Intel's Competitor to the Crusoe Processor 115
TJ6581 writes "C|Net news has an article up about Intel's new competitor(s) to the Crusoe Processor. Apparently the new chip uses half a watt of power and did not require a major re-design. Also mentions in the article that IBM will be using this processor in the notebook originally designed for Transmeta's processor." Update: 01/30 06:48 PM by H : This is the update to the story I posted last October - Intel has come through.
I love it! (Score:1)
Regardless of what you feel about Intel (Score:4)
...and the geek shall inherit the earth...
Re:Transmeta's effect (Score:1)
There's only going to be competition if there's a chance that you can win. However, the general sentiment (especially around corporations) is "Thank goodness these alternatives exist. If they didn't, we'd have to spend much more on our industry standard stuff."
Don't expect to the reap the benefits of competition forever if you have no desire to look at a competitors product. Of course, that's hardly a warning that Slashdot-ites need to hear
Re:Not exactly powerful (Score:2)
Where did you get this information? MHz and Watts have no direct, mathematical link between each other. Who did your math?
My calculation for a standard 300Mhz processor is way over 1 watt, at 14.3 Watts.
P=VI; 4.33A (Socket 5) * 3.3V = 14.289Watts
Re:Not exactly powerful (Score:3)
It has everything to do with frequency. Where do you think those 4.33 amps come from? Modern processors are made using a process called CMOS (Complementary Metal-Oxide Semiconductor), which is just a fancy name for two different types of transistors connected together: an N-FET and a P-FET. FET == Field Effect Transistor. The N- and P- refer to whether the transistor turns on or off when the gate voltage is above the threshold.
A P-FET turns off when Vin > Vt and an N-FET turns ON when Vin > Vt. If you connect a voltage to the gates of both an N-FET and P-FET, and connect the FETs together, you get a nice situation where there is very little wasted voltage (prior to CMOS, people used NMOS, which used a pull-up resistor instead of a PFET, which meant that power was always being drained in the resistor).
However, there's a problem. The transistors can't switch from "on" to "off" instantly. There's a tiny bit of time when both transistors are in tho "on" state (roughly corresponding to a short across the source/drain), meaning that current can zoom straight through both transistors to ground, which is bad. This is where you lose power.
Now, you do the math: if a simple NOT gate (which is just two FETs as described above) is switching at something like 1 time per second, you don't lose very much power. But if instead it's switching a million times a second, you lose one million times that previous value. And if it's switching at a billion times a second (one gigahertz), then you're losing a billion times that much power.
You should learn a bit more about what you're talking about before you attempt to lay down the smack.
Intel vs. Transmeta (Score:4)
they did is produced the chip that works fast when the notebook is used as
desktop computer, with power adapter on. As soon as the power adapter is
unplugged, the chip goes to being slow.
Besides, really nice technology about Crusoe is that it can detect when
user is having high demand on its laptop, and speed things up, then switch
to low consumption when the computer is idle (like while you're watching
something on the screen, not doing much). Intel has not developed such
technology, it might, but not this year. What Intel is dumping the market
with slow processors that consume less power, something expected for a
while now, and make it sound like they're onto something big.
Quite an expected low punch. Good marketing and excuse for US PC makers
to offer slower laptops for higher price. It will take a deployment of
Crusoe based products, such as new Sony laptop, to debunk the myth. Will
the Transmeta survive? If they have enough money and one or two faster
chips for this year, I would think in deed.
Plus, Intel plans to produce those chips in a year. A long time, specially after delays.
Re:Intel will kill it (Score:1)
Re:Intel colluding with IBM? (Score:1)
Re:Sure (Score:1)
The only drawback of course, is when the big company doesn't compete on merit but uses behind the scenes coercion. Intel AFAIK has gained marketshare by luck (IBM PC) and going for good, but not risky solutions, and executing well enough.
I'm not sure you would want an Intel that was innovative, took huge risks, and never failed to execute. I'm sure AMD and Transmeta stockholders don't want that.
Re:Intel will kill it (Score:2)
Re:Not exactly powerful (Score:1)
Intel vs Transmeta vs AMD (Score:2)
Look at the Pentium 4 chips. They have flopped from the begining, and, according to a few sources, there are several architecture flaws and set-backs still within the chip.
AMD has obviously capitalized on the giant's fall by releasing lower speeds in MHz that out-perform faster chips. And AMD's 64-bit chips will show the same results. EV6 architecture is better than what Intel is doing.
Maybe it is just time that Intel quits pulling the Microsoft and work on improving their current collection.
MunITioN
Re:Not exactly powerful (Score:1)
Reduction of feature size greatly reduces C, and will also permit a reduction of Vdd.
So yes there is an almost linear mathematical relationship between frequency and power as long as you dont change the other significant factors (Vdd, a, Cl). However as the delay of a gate a bit simplified can be modelled as t=kCl/BVdd where k is a constant typically at around 3, Cl is the gate load and B the conductance of the open circuit to Vdd or Vss (ground). B depends on transistor dimentions (higher B means bigger transistors, which means higher Cl for the preceeding gate).
Thus when you lower f you can exploit this by lowering Vdd and/or reducing C. (Its then generally preferable to lower Vdd until you reach the limits of what you can support, then reduce C to speed, this due to the quadratic relationship between Vdd and P)
Re:No worries yet... (Score:1)
Re:Not exactly powerful (Score:2)
Runs real good on those Transmeta chips, too.
