Intel Pushes Low-Power Crusoe Challenger 110
axis-techno-geek writes: "It seems that Intel now is trying to fight back against Transmeta with their new chips. Intel plans to have their new speed-step Pentium III's out in about a year (which in computer time means about 18 months :)."
Re:That's impossible! (Score:2)
The Crusoe can dynamically change it's clock speed and power consumption i.e. if you are playing solitare, there is not much demand for power from the procesor so the clock speed will be reduced. If you are playing a DVD, the CPU demand will be high and thus, clockspeed will increase automatically to compensate.
Whooppee... (Score:4)
Personally, I don't see this as being any great news from Intel. As far as I'm concerned, the most impressive thing about Crusoe isn't the fact that it's much better for battery life (though that's certainly a good thing, and it's about time SOMEONE did something about it), it's the code-morphing abilities of the chip. This chip has the ability to emulate any processor that TransMeta puts its mind to emulate. And this can be accomplished with SOFT patches (no more difficult that upgrading your BIOS). For now they have chosen to stick the the x86 market, which makes perfect sense. But Apple and Sun better look out, once they get themselves firmly established...
Re:Still not working for me. (Score:1)
Re:At least Intel gives you a choice. (Score:1)
Likely it will only surprise their more clueless shareholders and investors, who haven't bailed out by then.
Let's face it, Linux weenies like Transmeta because Linus works there. If Linus had gone to work at a sexual novelty factory, you'd all be running around in the room going whee whee whee and turning up the level on your Transmeta vibrating butt-plugs.
ok, guys, and gals .... (Score:1)
Re:At least Intel gives you a choice. (Score:2)
Transmeta (Score:4)
Fight back against vaporware with more vaporware? Sounds like a fair fight to me.
-
it validates his company's approach to the market. (Score:3)
Ditzel says: [Speedstep] validates his company's approach to the market. This joke has been used so many times it's no longer funny.
As others have said here, Speedstep is not progress--it should be called "SpeedStumble" because it drops the clock frequency on battery power--it can't dynamically adjust the frequency depending on the CPU load, like Crusoe can.
The critical point here is if Intel can drop power consumption and thus heat production low enough to be able to do away with the fan. Laptops should not have fans, and should not be too hot to keep in your lap. With Crusoe (or other low-power CPUs) you don't need a fan--with Intel notebook CPUs you still do, so far.
|Re:At least Intel gives you a choice.| - stupid. (Score:1)
Oh, and transmeta rates the crusoe at
4xthe power=1/4 battery life | a %20 speed/Mhz advantage does not make up for a 300Mhz deficite.
Puh-leeeeze (Score:1)
Don't flatter yourself. Intel, in no uncertain terms, does NOT consider, nor should it consider, Crusoe to be a threat.
Silicon vs. vaporware. OK, some Transmeta hardware has been produced, but not in large volume, not dealing w/tiny profit margins, capacity issues, fab yields, etc.
Wake me when Transmeta is real - Andy
Re:a bit confused (Score:2)
When I was shopping for an electric heater to put in my drafty college apartment, I got a real kick out of something. Some of the boxes that the heaters came in were proudly labeled "100% efficient". I guess they made a safe bet that most consumers don't know anything about thermondynamics, and that those who do would simply chuckle.
Re:LongRun vs. Speedstep (Score:1)
What about a fully-static CPU? (Score:3)
Would this not work better than slowing down the clock?
Thanks
Bruce
Intel does it again! (Score:2)
Old news (Score:1)
Re:That's impossible! (Score:1)
More ideas to reduce power consumption (Score:1)
2. I don't need 52X CD-ROM. Many people don't. At least, there should be an option to slow it down. Playing MP3's requires 0.1X speed.
3. Same for HDD's. Take this marketing idea, it's a goldmine: ultra-low power HDD with variable transfer rate, down to 20kbytes/s. Imagine 10Gb Nomad Jukebox running for days on AA batteries!
4. 200MHz ought to be enough for everyone (to play MP3's, of course)
5. Make the screen power-up only when I look at it (eye-movement detection).
Re:That's impossible! (Score:1)
Re:Crusoe can emulate any arch... (Score:1)
Transmeta's Code Morphing technology is obviously not limited to x86 implementations. As such, it has the potential to revolutionize the way microprocessors are designed in the future.
