent us a nice article about Transmeta which basically wraps up all the loose ends hanging out, and sums up the leading speculation on what transmeta will announce. The officail launch date for Crusoe is Jan 19 (next wed) and
Chris DiBona is gonna be there with a richochet on his laptop and hopefully broadcast to IRC what is going down. So anyway, we've only got a few days left to speculate: the leading rumor is a VLIW processor which will be demonstrated in some sort of PDA or Handheld running Linux. But speculation is running rampant: I've heard that they were working on a superior version of The Kernel's Secret Blend of Herbs and Spices, and before that it was a carbon dioxide fueled teleportation device so who knows.
grimsaado sent us details on the Crusoe in another story with slightly more realistic speculation.
Strange, although this may be promising
The Kernel's secret blend, or the Colonel's secret blend?
And at the bottom:
"Connecting People with Information"
Ita erat quando hic adveni.
I wonder what "surprises" Ditzel is talking about with regard to this.
Pentium killer?
I do like the design
Tell us when the chip actually does come out instead of contributing to the hype around it.
:); }return(0);}
Funny, though, something tells me it won't replace my 486 cpu anyway...
#include <signal.h> \ #include <stdlib.h> \ int main(void){signal(ABRT,SIGIGN);while(1){abort(-1)
OFTC: By the community, for the community
I think I remember reading an editorial at a hardware site some months back about the future of processors. The writer said that a software-based solution would allow for better functionality and more flexibility in excution. I thoght it was a pretty interesting article (sorry I can't remember the URL) but it didn't mention anything about any of these chips, well, "Goin' Mobile", to quote the Who.
--------
Oscarfish.com: tropical fish with attitude. Way t
Actually, I'm happy to finally see what Transmeta has been working on, for a while, I thought they just had Linus on board to improve their IPO standing.
I'm very interested, however, in a good Linux based PDA, and Transmeta's delivery of it will be great.
"You ever have that feeling where you're not sure if you're dreaming or awake?"
"You spoony bard!" -Tellah
I was wondering when someone would say this again....I was cleaning out my bookmarks last nite and came across transmeta...and the msg is still in the source (unchanged sadly)....and I say it's the -Kernel-'s secret blend :)
A PDA or handheld running Linux? Can you think of the numbers of geeks who're going to buy these? That just sounds too cool. If it's got enough power to run Emacs and gcc, I wouldn't have to lug around this Dell Inspiron anymore. :-)
At any rate, nice to know the hype is over and we can see what all the fuss was (even if it shouldn't have been) about. But hey, if it runs Linux, it must be good, right? :)
However, what about the other x86 products? I would think they would include those products in the release as well, such as *BSD and Solaris. This would just broaden the market. and of course, these forms of Unix might be able to be ported to the new chipset natively.
Thoughts?
--
Gonzo Granzeau
Gonzo Granzeau
"Nothing the god of biomechanics wouldn't let you into heaven for.." -Roy Batty
hmm I seem to remember a company that did large scale CPU and was the basis for alot of VLIW research and theuy wrote a book about it
Anyone Know what I mean ??
BTW VLIW is not great ARM arch has prover that well designed RISC work very well what will be intresting is the power consumption and Die Size
regards
john
a poor student @ bournemouth uni in the UK (a deltic so please dont moan about spelling but the content)
I wonder how these are going to be priced? I hope well enough to keep them with in reach of the average joe. Maybe I will hold off buying any computer eq until this evolves a little more..
Microsoft aggravates my tourettes syndrome.
I was just going to throw a message over to the slashdot guys reminding them that the Crusoe was going to be released in just a few days.
Hmmm... With that possible target for use, the mentioned low power consumtion, and the really neat trick of reconfigurable, and the article hints, adaptive microcode, you gotta wonder where the compromise is going to be. I wonder how they will compare to conventional embedded processors like the StrongARM and PPC for speed.
I feel a little worried by this speculation. I'd hoped that we would be seeing a new high-end server or workstation chip capable of handling code from multiple platforms. Oh, well. This is still pretty cool -- especially with the hinted adaptive abilities. I'll have the fastest PDA for running Tetris in no time!
If it's for-profit but free, you're not the customer -- you're the product (e.g., the Slashdot Beta's "audience").
And now there really is a secret message on the web site! Check the HTML source. It was the first thing I did...
Baz
The hype and rumors surrounding Transmeta have constantly changed, but everyone seemed sure it would be the greatest thing since sliced bread. Honestly, I think Slashdotters are setting themselves up for a disappointment they haven't seen since the Phantom Menace.
hmm ARM is in the UK
and the UK gov want to give a new lic for more bandwidth to give Inet mobile phones
virgin and MP3
bob
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
At the risk of stating the obvious, employing Torvalds was the best move Transmeta could have made. How many millions must he be worth to them in free advertising, before they even have a product?
