Transmeta Webcast Today at Nine PST, Noon EST 345
Almost everybody in the world wrote in with something like "Transmeta's big Crusoe announcement is today, and it'll be Webcast live on ZDTV starting at 9 a.m. PST." Later, after all the hoopla is over, we'll have a follow-up story on the proceedings as viewed through the eyes of several people who are there.
Re:Real Player through Firewalls (Score:1)
But that doesn't help us guys that have MS-Poxy server to get through as well (assuming it hasn't crashed that is
Any concern for virii on windoze? (Score:1)
Listening to Broadcast (Score:1)
For a demonstration, they had Linus Torvalds playing Quake on Linux (on a 667 MHz Crusoe laptop) against a Quake programmer playing Quake on Windows (on a 667 MHz Crusoe laptop). That was pretty cool.
I'm only watching in bits and pieces, but if I get some more time later, I will update this.
ZDNET loves MS (Score:1)
Re:ZDNet and Real (Score:1)
Re:Is there a way to get a transcript or a video t (Score:1)
- The TV link won't go through the firewall.
- The Radio link gives an error as if the DNS lookup fails, and reports I'm not set up for cookies (although I allowed all cookies for it).
- The Chat just gives me a blank spot in the middle of the screen.
So much for watching history live.Re:x86 compatible == "Internet compatible"?!?!? (Score:1)
Well, without ActiveX support, you're SOL on a lot of pages (mostly Microsoft pages, maybe, but software updates are important). The fact is that most PC users run Windows, so Windows must be supported. There's nothing preventing PowerPC from being supported as well.
Not true. What about the Power line (G5 [S/390], etc.)? I'm sure there are others (embedded chips, etc.), but I can't name them off the top of my head. Picking nits, maybe, but it irked me. :)
--
Direct URL for RealAudio (sound only) (Score:1)
Enjoy!
ttyl
Farrell
Re:transmeta chip specs (Score:1)
wow... (Score:1)
3: Informative (Score:1)
moderators you suck
Re:The wait's over (Score:1)
Re:Nothing to be said (Score:2)
Unfortunately until there is some (and I mean *any*) hardware support this is pretty useless. The OS has to support the processor.
As far as the OS support goes, the did hire Linus, perhaps that's what he does there. As for support hatdware, we'll just have to see. I would be seeing now, but the server's overloaded apparently.
Mobile Linux is an interesting entry... (Score:2)
After all, it sure would be neat to run Linux on a VLIW 128 bit CPU, wouldn't it?
View | Source (Score:2)
Re:Looks like rumours have been confirmed. (Score:2)
Mighty observent, has me thinking..
Re:ZDNET loves MS (Score:2)
2) This has NOTHING to do with Apache or Linux. It's a processor for crying out loud.
3) G2 is available for Linux as well. Use your brain before posting please..
Re:Transmeta Transmission... (Score:2)
Will the idiot who marked my post "Off Topic" kindly reply to this and explain just -how- a post about the Transmeta Webcast, and the station it's on, can possibly be off-topic to a thread concerned with the Transmeta Webcast and the station it's on.
Having seen several "moderator wars" on a number of my posts, where rival factions mark them up and down like a bloody yo-yo, I would like to make it very clear that I have had enough. The moderator system isn't about whether you "agree" with the post or not, whether you like the poster or not, or whether the sky is blue or not. If you're not willing to read the guidelines, you can B*gger Off and screw up some other guy's news service.
If my karma is that much higher than yours, that your ego is getting bruised, try posting stuff that people would find interesting or useful. Not to scrounge karma, but because it's actually -enjoyable- to post stuff that's interesting or useful. If you're unwilling to bother, don't hastle others. There are only so many hours in a day. If you'd rather troll through moderation than discuss something, why not go outside? Have a hamburger. Go walk round a park. Surely, there are plenty of things you can do, besides marking VERY relevent posts (humerous or otherwise) as "off-topic", for no purpose but to harass others.
Re:Transmeta webcast... (Score:2)
You're not missing anything. Both chips are x86-compatible (in that the binary-to-binary translation software turns x86 code into the native code for the chip on which it's running, and runs it), and can run any x86 OS (assuming the OS can handle the rest of the system into which it's put, that is). The slower one could run Windows (or BeOS, or *BSD, or Solaris, or...), and the faster one could run Linux (or BeOS, or *BSD, or Solaris, or...).
Re:Isnt this just programmable ucode? Why a big de (Score:2)
As far as I can tell, there's no such thing. The chips' hardware doesn't appear to be reconfigurable; they have fixed VLIW instruction sets.
The chips will run, at the lowest level, software that takes code for an instruction set such as x86, and translate it to the native fixed instruction set for the chip on which it's running.
It's more like changeable microcode than like FPGA-style changeable hardware, but microcode is generally used to interpret an instruction set, rather than to translate it to a native instruction set not interpreted by microcode (no, the System/38 and AS/400 aren't an exception; the vertical "microcode" was just software, called "microcode" for legal reasons, as per Frank Soltis' Inside the AS/400 book).