SO as soon as they build a chip that isn't the Crusoe (a chip limited to x86 emulation) they can take advantage of the flexibility of their code-morphing software.
Kendall said:
I can't follow the link you provided...
I was fixing it, didn't finish, then did something else, forgot about it and posted. He's a wearble-computers researcher and invented the matchboxpc/server which he took commercial and now markets as the Tiqit.
Re:Sure, I wouldn't mind a portable Cray. (Score:1)
Well, now that you brought it up, sure. Why not have a laptop that can FFT a SETI@Home unit in 2 hours? That can compute the lightmap to 9plats (my custom Q3 level that's only 400K in file size but is also 1/5 of a mile wide) in 4 hours? I'm game for that - even if I have to plug it in everywhere or have only a 1.5 hour battery life.
Re:Crusoe can only emulate x86 arch (Score:1)
So I will amend my initial statement slightly. You are correct: "Crusoe" (Transmeta's product as it now stands) can only emulate the x86 arch. However the TECHNOLOGY behind the product (I guess you could look at it as Transmeta's I.P. base, but people get so up-in-arms around here the moment you mention I.P.
I'm Sorry... (Score:1)
The link you are all looking for: (Score:1)
Re:this isnt just underclocking.... (Score:2)
Some actual links.... (Score:5)
- http://www.pcworld.com/news/ar tic le.asp?aid=31482 [pcworld.com]
Also next year, Intel will probably announce an ultralow-voltage 500-MHz chip for subnotebooks that should be extremely battery-friendly. The company demonstrated a processor running at 300-MHz.
- http://www.zdnet.c om/ zdnn/stories/news/0,4586,2639424,00.html [zdnet.com]
When Intel Corp.'s Frank Spindler disclosed that the chip giant is fast tracking the production of a new ultra-low-power chip for notebooks, there was an interested eavesdropper nearby.
- http://cnet. com
SAN JOSE, Calif.--Intel said it will enhance energy-saving features in processors for laptops and begin shipping new mobile Pentium chips next year as competition intensifies.
Re:a bit confused (Score:1)
> electricity, does that mean it runs cooler as
> well?
Yes. The two are inherently linked. Basically all the power (watts) a chip uses gets changed into heat. So, a 15W chip is effectively a 15W heater. A 1W chip is only a 1W heater.
ttyl
srw
Re:competition is grand (Score:1)
My sig will not be *** censored ***
Re:Bad Link (Score:1)
timetable (Score:1)
Re:Bad Link (Score:2)
Can someone full-quote the thing for me, please?!
LongRun vs. Speedstep (Score:4)
to conserving power by reducing clock speed than Intel's Speedstep,
but how much difference does it make in practice? It's quite possible
that Intel's crufty hack might give 90% of the benefit. I'm waiting
for independent tests...
Re:Whooppee... (Score:1)
One thing is for sure, intel had to much power when it mattered, because IBM picked the slower.weaker x86 line of chips over the more powerfull and faster 68k series.
On nthe bright side, at least windows is running on inferior hardware. MacOS is weak, but G4s are fast and it almost makes up the difference.
Re:Bad Link (Score:2)
When Intel Corp.'s Frank Spindler disclosed that the chip giant is fast tracking the production of a new ultra-low-power chip for notebooks, there was an interested eavesdropper nearby. Transmeta Corp. CEO David Ditzel was attending the Microprocessor Forum this week when Spindler, vice president and general manager of Intel's Mobile and Handheld Division, said that a new ultra-low-power Speedstep-based mobile Pentium III chip was in the works and should ship next year. The chips, which consume less power than current mobile Pentium IIIs with Speedstep Technology, are being moved into production sooner than originally planned to counter Transmeta's low-power Crusoe processor, which began shipping in notebooks in Japan last month. As for the eavesdropper, Ditzel said he was glad to hear Intel is addressing the low-power market segment, because it validates his company's approach to the market. With their low power consumption, Transmeta's low-power Crusoe processor and LongRun power-management software and Intel's forthcoming low-power Pentium III are designed to extend battery life for mini-notebooks to up to 12 hours. Intel officials concede that Transmeta did wonders for raising industry awareness of battery life. But, unlike Transmeta, Intel expects that the mini-notebook will account for only a small part of its overall mobile chip business, whereas more fully featured notebooks with larger screens, integrated drives, and faster processors will account for some 60 percent of sales. Shaving wattage According to Ditzel, the difference between Crusoe and the forthcoming Intel chips will be about a quarter of a watt. Crusoe chips on average consume about .5 to .75 watt; the forthcoming Intel chips will consume about half a watt on average. (Currently, Intel is shipping a 600MHz mobile Pentium III with Speedstep that consumes about one watt of power when running at 500MHz in its Battery Optimized Mode.)