11.0010010000111111011010101000100010000101101000
The graphics on the www.transmeta.com thing is made with Photoshop 3.0 :-)
As usual... there's a secret message in Transmeta.com's source(html)
"Yes, there is a secret message, and this is it:
Transmeta's policy has been to remain silent about its plans until it had something to demonstrate to the world. On January 19th, 2000, Transmeta is going to announce and demonstrate what Crusoe processors can do. Simultaneously, all of the details will go up on this Web site
for everyone on the Internet to see. Crusoe will be cool hardware and software for mobile applications. Crusoe will be unconventional, which is why we wanted to let you know in advance to come look at the entire Web site in January, so that you can get the full story and have access to all of the real details as soon as they are available."
Ok, how to you build software for these babes? What sort of development tools are going to be available? GCC probably can't do VLIW very well. I bet they have their own custom compiler. I wonder if it will be free, or if there is going to be a steep ``initiation fee'' requiring a major investment in tools.
Given that it's called Crusoe, I think it's most likely that the processor will be used for rescuing people from desert islands, so it's probably some kind of GPS device!
"Destroy science and religion. Science would re-emerge exactly the same; but not religion." - Penn Jillette, paraphrased
In my opinion there's no contest: If this thing is for real (which I pray to God it is) then Linux might chew Windows up and spit it out. Being a user of DOS-similar apps for 13 years, I simply can't WAIT when Linux becomes standard and everytihng else simply GOES TO HELL. But if this company CAN challenge the AMD Athlons and Intel Pentium III's, then things are going to get very interesting in the way of price wars and the availability of low cost copmuters.
But what about me? I own an 80286. Yes, you read that right, an 80286 that runs Linux. I use it as a network hub, and it whoops ass (I have better). Linux and Windows both have their advantages, but for some reason, I think Linux is just more dandy (Actually, I'm partial to MS-DOS 6.2, but who cares?). Of course, you must remember, that if Linux DOES become a standard, it is my belief that Red Hat will lead the way. Why do I say this? Cause anything is better than Gnu. ^_^
My 80286 is like the Bible: I swear by it every night when I try to run something.
Given a GCC code generator for "native Crusoe," or whatever they'd call the not-involving-emulation instrution set, it ought to be more sensible to run Linux natively on the chip, as that should be faster and more efficient since it takes direct advantage of the CPU's facilities.
That can be made more adamant if we're talking about PDA applications where it's likely that applications are "embedded" and where it's pretty certain that source code to the applications would be available.
If you look at some of the major proprietary applications, it's still the case that they may be recompiled for alternative architectures. Applixware is available for Alpha as well as IA-32. Mozilla is getting deployed on various architectures. StarOffice has been available for PPC. That may not represent an exhaustive list, but a PDA is not likely to have an exhaustive set of software installed on it.
If you're not part of the solution, you're part of the precipitate.
... is that whatever it does, it uses small amounts of screaming-fast DRAM. Transmeta has been pretty busy in the DDR SDRAM standards process, and their focus has always been on single-bank-to-controller applications at insane speeds (like 400 MHz!)
BTW, having followed Transmeta in the news for quite a while, I agree with some of the various article writers that Transmeta has all of the ingredients it needs for huge success: strong financial backing, partnering with established companies a focused product, a hugely competent technical staff, and enough name recognition to get the media attention required to explode into production.
Hmmm. Interesting thought. I wonder if /. has enough karma points with Linus to get CmdrTaco and Hemos up close and personal on launch day... Think I'll send a message with that suggestion to the good folks in Transmeta's Business Development department. There. Done. BTW, anyone care use the link and make the same suggestion?
Rob, Jeff y'available?
...Open Source isn't the only answer -- but it's almost always a better value than the alternatives...
And of course, it's also a hardware JIT for Java, which fits the oldest rumours about Transmeta... that may have been their original plan before they decided that Java wasn't taking off fast enough to bet the company on.
its hard to compile linux without gcc
this has brought about nice mails from the egcs crew and linus
its possible though
remember they dont have to release the code until its not dev anymore
well
have fun
Based on what I've heard from inside sources: the FT article is dead on the mark. Listen to the logic: there are processors designed from the ground up for high-powered workstations (Intel Itanium, Sun SPARC, DEC Alpha). There are processors designed from the ground up for personal computers: AMD Athlon, Intel PIII/Celeron, etc. There are processors designed from the ground up for embedded computing: StrongARM, MAJC, etc. Now, please, name me one processor designed from the ground up intended for laptops. You can't, can you? Intel and AMD retrofit their desktop processors for laptops. Every other component is now specially designed for laptops - think IBM harddrives, LCD displays, etc, etc. The end result is that the CPU is the (physically) biggest item on a laptop MB and it consumes the most power. Transmeta Crusoe, the first processor x86-compatible processor designed specifically for the laptop, will allow laptops to be smaller, lighter and run for longer all with less battery power. The logic is infalliable. The market is untapped. If their processors are fast enough, in 2 years Transmeta may be the Intel of laptops CPUs.