Windows Media Player (no other content in message (Score:2)
What is the emulated speed? (Score:2)
Ben
More data (Score:2)
LongRun: Adjusts megahertz to save power when full speed of processor isn't needed for current task.
Performance Benchmarks
1) 700 megahertz: Similar to P3-500 for far less power...
Re:Oh crap... (Score:2)
Did you honestly think this would get by WITHOUT being posted on Slashdot? Please....
-- Give him Head? Be a Beacon?
Re:News articles so far include (Score:2)
Take it everywhere, always on. Shut the screen off to save power.
Jeezus.
Combine this with wireless networking. Make it wearable. Recharge while I sleep. Wow.
Swirl Logo? (Score:2)
Maybe all their orginality went into chip design and they were tapped out when it came time to make a logo? =)
Re:Correction correction (Score:2)
I don't know the details about Crusoe's architecture, but in a general sense what you're saying isn't right.
In microprocessors, the instruction set is implemented on the chip in a kind of firmware called "microcode". Directly underneath the microcode is just bazillions of logic gates, which possess no instruction set other than "gimme a 1" and "gimme a 0".
So the microcode provides an interface between the software instructions arriving at the chip's pins as a set of electrical voltages, and the underlying low-level hardware.
It sounds to me as if what Transmeta have done (and shoot me down if I'm wrong) is to replace the upper layer of the static firmware with dynamic software which can be reprogrammed on the fly. Whether or not the chip can meaningfully be termed as having an "instruction" set depends on how much is still implemented in firmware.
Consciousness is not what it thinks it is
Thought exists only as an abstraction
Re:webcast notes [Errata] (Score:2)
As to other RISC features such as superscalar architecture, multiple execution units with pipelining etc which have recently also been incorporated in modern CISC designs such as the latest Intel processors...I don't know if Crusoe has these but it ought to otherwise I don't think it could compete in performance terms.
Overall, Crusoe's architecture is radically different from both CISC and RISC. It really deserves a new category of its own.
Consciousness is not what it thinks it is
Thought exists only as an abstraction
Re:Or archive the stream? (Score:2)
Consciousness is not what it thinks it is
Thought exists only as an abstraction
From the anouncement page... (Score:2)
Download RealPlayer G2 Alpha (for Linux)
Download the Windows Media Player
Well, whatever it is they are doing, I guess it's not anything that Mac users would be interested in....
wel.. (Score:2)
Anyways. Have they actually built any, or is all their information based on software simulations? Granted, the simulations can be pretty accurate, but then Intel thought that Merced was going to perform a lot better than it did when it reached silicon.
Listening to them describe videotaped presentations, though, doesn't sound very exciting. I wish they had a Crusoe motherboard/chip vs. P3, so it would be more first hand... After all - Microsoft has taught us not to trust video tapes any more....
Correction correction (Score:2)
Now, I didn't see the announcement, but I know for a fact that a processor MUST have instructions. That's sort of the point of it. These need not be x86, and in the case of Crusoe they're not. But there MUST necessarily be native instructions on this chip. It is these instructions that the x86 instructions (mapping in software) are morphed into.
These instructions on the Crusoe, are in almost all certainty RISC rather than CISC, since it's easier (drawing on intuition) to decompose CISC instructions into sequences of RISC instructions.
It actually makes a lot of sense. You have a lookup of CISC instructions for a particular architecture (x86 to start) and their corresponding RISC (native) operation sequence. It's surely more complex than this but: when executing x86 code, you encounter some instruction, you look up the corresponding native sequence and execute it.
Anyhoo, while Crusoe is not 'just a RISC chip', it's native instruction set must be simpler than x86 CISC, since you can't reasonably go in the opposite direction. The magic of this chip is that it seems to do runtime optimization of instructions... And executing only what it needs to??? Freaky!
If you can't get a feed (Score:2)
Re:Transmeta Transmission... (Score:2)
Uhh... MSNBC?
By the way, I think I've figured out the AOL/Time thing: Steve Case just wants to make himself Man of the Year.
Thoughts... (Score:2)
$ prices seemed pretty good, the the smaller chip to be under $100 and the larger chip to be under $200.
I don't think I've been so excited since I saw the Apple ][ for the first time!!
Re:Thoughts... (Score:2)
Engineering briefing to follow.
Full Up (Score:2)
The "Windows Media" still appears to be accepting connection (at least it was when I connected 10 minutes ago).
Here's waiting... 13 minutes to go....
P3 Power Dissipation (For comparison) (Score:2)
500 15.8
550 17.4
600 19.6
650 21.2
700 22.9 -- That means Crusoe is over 23 times cooler and has 23 times longer battery life!
750 24.5
Re:ZDNET's NT servers are full (Score:2)
p.s. This is too on topic! :-) We're talking about the webcast and why we won't be able to see it.
========
Summary (Score:2)
Well, technical difficulties caused me to miss the technical questions, but I'll give you what I got (which is a lot).
Let's start at the very beginnng...