Ditzel showed ZDNet News a Hitachi notebook with a 600MHz Crusoe TM5600 chip. The notebook, along with three similar models and a Crusoe-based Web Tablet, debuted in Japan on Sept. 27 and will be officially announced later this month in the United States.
Ditzel demonstrated how the Hitachi model scaled in 100Mhz increments from 300Mhz to 600MHz, but spent most of its time running at 300MHz unless it was loading an application or playing a movie. According to Ditzel, the notebook yields a battery life of between 10 to 12 hours.
Intel's first ultra-low-power Pentium III chip, due in the first half of next year, will be a 500MHz, which drops to 300MHz when on battery power. When running at 300MHz, the chip will consume about half a watt of power.
Active power consumption for the chip will be closer to 1.5 watts, Intel officials said. The chip's core voltage will drop to 1 volt, compared to the 1.1 volt of the current 600MHz low-power mobile Pentium III with Speedstep. A 600MHz version of the ultra-low-power mobile Pentium III is planned for the second half of the year.
Speedstep, which acts as a switch to drop clock speed and power consumption of a Pentium III running on battery power, is not dynamically variable like Crusoe, which can move through a number of clock-speed and power-consumption states based on demand from an application.
System-level approach Intel is also focusing on chip sets and other system components, including screens and batteries.
At the Microprocessor Forum the company demonstrated a version of its 440MX mobile chip set with active power management. The chip set's average power consumption was shown to be about half a watt, making the entire processor and chip set package average less than one watt. Intel is also investigating ways to get more life out of lithium-ion batteries and to reduce the power consumption of notebook screens. The company may make investments in companies working in those areas; however, it has not disclosed any such plans to date.
Transmeta's approach to system-level power management was to integrate the memory controller, a major part of a chip set, into its Crusoe 5XXX line of chips. This works to save both power and cost.
The first ultra-low-power mobile Pentium IIIs will be based on Intel's current 0.18-micron manufacturing process.
They "require nothing special in terms of manufacturing," one company official said.
Re:Intel's brilliant new idea (Score:1)
Really? I didn't know that PII 300's ran at 1/2 W.
Re:So it's... (Score:2)
Vapor for Athlon, vapor for Crusoe.
--
Give me a candidate who speaks out against the war on drugs.
Old news, innit? (Score:1)
ISTR reading about this on The Register many months ago.
--
Re:intel is losing it... (Score:1)
Bought a t-bird 900 today, and successfully o/c'd it to 1.25GHz.
Crusoe can only emulate x86 arch (Score:1)
the chip. This chip has the ability to emulate any processor that TransMeta puts its mind to emulate.
That is true...as long as the processor T$ wants to emulate is an x86. According to Professor Vaghan Pratt of Standford University [stanford.edu], the Crusoe series is only capable of emulating the x86 processor architecture as it's wired into the chips.
Re:Intel's brilliant new idea (Score:1)
HP omnibook 900b's currently come with the exact same thing, you buy a 650 and it clocks down to 500 when on battery, if you had read the article you would have heard something about that.
Re: AMD problems (was intel is losing it...) (Score:1)
My advice to people with AMD instability problems is to flash the BIOS to the latest revision, and download new motherboard drivers (and if that doesn't help, check that the power supply is up to the job).
Re:Gotta hand it to Intel's marketing department (Score:2)
But you missed the cool part - you can overclock it and get back to where you started!
They'll doubtless market that as the Speedstep Pro.
--
Give me a candidate who speaks out against the war on drugs.