Based on what I've heard from inside sources: the FT article is dead on the mark.
Listen to the logic: there are processors designed from the ground up for high-powered workstations (Intel Itanium, Sun SPARC, DEC Alpha).
There are processors designed from the ground up for personal computers: AMD Athlon, Intel PIII/Celeron, etc.
There are processors designed from the ground up for embedded computing: StrongARM, MAJC, etc.
Now, please, name me one processor designed from the ground up intended for laptops.
You can't, can you? Intel and AMD retrofit their desktop processors for laptops. Every other component is now specially designed for laptops - think IBM harddrives, LCD displays, etc, etc. The end result is that the CPU is the (physically) biggest item on a laptop MB and it consumes the most power.
Transmeta Crusoe, the first processor x86-compatible processor designed specifically for the laptop, will allow laptops to be smaller, lighter and run for longer all with less battery power.
The logic is infalliable. The market is untapped. If their processors are fast enough, in 2 years Transmeta may be the Intel of laptops CPUs.
I submitted this too, but someone beat me to it
A couple of days ago, a friend of mine that works for IBM told me that he is working on a project with Transmeta. He said even Linus Torvalds is working on this project. When I asked him what he was doing, he told me "cool stuff" with a smile. He is also under NDA, so he couldn't give me any more information but it will all be release to the public soon. With the article stating "unspecified partner", putting 2 and 2 together, is this partner IBM?
Steven Rostedt
Steven Rostedt
-- Nevermind
...and Chris DiBona is gonna be there with a richochet on his laptop and hopefully broadcast to IRC what is going down.
Anyone have any clue where he'll be 'broad'casting this? I'd like to set up an 'automated client' (read bot) to fire it off to other networks (EFNet).
Perhaps it's me, but besides the high geekfactor (and that's a factor you can't always ignore ;-D), what's the use of a multiuser handheld, a handheld or other small device looking like it running linux? :) To me: not an aweful lot.
Perhaps it's not the linux history of Thorvalds what brought him to transmeta but the technology knowledge he has.
Never underestimate the relief of true separation of Religion and State.
They aren't looking to just go after Intel but rather will have a surprise announcement of Multi processor Crusoes being launched by Apple running OS X server and emulating PPC faster than Motorola can make the chips run. 800 Mhz dual processor boxes running a *BSD variant that is easier to administer and cheaper than equivalent NT boxen (no CALs guarantee a price advantage).
The tip offs? IBM, a Motorola rival producer of PPC chips is producing the Crusoe but IBM's own PPC chips would compete with Crusoe if it were truly an embedded only chip. There's no advantage for IBM to do that.
Apple also plays into this with its stealth update of Mac OS X server to 1.2. With delays on other fronts (Pismo, the P7, Mystic) it would have been a natural to hype the upgrade to their server software at MacWorld SF but they didn't go for it. The stock price was hammered because of the lack of news.
A Transmeta tie in would give Transmeta an instant market for chips in a controlled hardware platform that has a good reputation for quality. This would ease Apple's supply problems for chips since they would suddenly be able to go to Crusoe whenever Motorola fell down on the job. It would also reinvigorate Apple's stock price by showing a business strategy that's been assembled piece by piece for the past year+, hidden in plain sight. It's just the sort of 180 degree coming from nowhere gotcha that Steve Jobs would love.
DB
Ask yourself: what is different about the assumptions of hardware computer architecture in an era of open source?
Answer: you can presume that the user has not just the binary code but the source code
Premise: programmable logic can optimally execute algorithms for which source code is available
Impliciations: Rather than using pre-defined, general-purpose execution units, the proper OS could compile open source directly into linked transistors on a programmable chip (e.g. Crusoe) in a JIT-like process when programs are loaded for execution. By more efficiently using transitors, applications would take less power (like Crusoe rumors,) and the one-time-each-runtime compilation burden would make this design most useful in single-task environments (like the embedded tasks in Crusoe rumors.)
Is this interesting enough that I should elaborate further?
--LP
Perl
If their processors are fast enough, in 2 years Transmeta may be the Intel of laptops CPUs.
Not just laptops though. If the chips are as small, fast, and energy-saving as described in the ft article then you will see them in just about everything that will be performing computing and IP functions in 2 years.
Expect this to include not only cell phones and PDAs (of course), but also watches, cars, TVs, cable-boxes, home phones, MP3 players, gameboys, etc.
Work for Change & GET PAID!
Security on a networked device, from separating out root, users, daemon groups, etc.
Having a faster/cheaper CPU does not establish anything about superiority of OS platforms.
There are any number of pieces of Win32-only software that people have, unwittingly, married themselves to, where it would take a messy divorce to extricate themselves. This list might include:
There may exist vaguely analagous software on Linux, and more and better in progress. That doesn't mitigate the messiness of the "divorce."