Transmeta was founded in 1995 in response to problems they perceived in the microprocessor industry. Namely that processors have gotten so complex they take
5 years or more to develop, require teams of hundreds of people, and contain millions of transistors which cause them to run extremely hot an consume large
amounts of power. They realized that a lot of these problems could be solved with new processor architectures, but then you lose the compatability which is so
important.
So the two big problems are compatability and complexity.
Transmeta's idea was software based microprocessor, in which software itself is an integral part of the processor. They criticized semiconductor companies for
trying to do everything with more transistors. The benefits of this approach include simplier hardware, which makes less expensive and less power-producing
products. By using software to handle some of the functions normally done in hardware they don't need as many transistors.
The benefits of software are not limited to reduced complexity, however. They later discussed their "Smart microprocessor" technology in which the software
could optimize certain things about your programs as they were run. Basically, the more you used it the better it ran.
The essential thing which makes the Crusoe processors work is their "Code Morphing" software. This software is the only thing that is ever compiled for the
Crusoe's VLIW instruction set. Everything else is in whatever instruction set it happens to be running.
Here is a summary of the "Code Morphing" software.
Morphs x86 into VLIW
128 bit wide instructions
3/4 of functionality implemented in software
Completely invisible dynamic translation
Processor can learn about application program
Very simple and very fast hardware chip
Reduction in number of logic transistors
Their goal is a single solution with extremley low power, x86 compatability, and competitive performance.
There were two processor shown today, the TM3120 and the TM5400. As I mentioned in the previous post, the TM3120 is aimed more at small "Internet
appliances" while the TM5400 is aimed at the more traditional laptop market, although both are completely x86 compatable. Both feature an integrated north
bridge and the TM5400 has DDR memory support. The TM3120 really seems to be being paired with a slightly modified version of Linux they are calling Mobile
Linux. Mobile Linux will be able to fit into ROM modules so these small webpads, etc. do not have to have hard drives and it will include all the basic applications
that one would expect from such a device. It sounds to me like some very interesting products could be made out of this. The TM5400 is aimed at more
conventional laptops (In the entire program they never once mentioned desktops. They are really targetting the mobile market.) and will feature better
performance, but it will still dissipate only 1 Watt of power. Expect traditional windows based machines using these processors, although I certainly wouldn't
mind a good Linux laptop.
There was a lot of marketing speech in the whole presentation, which I will try to filter out. Basically they went on (and on) about how they are the only
semiconductor company that is really designing exclusively for the mobile market. They pointed out how previous mobile processors are just modified versions of
desktop ones, but their processors are different. They compared current notebooks to the first generation cell-phones: bulky, heavy, and impractical. They say
mobile computing is currently hampered by the form factors used.
They went on to discuss compatability. Crusoe processors will be completely compatable with normal x86 instructions as well as MMX. There currently is no
support for either SSE or 3DNow!, although it might be added in the future. They pointed out that where current internet devices have the most difficulty is with
plugins. Plugins are basically small, x86 compatable programs that are downloaded, and most internet devices just aren't x86 compatable. The Crusoe, being x86
compatable, will have no difficulty at all handling plugins.
Next on the agenda was battery life and weight. These are very important things for mobile computing and is the one area where current mobile processors have
the most difficulty. This area really impressed me (as I'm sure it was supposed to). The fact that the Crusoe dissipates only 1 Watt while in operation is just the
beginning. It also features a special sleep mode that will allow it to dissipate less than 20 milliwatts. Also, the smaller number of transistors when compared to
most processors means that the Crusoe generates less heat and doesn't need a fan, resulting in less noise made and less power consumed. But the really cool
part is their "LongRun" technology.
"LongRun" ties in with the "Code Morphing" software and will manage your mhz and voltage on the fly. It actually determines how much processing power your
applications require and throttles the processor back to exactly that level while you run them. It happens instantaneously whenever you start or stop an
application and there is absolutely nothing the user has to do to take advantage of this technology. This simply wouldn't be possible if it wasn't for the software
nature of the Crusoe processor. They call it, "performance on demand".
They also talked about the company, but I'm more interested in the technical stuff (as I'm sure you are), so I'll make it breif. They have 200 employees worldwide,
including Japan and Taiwan because that is where a lot of the manufacturing happens. They were very proud of their design teams, especially the way the
software guys and the hardware guys worked together in a way that just doesn't normally happen. They talked about some of their partners, confirming that IBM
will be manufacturing their chips using their
have the same kind of crappy deal Cyrix got. In fact, they said that IBM had an entire team in Vermont dedicated to nothing else besides supporting the Crusoe.
One of the (many) interesting things about Transmeta is that they are not simply a chip company. They designed their own BIOS, they designed a lot of the apps
for Mobile Linux (especially the power management stuff), and although they never actually confirmed it (at least if they did, I missed it), I'm pretty sure they
designed their own chipset, seeing how they have a custom BIOS and a processor with and integrated north bridge. They also designed a lot of the prototypes for
various internet appliances which could be picked up by OEMs. Basically, they are committed to delivering an entire platform, not just a processor. Maybe AMD
could learn a thing or two from these people.