Re:first prost (Score:1)
Holland? (Score:1)
The recent tally... (Score:4)
Intel recalls the 1.13 P3
AMD rolls out 1.1 and announces near term avail of 1.2 GHz Athlons.
Intel repurposes P4 systems after performance issues, show that tick for tick they run slower than P3, Athlon & Duron. (at least until they can shrink the die small enough to run at 2GHz which puts them on par with todays chips)
AMD announces SMP chip set with initial 2 way (this year) support for their Athlon line. And the SMP is higher bandwidth than current Intel systems...(Finally!!)
Micron's MAMBA DDR SDRAM chip set is announced for Athlon systems - with 8 MB built in cache!
Intel initial chipsets for the P4 will be RAMBUS ?
Now they are going after the Crusoe chipsets with more underclocking tech ?
No I have to admit my info may be flawed, as I have not been watching the processor news for the past 20 minutes, but it seems to me Intel is now fighting battles on all sides (in the CPU arena at least) and, at least right now, losing ground on most of the fronts. Of course if I were Intel I would be most concerned with Athlon SMP invading the traditional server and highend workstation market...but that's just me.
Re:What about a fully-static CPU? (Score:4)
Posted from a nfs-booted iopener, progress in disguise...
//Frank
Re: (Score:1)
Re: (Score:1)
Re:Intel's brilliant new idea (Score:1)
Re:No one can touch Intel (Score:2)
1. AMD has overtaken Intel on the "power-of-cpu" camp. This means that they're going to become more and more popular.
2. The law of exponential growth. AMD Should overtake Intel in less than 3 years. AMD's popularity is increasing by the day, while Intel is falling out of favour.
Re:Whooppee... (Score:1)
Re:Crusoe can only emulate x86 arch (Score:1)
even if the chips you can buy RIGHT NOW don't do it, it would require VERY minor changes in design and fabrication to "tune" the Crusoe to, say, a Sparc-emulating version. "Very minor" as opposed to the amount of effort required for Intel or AMD to design and fabricate chips to do the same.
I'll meet you half-way right here (though not to be an extremist of the middle -- and there are many of those around) and say that my knowledge isn't suffuent to disallow that it may be the case that the Crusoe could be modified to non-x86 use in a future iteration. (Sorry for the wordiness, but I'm trying to be precise.) In fact I have no direct knowledge of how the Crusoe works -- I'm analyzing what people say about it. I think it's important for the IT community to be realistic about the capabilities of individual technologies -- and self-destructive to be unrealistic.
What I read into Pratt's comment's about Crusoe's having an 80-bit FPU was that Crusoe would never be able to emulate anything else and that people saying that it could were jumping to uninformed conclusions about the possibilities of its code-morphing on the Crusoe platform. Perhaps I read too much into Pratt's comments and it's simply the case that the FPU can easily be changed -- not as easily as the software but easily enough.
Thanks for the interesting exchange. More is welcome.
Re:That's impossible! (Score:2)
My biggest complaint with computers is how quickly they respond to the users. Quite often, low-power mode is a terrible culprit here, especially when the HD spins down. You move the mouse, click on something, and wait 30 seconds while the machine pulls it's head out of it's ass and spins up the hard disk. If you do this often enough (if the cylce is short enough) it probably takes more energy to spin-up the disk that it would have taken to continue running it (newton's second law anyone?); and makes me sit and wait for my $5000 new spiffy super-fast 1.4 GHz PIV (exaggerating, but nonetheless, point is made, no?).
I'm not saying this is bad, but I'm saying that it's USUALLY implemented to make things worse. If the chip goes INSTANTLY from 300 to 500MHz the very moment I start moving the mouse after letting it sit, then it's okay. But latency sucks. Which is probably why there are now strong arguments that a 500MHz CPU is enough for most people. It's not, but when you pay the extra dough for the 1.4 Ghz one, and you don't see a performance improvement for MOST day to day tasks, the problem is your suffering from latency, and often reading data that was not cached. Poor implementation.
Re:Intel's brilliant new idea (Score:2)
Re:Whooppee... (Score:2)
Fucking Motorola better watch out.