If Transmeta were producing a chip that provided support for things like:
- Segmentation in the style of Multics
- Hardware-supported garbage collection, useful for either JVMs or for Lisp variants
- Something otherwise better than merely expanding chips from 32 bit addressing to 64 bit addressing
I'd agree that there could be something fundamental here to, in the long run, change computing.When all that it seems to bring is power consumption reduction, space reduction, hopefully faster performance, and perhaps less damage to the wallet, that's not really news. Every generation of CPUs since the 8008 has provided some mixture of those improvements.
Unless "Crusoe" offers more than that, I can't get overly excited.
If you're not part of the solution, you're part of the precipitate.
You don't get it:
StrongARM is small enough and low power enough to be used in handhelds. When determining if a product will impact on a market the question you must ask yourself is WHAT IS THE COMPETITIVE ADVANTAGE?
What is Crusoe's competitive advantage vis a vis StrongARM? It can emulate the x86.
Is this important in handhelds? NO! What would PalmPilot gain from x86 compatibility? Running MS-Windows apps? On such a small screen? Yeah, right! Handhelds assume incompatibility w/ desktops.
Is this important on laptops, hell yes. Laptops assume compatibility w/ desktops.
I'm not saying that Crusoe will have SOME impact in handhelds, but it's biggest COMPETITIVE ADVANTAGE is in laptops. Hence, it's biggest impact will be in laptops.
Learn some market analysis, buddy.
Jeez - if this is yet another CPU I am going to be extremely disappointed in Linus.
With Intel's move of making the Itanium wiretap ready, we NEED open-sourced CPUS!!!!
There are chips designed for laptops. They are called PowerPC chips :) Motorolla/Apple/IBM have always wanted to design two chips in parallel, one for the power desktop, on for use in low cost, low power consumption, low heat generation situations. The Second Generation PowerPC chips were the 603 and 604, where the 603 was the ideal laptop chip and the 604 the power user chip. The Third Generation PowerPC chips were developed the same way, but the low power consumption model was so damn good Apple never used the other, hotter, chip. (This is why the G3 chip was used in Laptops from the get go and why Intel always has to wait about 6 months after doing a new chip before they can get it small and cool enough to use in a laptop.) I don't know what happened with the fourth generation powerPC chips that only the G4 is in widespread use. ---matt wecksell
I'm not sure if it's been mentioned before by Linus himself or elsewhere, but why does a project involving Linus have to involve Linux? Is it possible that he could be working on a customized o/s for an all new platform? Is it possible that he might be working on other things altogether?
This is not a flame and I'm probably misinformed, but I havn't seen any references explicitly stating that Linus is working on an o/s and for that matter, Linux.
Furthermore, judging by the information on the patents filed by Transmeta, this appears to be a radically different processor design. This might need an equally radical approach to software development, especially in the compiler and o/s departments. Maybe Linus is stretching his wings at Transmeta and doing something completely different.
-- kwashiorkor --
Pure speculation gets you nowhere.
-- kwashiorkor --
Leaps in Logic
should not be confused with
Jumping to Conclusions.
To quote someone: "post or moderate? Post or moderate?"
Well, I'd have to disagree with you there, Norm.
I've heard that the LCD is the single biggest power hog on there. But you're right, I'd like to see a less power-hungry cpu. But I don't need no stinkin' multimedia device.
What I'd like to see is someone manufacture an older cpu (as I can't really justify needing more than a P166 (if even) equiv in my laptop, and would gladly trade battery life for speed. I suspect that many would agree with me) using modern fabs, thus making them small, cool, and power lean.
Make the back of the display white/detachable/translucent, so that I can use ambient light when availible.
Basically, if you look at the triumvariate (hoping I'm using the word correctly) of display, speed, and batterylife, the display and batterylife have been mostly constant througout the evolution of the laptop, whilst speed has steadily increased. I don't want that. I'd rather see steady increases in batterylife, and only incidental improvements in the other two.
Except that the screen drains a pretty hefty part of the battery and that the StrongARM you mentioned is perfectly suited for a laptop.
Transmeta needs a fab; IBM is a logical partner. But--a thought: IBM would have been in this partnership for a year or more; would knowledge of the coming Transmeta products have been a reason that IBM has gone all out for Linux?
1000 SlashDot sigs
The best and worts thing that Transmeta has done is not tell anyone what they're doing.
They're creating incredible hype generated by people who expect the best thing since the first protein, which gets Transmeta's name around in preperation for both product release and IPO.
But when they do release what they are going to release, it can not possibly be as good as everyones expectations which will at least in some way hurt the company and their image.
What they are going to release will be great, from a certain point of view, but don't get too angry at the company if you've been misled to think that it will stop the world from turning.
What I think about when I see the combination of cheaper processor and order-of-magnitude less power consumption is: SMP laptop. That's what I want as a software developer. Sorry to all you who say that we don't need faster/stronger CPUs; I say 'until I have instant compiles, it's not fast enough' (meaning disk, bus, memory, CPU, the whole lot. I'm GREEDY!).