On to some of the specifics about the processors. They don't have any sort of registry renamers or reorder buffers in hardware, it's all done in the software. They
actually used a very modest VLIW engine. (Although one of the really interesting things they mentioned is that their different processors don't have to use the
same instruction set. In fact, the TM3120 and the TM5400 have slightly different instruction sets optimized for different tasks. This has a lot of potential for
future scalability.) They weren't very clear about this point, but they seemed to imply that their software could actually theoretically be ported to other
architectures, although with very poor performance because the architectures would not be optimized for using software in that way. But the point was that it's
the combination of software and hardware that makes the Crusoe special. They are really designed to work together. Many things are implemented in software to
save on hardware, and other aspects of the hardware make it well optimized for using the software in this way.
There are no x86 decoders in hardware at all. The x86 instructions reach the software where they are decoded, translated, and stored in a special memory area
that essentiall caches translated programs for future reference. This means that the processor does not have to be constantly translating the same instructions
again and again, although it will only cache the part of whatever programs you are running that you are currently using.
The Code Morphing software goes beyond simple translation, however. There is a "back end" that can actually learn about and optimize your programs as they are
run, actually justifying the marketing term "smart processor". The Code Morphing only takes time to optimize the most commonly used parts of your programs. In
one example they gave, it was possible to take 20 x86 instructions and do the same thing with only 10 VLIW instructions. Making the programs smaller in this
way can also help lower the power consumption.
They went on to discuss different benchmarks. This part reeked of marketing and I think they tried to cover up some things, but I'll tell you what I can. Basically,
they criticized current benchmarks for laptops for not taking into account the amount of power consumed in going through the benchmark. I suppose there is some
truth to this claim, but that doesn't justify what they did. They actually created a benchmarking program that (somehow, we don't know the actual formula used)
actually integrates power usage into the score. That's right, there isn't a seperate performance score and power usage score, they are part of the same number.
I'm sure this is great for a company advertising a really low power usage processor, but does it make sense. To add to the complexity, the entire benchmark was
sort of calibrated around a reference system. They said the system was a P3 500, although they didn't mention anything else about it. The reference system is
given a score of 10 and everything else was compared to that. Basically, the Crusoe scored in the 30-40 range, which might be really impressive if I knew what in
the hell that score actually meant. Oh well.
They did mention some things about real-world performance, although no other benchmarks were given. They said that DVD decoding and mp3 playing should
perform comparably to other x86 processors, although I don't know whether that is compared to the processors out now, the processors that will be out when the
Crusoe is release, at the same mhz, or what. So I don't know exactly what this all means. They also said that other applications should perform at 80-90% of
other processors, and in this case I have two "others" I don't know what they mean. Later on in the Q&A they said that most business apps should perform
slightly better. Better than what? Either they just weren't clear or they're trying to hide something. Performance might not be the Crusoe's strongpoint, although
frankly I can live with that if the claims about battery life are true.
The next part of the program was very amusing. We all got to watch Linus Torvalds play against Dave Taylor (one of the original Quake designers and also a
Transmeta employee) in a game of Quake (running on TM5400s of course), although they didn't say what version. Linus basically got his ##### kicked. That's
about all there is to say about that part.
There was a lot of stuff about business forcasts and the like after that. They looked at the laptop market and pointed out that the fastest growing segment of that
market is in the thin and lightweight models pioneered by Sony's VAIO. Obviously if the claims are true, Crusoe is perfect for that market. I wonder if Sony will
make a VAIO based on the Crusoe? They also talked about Mobile Linux and how it will be able to fit into ROM for internet appliances. They even went on to talk
about how there might eventually be a palm-sized device based on the Crusoe. Obviously, that would be just impossible with any other x86 compatable processor.
One story I would like to mention deals with the software nature of the Crusoe. They talked about a situation in which some guy in Japan came across a bug in an
early version of their product. Instead of having to fix the design in hardware and fab a new sample, which could take months, they fixed it in software and sent a
fix to the guy over the internet. Everything was working fine in under a week. The Software-upgradability of the Crusoe really has some interesting applications.
Finally, some mention of prices. They said the TM5400 would be in the $119-321 range and that the TM3120 would be in the $65-89 range. PCs based on them
could be around $1200. They said that internet appliances based on the TM3120 could be anywhere from $500 to $1000, as there is a lot of room for variation in
their configurations. No specific products have been announced yet, they are leaving that up to the individual OEMs. They did say that a lot of companies are
interested in their technology, and I don't blame them.
Re:What is the emulated speed? (Score:2)
Re:What is the emulated speed? (Score:2)
Re:webcast notes (part n) (Score:2)
Or how to shoot your karma to the stars.
There should be more Transmeta announcements
(But thanks from all those that can't see it directly nor hear it
Re:Correction correction (Score:2)
There is an instruction set to any processor, or it doesn't work. By defintion.
The point is that, with something like a P-II or a Crusoe (from what I can see right now) the instruction set really isn't very relevant. It's not accessible and it probably wouldn't even be especially easy to use if you could. The simple idea of this is that they can make this core full of micro-ops run faster after translation than they can get the standard core, so they do it that way.