My guess, is that Transmeta probably has a chance at making one of these deals EMULATE a PPC chip FASTER than the fastest G4 Motorola can currently produce. (which is the same speed they had 12 months ago).
i don't mind all that much (Score:1)
Re:this isnt just underclocking.... (Score:2)
competition is grand (Score:4)
Re:Crusoe can emulate any arch... (Score:1)
That would mean the chip can still do any instruction set just fine. You just need to tell the Code Morpher what to do with the incoming instructions for any other architechures you wish to support.
As for the link, I think your link was correct - the researchers own home page had the same link with the same problem! Give that the domain is "wearable.standford.edu", perhaps the web site was out at the cleaners... I'll check the link later to see if it's up.
That's impossible! (Score:3)
Is there?
--SpookComix
The Chips link is broken (Score:1)
Does anyone know the real URL I'm too lazy to hunt
Re:More ideas to reduce power consumption (Score:2)
4) You can never have enough speed. Not even on a laptop.
5) Uh, that would be hard.
Bad Link (Score:3)
Here's [lycos.com] the right one.
This begs the question (Score:4)
Yeah, but will they look like barf?.
--Jim
Re:Transmeta (Score:2)
Ummm... well... (Score:2)
Mark Duell
That's nice... (Score:5)
Re:At least Intel gives you a choice. (Score:2)
Intel's brilliant new idea (Score:4)
Intel calls it "Speedstep technology". I call it underclocking. This is nothing new. The *really* sad part about all this is that it's going to take another year for this *amazing new technology* to be available.
this isnt just underclocking.... (Score:1)
Re:At least Intel gives you a choice. (Score:3)
Finally, keep in mind, a year from now, when Intel finally rolls out their "new" technology, Transmeta will more than likely have a new surprise.
Re:LongRun vs. Speedstep (Score:1)
Re:At least Intel gives you a choice. (2nd try) (Score:1)
You're obviously confused: SpeedStep only offers you two different steps. Crusoe has a gradient (i.e. it can operate at a number of speeds between it's min and max speeds) and is much more intelligent about deciding what speed to run at. If not much is happening, it drops down a notch or two. If it needs more cycles, it hops up again. With Intel, it's all or nothing. And I think we all know that comparing raw clock speeds (MHz) is just crap. YES, a < 600MHz Crusoe, performs on average more like a 600MHz Pentium (though it's more like a P3 450 or 500, I don't know where you got that P2 250 figure), but it's a trade-off between speed and power consumption/heat generation. CURRENT P3's with SpeedStep (let's not forget, this "new" thing won't be available for a year) may give you a minor advantage, but in exchange for that 100MHz drop in speed you're still only going from 2-2.5 to 4-5 hours on a standard battery; Crusoe machines can run 10-12 hours! And yes, some of that is due to smaller monitor, etc., but there's no way getting around that P3 w/SpeedStep (either now or this "new" chip) loses hands down to Crusoe in the performance vs. battery life arena.
Finally, keep in mind, a year from now, when Intel finally rolls out their "new" technology, Transmeta will more than likely have a new surprise.
Re:Well, it depends [This begs the question] (Score:1)
Re:this isnt just underclocking.... (Score:2)
Re:intel is losing it... (Score:2)
Re:LongRun vs. Speedstep (Score:2)
Intel doesn't have a prayer at catching Crusoe in the immediate future on power consumption. The best that they can do is get it low enough that the monitor is the main limiting factor for battery life. The smaller die size and integration of some of the memory interface components on the crusoe could mean that the cost savings become the greatest benefit of the technology.
Re:Intel's brilliant new idea (Score:1)
You don't think Intel is sending a message to the machine manufacturers do you?
Re:That's impossible! (Score:1)
Re:That's impossible! (Score:2)
Now that Tranmeta has produced a chip that threatens this sub-market, Intel actually has a reason to produce chips that aren't so power hungry. The expense of production with hitherto dormant technology is offset by not loosing (as much) market share to this upstart company.
`ø,,ø`ø,,ø!
Re:That's impossible! (Score:1)
The Crusoe can dynamically change it's clock speed and power consumption i.e. if you are playing solitare, there is not much demand for power from the procesor so the clock speed will be reduced.
That means if the benchmarks mentioned here [slashdot.org] a while back when the new picturebooks were announced, that found a crusoe 600MHz to run at between a P3 400 and P3 600. That makes the P3 500 that slows down to 300MHz when on battery very very LAME, not really that nice IMNSHO!