Anyway, think about it. 10x less power consumption. To me that doesn't mean I work on battery for 10x longer, that means I can have up to 10 CPUs with the same power draw (& heat, too? I'm not too certain if those necessarily go hand-in-hand).
Furthermore, since (as other people have pointed out here) this chip wouldn't theoretically be limited to the x86 CISC instruction set, couldn't these processors be tailored for graphics/sound/whatever processing? So what if we have a computer with 8/16/64 processors in it, and dynamically allocate them to take care of the different processing jobs you have: graphics, sound, and data processing. In graphics intensive/sound intensive, use 40% of your CPUs for data processing and 60% for DSP (video and sound). Or some more optimal mixture.
This is all just hand-waving, but MAYBE it'll come to pass. Or is this just total non-sense?
---------The early bird gets the worm, but the second mouse gets the cheese.
The recently announced Oxygen Project at MIT refers to the RAW processor as part of its design. The chip will rewrite its internal wiring using logic gates and microcode, compiled in some HLL. Sounds like chips are getting so big that it isn't worth the effort to customized them anymore. Just install FPGAs in everything and change the configurations as needed.
Something still makes me think that the fact that CRUSOE is an anagram for SOURCE had something to do with picking Crusoe as the 'code' name.
Perhaps they're open sourcing processor/chip development.
--
'...let the rabbits wear glasses...'
Y2035/38 consulting
In the February Popular Science, they said that the new Amiga machine had a TransMeta chip in it. Popular Science named no specifics, however, I think that this must be the chip they're talking about.
Phyrkrakr
"God doesn't play dice"-Einstein
Psychic spies from China try to steal your mind's elation.
Making worshipful cow eyes to the effect that it would be just TOOO SWEET! doesn't establish anything about one's ability to do that. The matters patented, such as Memory controller for a microprocessor for detecting a failure of speculation on the physical nature of a component being addressed (US5832205), might be helpful in making an emulation a bit faster, but does nothing to support the software side of the matter.
The only way your theory is "prime possibility" is if Transmeta has actually been working with developers of something like VMWare. If they have, then there could be some synergies.
If not, then all we have is a somewhat faster chip. Plus, of course, a bunch of members of the Cult of Linus, all worshipful, but with no actual merit to their worship...
If you're not part of the solution, you're part of the precipitate.
"The Web pad will have been designed with an unspecified partner, those sources speculated"
aren't the over-the-top specs people have been whispering about the new apple handheld just begging to be running on top of a transmeta chip?
hell, darwin's based on BSD, why not a new Apple PDA based on linux?
- Entertaining Bits from the Ancient Kernel Tree
All this speculation is worthless. When January 19 comes and the world receives its very first Mobile Internet Toaster, maybe you'll understand.
The only reason you care anything about the unveiling of yet another 'innovative' processor is because Linus is there. Be a rebel, think for yourself; if a toaster company pulls a media stunt, treat them like they didn't. Beat the system
With wireless IP (ricochet) and CDPD... it could be a pretty cool world... code up a paging/alert type app that monitors your servers and stuff... or whatever...
Crusoe may be pretty slick and fast stuff, but is merely hardware. It doesn't, simply by sitting on a PC board emitting heat, solve software problems.
IBM picked Linux because they figured they could sell software services out of the deal. I don't see Crusoe in that picture.
If you're not part of the solution, you're part of the precipitate.
The first patent doesn't seem like a big deal, at first. Translating x86 instructions into other instructions is what every Intel processor has been doing since the PPro. In fact, the PPro includes microcode that handles this conversion, allowing dynamic bug fixes by re-decoding x86 insts to not use the buggy native RISC inst (but only like 3 people on earth know how to do it, due to fear of microcode viruses).
;)
However what would be cool is if the translation layer was general enough to allow it to run many different instruction sets, perhaps even simultaneously. You could run apps compiled for any platform on the same machine at the same time. Finally, every geek's dream come true: The ability to run WinNT and AIX on the same machine at the same time!
I suppose VLIW in itself is interesting, more as an anomaly than as a technological marvel.
I guess we'll see on the 19th, but for the meantime it's more fun to guess.
The enemies of Democracy are
Intel would shit bricks if that ever happened. LOL.
Can you imagine what would happen if Linus did that!?!?!?! Wow! He'd have hit Two fantastic homeruns! And have changed the entire computer industry more than anyone else!
Mips/Watt wise at any rate. PPC cant compare with (Strong-)ARM SuperH etc in that respect.
Yes! "Plunge your very own 9.6 volt Sperm Whale to the inky depths of the pacific ocean!"
You buy stock. Stock goes up on vaguely Linux related news. You sell stock. You roll in dough.