I agree that it would appear that they've replaced the static translation firmware with dynamic, reprogrammable software. It definitely has an instruction set - and this could get even more interesting if that's accessible - but the whole point of this is that the translation is blindingly fast.
Whatever, I'm interested
Greg
Partnership with IBM announced (Score:2)
Re:Transmeta webcast... (Score:2)
BTW: 400 MHz for Linux and 700 MHz for Windows.. LOL. But 1W/chip at 700 MHz with 400kB cache...that's impressive.
OTOH, anyone can do that... (Score:2)
I can't wait until we get real software chips: imagine a grid of tiny computational elements that can be set to perform one of several functions (like, xor east and west inputs and output to north, inverted to south). There are some really cool ideas for this out there, like making the data and configuration lines the same, so you have to start programs at the edge and make them worm their way inwards.
Imagine being able to rewire all the transistors in a modern chip into parallel vector crunchers for graphics apps! Even if you wasted 90% of the transistors in switching, you'd still probably have 20 to 100 times faster calculations.
This sure is a step in the right direction, though. The big problem with current designs is that only one tiny portion of the chip is working at any one time. The hardware needs to be split up and made more versatile so more stuff can be run in parallel.
Re:webcast notes (Part 7?) (Score:2)
:-)
Thimo
--
No native instruction set to be released (Score:2)
~~
Re:Traditional press are a bunch of assholes (Score:2)
The Engineering Q&A is MUCH better...
Re:AAAAAARRRRGGGGGHHH!!! IT'S JUST ANOTHER x86 CHI (Score:2)
It's not "just and x86 processor." That was, understandably, their first target.
Phew! 10mw on standby! The 667MHZ chip runs faster than a P-III 500 MHz, and uses less power playing DVD than a PIII does idling.
TRANSMETA Website to go live NOW! (Score:2)
TRANSMETA Website NOW LIVE! (Score:2)
Wait, I got through. The logo for Transmeta is downward pointing teal triangle with a wavy line crossing it.
they have a
Still to loaded to actually vist these paghes though.
minor pricing correction (Score:2)
The 500- and 700-MHz TM5400 versions will list for $119 and $329, respectively. The 333- and 400-MHz TM3120 devices will list for $65 and $89, respectively.
--LP
Re:Transmeta Corp. IPO? (Score:2)
--LP
News articles so far include (Score:2)
one by techweb with more details than the other articles, [techweb.com] and
one from Infoworld [infoworld.com]
--LP
Re:AAAAAARRRRGGGGGHHH!!! IT'S JUST ANOTHER x86 CHI (Score:2)
From the sounds of this chip they can do whatever they want with this chip. First they will get the market with it, then they will change it to whats best.
Updates (Score:2)
For those not watching, Crusoe processors are long-word processing systems that load the whole of various instruction sets (x86, etc.) in software. Designed with mobility foremost, leading to low power and low heat output, targeted towards clients rather than server platforms.
More interesting is the idea (mentioned briefly by the monkey on stage) of Crusoe's software adjusting its performance and configuration to software. I drooool.
Re:P3 Power Dissipation (For comparison) (Score:2)
500 15.8
550 17.4
600 19.6
650 21.2
700 22.9 -- That means Crusoe is over 23 times cooler and has 23 times longer battery life!
750 24.5
WRONG! WRONG! WRONG!!
This is disinformation.
The Baseline p3 gets a rating of 10, so 22.9 means it's TWICE AS COOL, and TWICE AS MUCH battery life. NOT 20 times.
This is backed up by the Temperature stats where it the P3 runs at 110 degrees C, and the Crusoe runs at 45 degrees C.
Kintanon
Re:P3 Power Dissipation (For comparison) (Score:2)
Though yes, it does use FAR less power.
Kintanon
Re:Is there a way to get a transcript or a video t (Score:2)
Looks like rumours have been confirmed. (Score:2)
For those not in the USofA (Score:2)
Re:Nothing to be said (Score:2)
Unfortunately until there is some (and I mean *any*) hardware support this is pretty useless. The OS has to support the processor.
Offtopic: Your PEZ comment (Score:2)
As for the electric dispensers, I had a circular one that I bought at K-mart, but it kept jamming and then it just quit working after about a week. We need Tux Pez dispensers.
Re:webcast notes [Errata] (Score:2)
Sounds like the Instruction Set has been greatly Reduced. That qualifies as RISC in my book =)
Re:ZDNet and Real (here I go again) (Score:2)
Re:Panic Time (Score:2)
Fine typographic taste ... (Score:2)
It's a nice informal font (that's why we are giving it away), but if they'd asked me I wouldn't have used it for a corporate logo.
-Martin
A summary of what was shown. (Score:2)
Two chips were shown, a 400 MHz chip for internet appliancies and a 700 MHz for ultralight laptops. Using "LongRun" power management the chips use as little as 1 watt of power during average use. A demo showed a 667 MHz chip drawing about 2 watts to play a DVD movie, while an Pentium III 500 used 7 watts. A real world system's supposed to have about twice the battery life of an equivalent Intel-based system. The 667 MHz chip performs like a PIII 500, and the translating software is still being tweaked.