How every version of MICROS~1 Windows(TM) comes to exist.
Re:first prost (Score:1)
I'm sure they were, since that (sort of) translates to "Hello! My CmdrTaco flame!"
Re:At least Intel gives you a choice. (Score:1)
Re:Crusoe can only emulate x86 arch (Score:1)
Crusoe can emulate any arch... (Score:1)
However, I remember reading about demos where it was running Java bytecode directly. Here's an extract from one review:
After this Q&A was complete, we moved to the other hall to look at the goods. There were a number of machines running the TM processors. The most impressible demo was a chopped up version of doom that ran it's main loop as PICO Java and the graphics routines and such at an x86 machine. There was a graph above the screen where it showed the machine swithcing between instruction sets in the middle of the VM.
The page I got that from is here [dibona.com].
AMD (Score:1)
Re:At least Intel gives you a choice. (Score:1)
Personally, I am a lot more impressed by AMD's Power Now! [slashdot.org] (perhaps not by the name, but the idea) than Intel's SpeedStep. AMD has implemented dynamic clocking based on usage for the mobile K6-3's, and I hope to see it in the mobile Athlons. This to me is a much better approach.
any new transmeta chips (Score:1)
-MSD.dyndns.org [dyndns.org]
"Sucks to your ass-mar"
[OT] Chromatic wishing for a miracle or three (Score:1)
What, relying on miracles at the process level wasn't more significant? :-)
As I recall, they were doing all sorts of out-there things, such as a metal layer on the die that was "in the air" -- the rest of the surface was etched away around it to reduce sidewall capacitance or somesuch. Some aspects of their design were far too researchy to be viable for production, IMHO.
But, then, I'm just a partially informed spectator, so what do I know?
--Joe--
Re:Whooppee... (Score:1)
You should _read_ the post before replying. All parts of it.
First: I talk about Mac OS X *Server* which is *NOT* a beta operating system.
Second: The 16% figure in the first paragraph is not disk-based. It is clearly said that 'the very same (real life) code is faster on the Athlon,'. Real-life-code means compute intensive code with about zero call to the Operating system. I agree that it depends on compilers, but the very same (The gcc derivative furnished by *apple*) compiler was used with the very same options.
Third: The file system perf paragraph is not based on a beta implementation of anything. I hate NT, but the startup boost on NT is very very noticable. I admit that it does not say anything about G4 (the processor). That is why it was not included in the 16% figure and deserved a different paragraph.
Fourth: I watched the jobs demo. I *was* there. I attended to every Jobs demo in France since 1991, if this gives you a clue of anything. I know what jobs demo machines are. I touched some. (At this time there was a couple of dozen of person attending to the demos. It was even easy to talk with him before/after the demos)
Fifth: When Jobs did the Photoshop demo, it weas clearly to prove that G4s were better than PIII. Not that altivec-enhanced-photoshop-on-probably-hacked-con
I welcome you to re-read my post.
I would love beeing able to say that G4 blows off K7/PIII, because I hate IA32 architecture. Ppc assembly is a beautifull thing, from an engeneering point of view. I deserve to kick ass of bloated IA32. But it don't, in my personal experience.
Cheers,
--fred
Re:competition is grand (Score:1)
They would ship five in 2002.
So it's... (Score:1)
--
Re:That's impossible! (Score:1)
Re:competition is grand (Score:2)
Later they would be recalled
At least Intel gives you a choice. (Score:3)
Speedstep is NOT a NEW thing (Score:3)
Re:a bit confused (Score:2)
Anyway, did you do an EMC test on the heaters?
If it's throwing off heat, it's either throwing off light or sub-luminal frequencies?
In fact, just an "on" light would be good enough to destroy their claim!
Phil
AMD's PowerNow! (Score:2)
It's already available in the K6-2/3+ line and should be available for the Athlon laptops sometime early next year.
Unlike Intel's technology, PowerNow! users a gradient technology (though certainly not quite like Transmeta's level) instead of just fixed speed steppings.
Gotta hand it to Intel's marketing department (Score:3)
Re:Ummm... well... (Score:5)
Mark Duell
a bit confused (Score:2)