Think about Crusoe the character. Stranded on an island, he had to make due with what he could find. He used what was given to him, he turned things into tools and useable stuff. I believe crusoe will be a small pentium2 class CPU that can be software configured dynamically to handle a variety of platforms, it will make due with what you give it. It will be very small and very power effecient, designed with the goal of being incorporated into ANYTHING AND EVERYTHING. No doubt a special embedded linux type thing will be the thing to run on it. It will probably have the ability to use multiple CPUs in large arrays, so that items needing more powerful CPU capabilities can incorporate several of the processors into them. Transmeta, think about what that name means. Trans = change. Meta = everything (the big picture). Transmeta is going to change the world.
good joke
Dataflow has a very bad name so they dont want to call it that but it is IMO :) BTW read this line and try to pick out the FUD "Raw chips will be able to achieve clock frequencies on the order of 10 to 15 gigahertz by 2010, compared with frequencies of about 500 megahertz for today's microprocessors" IMO their design has too small a granularity for the desktop, its more suited for number crunching. One size does not fit all.
Dataflow has a very bad name so they dont want to call it that but it is IMO :)
BTW read this line and try to pick out the FUD "Raw chips will be able to achieve clock frequencies on the order of 10 to 15 gigahertz by 2010, compared with frequencies of about 500 megahertz for today's microprocessors"
IMO their design has too small a granularity for the desktop, its more suited for number crunching. One size does not fit all.
We've been playing around with software CPU emulators for decades, Java being the most recent example: we'll just have everyone compile to an imaginary CPU that'll be emulated everywhere!
Too bad Sun didn't see that a platform was more than the just CPU. Transmeta just might.
Don't reinvent the wheel! Why not refactor the common stuff into an emulator accelerator?
It's like the 3D hardware revolution for the last couple of years. We finally went from experimenting through software, to accelerating known algorithms in hardware. Hardware acceleration of Windows 2D video calls used to be the best example of this.
To paraphrase an old programming adage, "Handle the repeated stuff in hardware, handle the differences in software."
If everyone else kept really quiet about product development and maintained the level of secrecy Transmeta obviously have, then they would have shouted their faces off about what they are doing right now?
Just wondering...
Check http://www.pixelfusion.com/
They are applying the same idea commercially.
A custom circuit can be from a couple of times to a couple of magnitudes better than FPGA, in performance and power usage, that is not something wich is acceptable just yet. You need a good mix between development costs flexibility and efficiency, and it does seem that part reprogrammable chips are where its going to be at. (mainly because of the ridiculous costs of fabbing, if some revolutionary technique lowers that cost a lot we are going to be right back to custom circuits for a while)
Another point is that IBM is so large and complex (incoherent?) that its different divisions often have competing products. E.g. AIX, OS/2, Linux and Windows NT are all sold by IBM.
perl -e 'fork||print for split//,"hahahaha"'
It cant compete in Mips/Watt, .18u StrongArm is going to devestate the embedded market... Crusoe better be .18u too, otherwise they will need a hell of an architectural edge to beat out the process gap.
Every single synchronous digital circuit in existence can be reduced to a "cloud" of asynchronous logic (i.e. some arbitrary many-to-many mapping), followed by a clocked register. The register contains the state of those outputs.
Cascading many of these together makes for fast, pipelined circuits, as long as the "clouds" are small enough for their total propagation delay to be less than the clock period of the registers (plus some setup and hold time, etc.)
An FPGA is a whole matrix of logic cells feeding into what can arm-wavingly called "routing logic" - in that sense the previous poster was right. But the important thing about FPGAs is that each cell (which is used to implement logic functions) isn't conventional digital logic at all (as you find in PLAs or PALs, for example, which implement AND-OR trees), but an SRAM lookup table, typically of 4-bit width.
In other words, by precomputing and filling the 16 outputs of a 4-bit-wide SRAM you can implement ANY arbitrary logic function! You simply read from the truth table, and instead of trying to optimise it using logic, you put it in memory!
The advantage of this is (1) constant speed (every function is constant time) (2) extremely regular (=dense) layout and (3) fairly easy to implement for something like a VHDL synthesis system.
Hope this was of some use ... reply if you want to discuss/argue anything.
1) He's a Linux geek, and they're using Linux.
2) He's well-known, respected, and loved.
Jus' my opinion, moderate down at will...
- IBM has an x86 license thereby providing patent protection to a potential x86 clone (or in this case emulator)
- coppper - they pioneered it - by now it's probably not much of a premium
- they are pretty much on the cutting edge for process
On the downside they probably can't compete on price as well with the Taiwanese or Koreans, and a companyOf course TM could be doing the chipless-fab business model where they don't build or sell the chips - just design them and sell the software to make them run
"Software based processor" sounds a lot like the winmodems: they cut things in hardware to do them in software and the result is not good. Is this the same thing?
:-)
That was the impression that i got but i'll be glad to be wrong (at least this time
"Now you can see that evil will triumph, because good is dumb!"