Transmeta will not release the nativer architecture of the chips. Thanks to the efficent emulation software, they can give good x86 performance and still change the silicon without worrying about backwards compatability. The two chips announced have partially different ISAs, but both appear identical to a Pentium to applications.
There's real functioning silicion, shown running both Windows and Linux. IBM will provide semiconductor tech and fabs when manufacturing starts in mid-year.
Transmeta has developed a special "Mobile Linux OS", a mini-distro for diskless handhelds and the like. (note: the OS runs in x86 emulation, not natively) The power management enhancements made to Linux will be merged into future kernel releases.
Re:AAAAAARRRRGGGGGHHH!!! IT'S JUST ANOTHER x86 CHI (Score:2)
Give Transmeta a chance..It makes complete sense for them to focus on the x86 compatibility of the chip. If they started talking too much about the software reprogramming of the CPU it would just confuse their launch message, which is all about the benjamins, not the geek-factor of the technology.
Re:webcast notes (Part 7?) (Score:2)
Linus talked about how TM flew him out from Finland, because he had to sign the NDA before they would tell him anything. So then he had a one day discussion of what TM was working on, and then he went back to his hotel room and thought "what a bunch of crazy guys".
He said that he was looking at a job which would offer him something fun, a chance to work on a cool new technology.
Then a journalist asked him what job offers he turned down, and he repeated his common line about getting lots of offers from Linux companies, and how he really didn't want to polarize the community towards one 'official' linux to the detriment of others. So he chose a company that offered him a chance to work on a fun project that included linux but wasn't a distribution.
Real Player through Firewalls (Score:3)
Is Crusoe Revolutionary? (Score:3)
It struck me as odd that Ditzel claims Crusoe is the first processor to make use of a software solution. It may be the first designed to do this from the ground up, but the code morphing idea is not really all that new.
The IBM S/390 has the ability to completely recode the instruction set through microcode. The Alpha has PAL code for VAX compataility. FX!32 provide x86 compatability and "learns" about the program through profiling and reoptimization. Dynamo is HP's answer to FX!32, providing similar functionality to PA-RISC and probably IA32. While none of these are quite exactly like Crusoe, it's the same general idea.
Dynamic recompilation research has been going on for a while. Translating JVM's are only the most recent example. In fact Transmeta hired some big names in this area, which of course makes a lot of sense.
We've been kicking around the idea of self-optimizing computers for a while now. The idea being basically what Crusoe does - examine the software running and adjust accordingly. Now if a couple of inexperienced graduate students can come up with this idea after reading a few papers, it can't be all that revolutionary. :)
What I did like was the fact that someone finally publicly (i.e. through marketing) expressed the value of being able to change the underlying native instruction set. The question of whether the native ISA would be exposed was asked. The answer was no. This is a good thing, as it frees the hardware designers from having to carry around a bunch of compatability cruft. It's almost a necessary thing when your core is native VLIW.
I found it humorous that the bigwigs at Transmeta feel that Linus is the guy you want on a project like this. I figured you'd want someone with expertise in computer architecture and compilers. Yeah, I know it was a throwaway line, but we had a bit of a chuckle. I'm sure Linus us doing a great job. :)
As for the power consumption, it's not all that spectacular when you consider that the StrongARM has been sub-Watt for some time now. I'm not sure how many hardware power tricks were pulled with Crusoe, but the power demo was very entertaining. In any event, no one has been able to run x86 at a Watt, so that is certainly worth some praise. I know that Intel is very, very concerned about power (from a friend I have there). This could make them a bit nervous. I'm interested to see if they have anything that can compete.
But my big question concerns the web pad. First of all, that screen is tiny! No wonder it's low power! Aren't the disk and screen the largest consumers of battery power? It doesn't seem like Transmeta does anything about that. Hopefully the unnammed (and apparently unsigned) OEM's can solve that problem.
Even ignoring the screen/disk problem, why is Linux "the obvious choice" for a ROM-based OS in a web pad? If their main goal is a "fully internet compatable" low-power mobile device, it seems as though they've shot themselves in the foot by relying on an OS that has sub-average plugin support at best. They kept touting how the x86 compatability provides full support for plugins, but the OS question was completely ignored. What's the answer? It doesn't seem fully internet compatible to me.
--
transmeta chip specs (Score:3)
Chip 1
700mhz
400k cache
1w power use (yes, *1* watt)
100% x86 compatible
chip 2
400mhz
100k cache
1w power use
Re:Transmeta webcast... (Score:3)
No. The software that runs natively on the VLIW core does all the translation, optimization and such.
But oh my, what optimization! They're talking about ass-end compiler optimizations.. very 'big' algorithms that won't fit on silicon.
I wonder if you could buy the code optimizer and have it spit back hyper-optimized x86. You know, the code runs and self-profiles, rewrites and then retranslates to x86.
I also have to wonder how scared Intel and AMD are right now. This puppy runs COLD, FAST and is SMALL and CHEAP to build. Plus it is self-optimizing and 100% x86-compatible. I think Intel et al just lost their mobile market.