Yes, there is a secret message, and this is it: Transmeta's policy has been to remain silent about its plans until it had something to demonstrate to the world. On January 19th, 2000, Transmeta is going to announce and demonstrate what Crusoe processors can do. Simultaneously, all of the details will go up on this Web site for everyone on the Internet to see. Crusoe will be cool hardware and software for mobile applications. Crusoe will be unconventional, which is why we wanted to let you know in advance to come look at the entire Web site in January, so that you can get the full story and have access to all of the real details as soon as they are available.
The rumor I heard is that a company named Xybernaut has an NDA with transmeta.
Fact: [that has been checked] IBM has invested Xybernaut.
Long-standing rumor: has been that IBM will be (one of?) Transmeta's fabs.
Another rumor is that Tuesday Linus will be seen wearing a wearable computer (At the Auto show?).
Fact: Xybernaut makes wearable computers.
Fact: Wednesday Transmeta will go public DEMONSTRATING their CPU in something(s).
Rumor: demonstrating their CPU in something portable like a PDA.
Fact: xybernaut stock has gone up 35% in the last two days
Conclusion: Xybernaut has a wearable computer using transmeta's CPU, that will be fully disclosed on Wednesday.
As far as being a software based CPU (ok maybe not based but you know what I mean)... will it suffer the same problems as a winmodem?
"...I'm just wondering... How the hell would you use emacs with a handheld?"
Have you heard of pen computing? Anything the GUI doesn't provide can be implemented as scribble-type input. As for chording, that's why you have shift-states, command strokes, etc.
And really, how much Emacs functionality do you suppose -wouldn't- be covered by 1: a command stroke equivalent to CTRL, and 2: one for Meta-X? (OK, you might need one stroke just for Meta or Escape itself... But not much)
Windows boxes are appliances. Linux boxes are tools.
---GEC
Bow-ties are cool.
are they(transmeta)publicly traded? if so, anyone know their symbol?
I just don't get it - what's the point in getting Slashdot source anyway? It seems as though it's a very well put together system, but did the Slashdot creators make any substantial pieces of software for the implementation of the system that would be useful in other contexts? Apart from the fact that it's been refined to a nice level of polish, is there anything particularly challenging about its implementation? What exactly would someone do with the source if they had it? It seems to me as though Linux users already have plenty of web tools available to them - Apache, various databases, PHP, GIMP, etc. - If you can't make something out of that, maybe you should try something simple, like "Welcome to my homepage, here you'll learn all about me, but bear with me, it's under construction..."
---GEC
Bow-ties are cool.
Okay, so it wound up being news in some circles... Now, why do you suppose Transmeta hired Linus? I wouldn't discount the fact that he's popular and well-recognized in various circles, but don't you think he has the skills to pull his own weight?
Think of it this way - you're looking for a job - something to take care of the bills and family, you need to write your resume. Do you suppose they'd be impressed by a line like
PROJECTS
Initial developer of the LINUX operating system. Served as project leader of an international group of developers through the continued refinement of the system.
Whether or not they'd even heard of Linux? I think they would, here's why -
1: It's one hell of a project, no matter how you slice it, and it's only more impressive the more the prospective employer researches it. This plays into the Linux name recognition thing.
2: It demonstrates that he knows a lot about programming, PC architecture (as well as other architectures, special hardware and configurations, etc.) - in the interview they'd probably ask him how the project started, he could tell them all about Minix, early experimentation with protected-mode programming and context switching, leading to the earliest releases and growing interest.
3: It demonstrates that he has the discipline and management skills to establish a project like this, build it up, and then coordinate the other volunteer developers through its continuation.
So could it be that Transmeta hired him because they knew he was skilled, competent, and able to work and play well with others? I think so.
---GEC
Bow-ties are cool.
Okay, I had this idea when I first read Transmeta's patents. In fact, it's all old hat. The challenge is in the real processor, VLIW, EPIC or whatever.
If you have some device like a FPGA, reconfig is going to take some time. Although there have been advances in that field, I have a very cheap solution to it.
All that the hardware must assume is the quantum of the pre-emptive task scheduler in OS. I think you can make a translator chip that reconfigs within a quantum. Then just double buffer the translators. Since the next quantum can be known beforehand, reconfig the idle translator while the other is operating. Then switch translators when the quantum is done. The instruction set for the upcoming code stream will be ready. Voila!
That way, I think you could do all sorts of emulation tricks, plus an array of DSP, 3d, etc. features easily. I won't elaborate further, but the implementation should be obvious, (you need a *bit* of software support to take advantage of this design...)
exa
--exa--
The low power consumption is also an essential part of a product they are developing. Lower power equal less heat, no fan require in a closed, compact device is also a big selling feature. Imagine if you can have a intel PII 550 class small hand held pad that can run application well. Drop in a wireless module for networking and you have the best toy in town that every geek would like to have one.
- Etam
They are slow and burn a lot of power.
Asynchronous logic uses no clock. Supposedly a StrongARM CPU has been fabbed based on an asynchronous design but is not yet available commercially. It may be the case that the Crusoe is the first commercial CPU based on asynchronous logic. For more information check out the asynchronous logic Home Page.
One of the patents also reveal that the device has larger than 32-bit addressing. In a specific code sequence, they compare the -- otherwise unspecified size -- registers to check if it's outside the 32-bit range. That also may mean a bit closer cooperation with AMD to unify the x86-64 efforts.
Another fact is that at 4th Sep. 1999, Steve Chamberlain of Transmeta, has submitted a set of changes to the assembler of the Cygnus binutils to support the picojava architecture. Presuming this is not his lone interest, Transmeta should have a need for a picojava developer tool chain...
"Ten years from now, they could do it in a few seconds." -- The Racketeer of the Hellfire Club, 1993, Phrack 42
Wow, everyone waits for the 19th.. This tag found on the transmeta site makes yummy thoughts.. ;) Yes, there is a secret message, and this is it: Transmeta's policy has been to remain silent about its plans until it had something to demonstrate to the world. On January 19th, 2000, Transmeta is going to announce and demonstrate what Crusoe processors can do. Simultaneously, all of the details will go up on this Web site for everyone on the Internet to see. Crusoe will be cool hardware and software for mobile applications. Crusoe will be unconventional, which is why we wanted to let you know in advance to come look at the entire Web site in January, so that you can get the full story and have access to all of the real details as soon as they are available.
It seems obvious to me that the whole reason they hired Linus is to port the Linux kernel onto their architecture. They could hardly run Windows on it since Win98 doesn't even run reliably on half of the x86 motherboards out there. However the Linux kernel is small enough and "knowable" enough to make a reliable port feasible (especially in Linus' case since he's supposed to have reviewed every single line of code in there). And ask yourself: How much are they paying Linus? And how mnay people would it have taken (and how much would it cost) to port any other capable OS onto Crusoe?
Consciousness is not what it thinks it is
Thought exists only as an abstraction
I reckon one aim is to run the Tao virtual processor natively. It's like a Java VM, but for any language. It has the potential to achieve the write-once run anywhere promise that Sun failed to deliver. Pretty cool. Also, Tao have been linked with Sony and Amiga, who have both been linked with Transmeta.
Choice of masters is not freedom.
not to mention squishdot, a Zope/Python based slashdot-workalike implementation. The main reason Slashdot itself isn't using it ias because i) Rob knows Perl, not Python (although python is extremely easy to learn)
ii) the message database wuold have to start again, from scratch. All the previous stuff would be in a different format.
With Linus working at Transmeta, and Linus recently having submitted patches to the FreeBSD kernel (http://www.freebsd.org/cgi/cvsweb.cgi/src/sys/i38 6/i386/mplock.s), isn't it not a distinct possibility that Transmeta is using a BSD? Would it be because of the commercial appeal of using a BSD-style license?
That figures. Their patents talk quite a bit about caching translated code. Since the size of the translated code likely is rather large (think VLIW...), using normal caches would probably be quite a bit too expensive, hence the relatively fast DRAMS...
print "yet another p{ython,erl} hacker\n",
The Trasmeta chip is much crunchier that all it's competitors and it is able to soak up an hither to unimaginable amount of dip.
W S B Transmeta -- it goes yummy in your tummy...WSB
Thought someone here might be interested. I'm usually a lurker, but figured this was safe to post. http://www.zdnet.com/zdtv/zdtvnews/features/story/ 0,3685,2119139,00.html
Here is my thinking:
1. You can software instruction sets
2. People are talking like it is very dynamic
So, that got me thinking. I wonder how this would work with video game emulation? I can imagine having the instruction sets for a Sega Genesis, Super Nintendo, Sony Playstation, Nintendo64, etc on my system, and getting quality video game play even with games for the high end system.
I'm sure the video game industry is just gonna love that idea (sarcasm!)
If you see problems with this logic, let me know...
Dan
What does all this mean?
I speculate that Transmeta will release, under the GPL, a translator program that translates [x86/PPC/6800/PA-RISC/SPARC/whaddevah] instructions to native Crusoe instructions. But they will not actually release the specs to the instruction set. This will mean that in order to determine the native Crusoe instruction set, a programmer will need to have "consulted" the GPL'ed program, thus inheriting the GPL for her/his code. Not a problem for GCC, but it may be a problem for proprietary compilers. Proprietary programs will still be able to run on the chip, but not at "native efficiency", only via the translator program. So we now have a chip specifically optimised for Open Source software.
I started out thinking I was joking, but now I'm not so sure. I suspect the GPL would not work exactly as I have described in the above scheme, but maybe a similar licence would. It certainly would be unconventional wouldn't it?
Ahh - My eye!
The doctor said I'm not supposed to get Slashdot in it!