I was particularly impressed with the Longrun technology. If those were actual true power consumption graphs...
Transmeta Transmission... (Score:3)
For those impatient, don't worry! Using the latest bzip2 compression software, I have squashed a time machine into this box: [#]. Just extract, expand and go forward one hour. If you get back, could you post the summary here?
The Operation Begins (Score:3)
Twenty minutes, and the gloves come off. They've had years to prepare for this; now we get to see if the computer industry gets its first proper launch.
No more delays. No more promises. The most carefully marketed R&D house in Silicon Valley is about to open their doors--it'll be intriguing to see if they're as skillful on the open scene, under attack by the whispering galleries of competitors.
Hiring Linus was genius--you couldn't buy a more devoted audience.
Yours Truly,
Dan Kaminsky
DoxPara Research
http://www.doxpara.com
Curusoe: 2 Different Processors (Score:3)
2) 400 hegahertz x86-compatible 1 watt dissipation 100kb+ cach, targeted at Linux devices
No actual performance figures, at least yet...
more update... (Score:3)
Linus is closet q player..
Dave taylor runs quake on Windows, Linus on Linux..
667 TM 5400...
Makes a joke that if he loses its not the OS...
Linus bites it..
Runs really GOOD!! and FAST!
Linus bites it again.
Dave starts making fun of Linus !!
Bites it again..
Jim Chapman.. Marketing.
Intel/Cyrix..
left cyrix in 1996
2 lines
Emerging internet appliance.
existing mobile market.
mobile pc's focus on sub 4lb machine.
2 spindle hdd, dvd/cdrom
lightweight
1 spindle > 3lb's
fastest growing segment 80/90% dev on type product..
user lightweight/better battery
3lb single spindle == viao short battery life.
highest quality entern...
longest battery life in apps..
4x/5x's less battery usage.
less thermal solution reduces weight and cost.
dvd machines in slim form uses no more then 2hours in dvd mode..
mobile 17% crusoe to improve batter, weight, preformance.. new thinner form factors...
in the hands of the leading mobile makers in mid year..
Mobile inet device.
market neeeds standard..
uses linux as logicla chouce.
low power
x86 compat..
between 500 - 1000 dollars.
no way to get low power before crusoe.
can get linux in rom.
Transmeta:
End product enhancing
end user considerate
industry growth
original r&d
add value throught software
hardware optimized.
two memory controllers.
--- i can stop this since the transcript is in #transmeta..
Solipsistic Semi-Naked Savages. (Score:3)
IRC discussion during/after announcement (Score:3)
Re:700Mhz / 400Mhz (Score:3)
Low Weight and new Form Factors
Rid of noisy fans
That is exactly right. With all of the faux battles between chip manufacturers and video card makers resulting in 15% clock rate improvements here and there, the irrelevancy of it all is often difficult to see. An 800 MHz Athlon with some nutty next gen 3DFX chipset is completely pointless when what people really want are low power consumption devices that don't fit the clunky desktop model.
There have been a number of people focusing on developing low power CPUs that do what people want, but they've been mostly ignored because web geeks like to hear about overclocked processors that need to be dunked in liquid nitrogen. Crusoe could potentially take over the world, as the big PC chipmakers aren't even in the same universe.
more notes (Score:3)
they're putting the Linux OS in ROM for some devices
the form factor of a "vaio-like" device with Crusoe is smaller than previous existing models with comparable compute power; the prototype they show seems to be about 7"x5"x.75", although this is a very rough, rough eyeball.
some talk about using this as a wireless Internet client; per month pricing totally unclear so far
--LP
pricing (Score:3)
the $60ish to mid-hundreds range.
mobile crusoe PCs expected to sell around $1200-2500
wireless internet PCs pricing expected to work somewhat like cellphones, with business models TBD.
--LP
Is there a way to get a transcript or a video tape (Score:3)
Re:ZDNet and Real (I got audio) (Score:3)
http://www.zdnet.com/zdtv/static/radio/ram/list
Not sure if this is the right stream or not though. But I'm getting audio...
Embedded/mobile processor; Ideas of Ditzel, Hunter (Score:3)
Ditzel, since his days in academia, has been a champion of the philosophy that compilers and optimizers can make simple, less capable hardware outdo more powerful hardware. We'll see if this is so. (This approach didn't help Sun; it's unclear if it will help Transmeta.)
The biggest potential problem of the architecture, whose emphasis is on power savings, is choppy performance due to unexpected pauses for code translation and/or "shifting gears" from low power to high-performance modes. I don't know enough about the details yet to evaluate the impact of these issues. If I can actually test a machine with the chip in it I'll be able to say more.
--Brett Glass
The secret's out... (Score:3)
webcast notes (Score:4)
End part I
700Mhz / 400Mhz (Score:4)
so far he has showed two vers of Crusoe. One is a 400 MHZ version for handhelds and other smaller mobiles.
there is a 700Mhz version for Notebook size mobiles. All have X86 compatibility and only use 1 Watt of POWER!
more and more this product is becoming focused on the internet.
Crusoe supports:
1. Full PC Architecture
2. Runs all X86 Operating Systems
3. 100% X86 Application Compatiblity
4. Run WEB BROWSER & Plugins
Crusoe uses very little power. 20 miliwatts of power in STANDBY MODE.
Crusoe's low power offers what people want.
Long Battery Life
Low Weight and new Form Factors
Rid of noisy fans
Transmeta hired Marc Allen to help run Transmeta.
Now going over background of Transmeta...
Summary of introductory speech (Score:4)
- Crusoe is, from what I understand, an extremely simple and compact RISC processor that supports x86 entirely through an integrated emulator. It is as fast as "traditional" processors, cheap to produce and extremely low power (1 W)
- There are two models right now : a 400mhz/~128kb cache for notebooks, and a 700mhz/~400kb cache for laptops.
- Several APM techniques, like a "deep sleep" mode
that means you'll almost never have to reboot your laptop. Cool
- They both run every x86 operating system -- emulation is complete.
- More processors are forthcoming! Looks like Transmeta is planning to take over the mobile market. "If it has a battery and a web browser, it's going to be built with Crusoe".
Well, it looks like their gamble paid off
Broccolist
ZDNet and Real (Score:4)
How are we supposed to see/hear it? Anyone know a way of getting the stream without using the plugin?
Transmeta webcast... (Score:5)
Re:webcast notes (part 4) (Score:5)
Re:webcast notes (Score:5)
- Transmeta corporation ______________
- Transmeta hired a new CEO - Mark Allen of Netscape fame.
- Transmeta has 200+ people on staff.
- Manufacturing is overseas in Taiwan
- IBM partnership!
- Paul Allen invested in them.
- Buzzword Alert: "technology innovator"
- Transmeta will not sell directly to end-users. It is OEM only.
- Transmeta has modified the linux codebase with their code morphing tech
End part II. Opening to Q&ARe:webcast notes (part 3) - linus plays quake! (Score:5)
(( Missed part of the webcast due to work ))
End part III, on to tech details
More update.. (Score:5)
Targeted for internet mobile device.
400Mhz, L1 Cache
Integrated NorthBridge.
5400
Runs Windows/Applications.
700 Mhz, L1 128k, l2 256k
DDR Memory Controller.
Morphs x86 into VLIW instru.
Registers and other X86 Tech is done in software.
Simple fast, low power processor..
code morphin'
Translates x86 pc apps to vliw instructions
Generates x86 to vliw, optimizes, puts in translation cache that is hidden and only the processor knows of it.
only what is used is translated, done on the fly.. keeps programs in translation cache to aid in morphing on the fly.
learn and optimize.
Analyzing the x86 app that is running and optimizes the top code.
register renaming, reordering.
software can optimize like in the compiler
20 X86 Instructions--> 10 vliw instructions
removes useless code to get better preformance and lower power.
benchmarked on a windowing environment.
shows video with fast windowing environment.. translation doesn't get in the way.
P3 Playing DVD
105.5 C
Crusoe Playing DVD with Translation.
48.2 C
Thermal solutions are draining the Mobile battery, and crusoe puts out less heat, less thermal solution == less battery..
LongRun Technology in Code Morphing adjusts power consumption to match needs of the application and uses the lowest power required.
700 Mhz at 1.6V bumps down the MHZ then the V's
Can make the analysis on a frame by frame basis ( example used is playing a DVD ).
700 Mhz uses 1.65V is 100% then the processor goes to 400 Mhz using 1.4 Vs then down to 333Mhz
at 1.2v. Thats only using 25% of the processor.
Introduced new Benchmark Methodolgy:
1. Benchmark on a real APPLICATION
2. Measure battery AFTER the app is run
3. Combine the two for the score.
Demo Crusoe on Linux with Linus.. be back in 30mins.
Traditional press are a bunch of assholes (Score:5)
The Slashdot/Red Herring/Etc type reporters are the only ones asking intelligent questions about the products!
"Hi, I'm a total bitch. I came in late and didn't really listen. This isn't really a new idea, is it? Isn't everyone doing this? Can you explain?"
Sheesh. Someone must be training those people to be clueless and annoying, unless they are actually clueless and annoying and just don't know any better.
Re:webcast notes [Errata] (Score:5)
The Crusoe processor is not a RISC processor; the presenter was involved in the development of the RISC processor ~1980. The Crusoe family (there are two being presented, with more to be announced) is a small processor that does not have any instructions on the chip itself; it has a software couterpart that translates standard x86 instruction calls to the chip itself. It is completely x86 compatable and the two that were demonstrated operate at 1 watt, making it ideal for mobile computing platforms. To summarize, Transmeta has created a brand new processor family that uses software to handle many of the functionality that is traditionally embedded on the CPU itself.
Oh, and the "marketing geek" is David Ditzel, CEO and president of Transmeta, one of the architects of the Sun SPARC microprocessor and longtime advocate of the RISC architecture; I highly recommend checking out some of the books and articles that he has written over the past 25 or so years working with microprocessors, as they are quite interesting. I will admit that he did abuse the "Internet" line a bit much, especially since he was introducing a bloody processor.
Deosyne