More seriously, the chips are good these days, but bloatware sucks back the advances. If you want fast, try running some of the older stuff ... such as WP 5.1 on DOS, on a PIII. Your problems stem from one source - M$. Win2K? Bloat. Lousy memory management, but worse, 50 gazillion background processes sucking back cycles (OK, the number is closer to 50, but the point still stands) MS Office? Bloat. Doesn't even matter which version. M$ Outlook? Bloat^nth, on both PC (and Server, if your using one), and the biggest security hole to come out of Redmond. RTF? WTF! Use text or HTML, but avoid M$ formats. It also kinda sucks that certain M$ progs check files for M$ tags before loading ... and they load significantly faster if the tags are present.
Re:Transmeta needn't worry. (Score:1)
Pardon for being slightly off the thread, but can you think of any successful company that isn't marketing driven? I'm a tech person, not business, but it seems that every tech company has to make the switch from engineering driven to marketing driven in order to be successful (success doesn't always equal income though).
Anway your above phrase just struct me as odd..
Re:With Intel's Current Track Record... (Score:1)
Mike
Re:Intel vs. Transmeta (Score:2)
user is having high demand on its laptop, and speed things up, then switch
to low consumption when the computer is idle (like while you're watching
something on the screen, not doing much). Intel has not developed such
technology, it might, but not this year
Bullshit. It's called the HLT instruction and any decent OS executes this instruction when idle. It's been in Intel CPU's for ages and works excellent to conserve power: as long as you run a non-brain-dead OS.
When Transmeta publishes figures that actually show the power consumption when the CPU IS busy, then we'll see how far they actually are ahead of comparable speed Intel CPUs.
Breace.
Re:Transmeta needn't worry. (Score:1)
Clearly you've forgotten about Intel's XScale [xscale.com] architecture (the successor to the StrongARM [intel.com]). The ARM [arm.com] processor currently holds a HUGE [arm.com] segement of the embedded market, and Intel is promising the same low-power high functionality technology at speeds up to 1GHz in the near future. If anything, their presence in the embedded market is growing.
Re: (Score:2)
Re:Sure (Score:1)
Awriiite! (Score:1)
Alright, the first 100.000% x86-compatible CPU that consumes half a watt! I don't care if Linus isn't on the development team; he's probably half of Transmeta's troubles.
No matter what will happen in the laptop market... (Score:1)
uh....yeah (Score:1)
Re:Not exactly powerful (Score:1)
I'll bet that isn't far from the truth. Sounds like exactly the sort of thing they would do, and then turn around and tell the public that it's slow because those other, non-MS formats are "inferior" and just naturally take a long time to process.
Wrong formula (Score:1)
You need to use this one: P=CFV^2
P = power
C = capacitance
F = frequency
V^2 = voltage squared
As you can see, power varies linearly with frequency. Actually, it's better than that: at lower frequencies, you can use a lower voltage, so you get a kind of x^3 relationship with frequency.
--
Patrick Doyle
Java bytecodes (Score:2)
--
Patrick Doyle
marketing FUD (Score:3)
Re:Not exactly powerful (Score:1)
Hardly revolutionary, no, but well worth the savings on battery life.
Re:Not exactly powerful (Score:1)
Re:DAMMIT (Score:1)
Intel colluding with IBM? (Score:3)
I'm very suspicious of this flow of events... Intel hears that IBM is thinking about using a chip by another possible competitor, sends a representative over to do some arm twisting, bribing, etc. etc... "Just wait long enough for us to get our low-power PIII out the door..."
Pheeeew. That's got quite a stench to it.
--Fesh
Re:Intel will kill it (Score:2)
It's no longer a matter of hardware catching up to software needs but vice versa. We have so many idle CPU cycles (your welcome, SETI) but we're wasting electricty in CPU cycles, and then we waste just as much energy on cooling the heat that those CPU's generate.
I also agree that Intel will screw it up as they usually do. I mean, lets face it -- Every 1st generation Intel CPU release is about as reliable as an x.0 release of RedHat, so just work on making what already exists better and more efficent.
I think the Internet community would thank you for it.
"Of course that's just my opinion, I could be wrong"
--Larry Reckner
larryr@linux.com
Crusoe was not designed for low power (Score:2)
Now they are paying the price for not exploiting all the benefits of the technology they developed. I hope for their sake that they can find more "markets" for code morphing or they will end up just another company crushed by the Intel goliath.
Time Zone / Planet (Score:1)
It's now 14:25 EST. If Hemos is still on this planet, he must be in the U.K. or Portugal. Way to go Hemos, such dedication whilst on a European vacation is not common to see these days!
Re:Sure (Score:1)
What, compete?
marketing (Score:1)
Re:Not exactly powerful (Score:1)
Re:Not exactly powerful (Score:1)
In a CMOS circuit, the electricity involved is static (non-changing) and does not require a current flow. For those interested in electronics design, this is why you need pull-up resistors when interfacing CMOS to TTL. You need to explicitly supply a return path for the voltages, because the chip doesn't connect them through normally.
excellent use of fallacy (Score:2)
And he was probably correct about the intel chip running at 300MHz consuming less power than the Transmeta chip.
But I have to ask: Why didn't he combine the two?
How well does the intel chip perform in castration mode? And how much power does it consume at 500MHz?
Of course that's a moot point, because you can only run it at full speed on AC power, even if you wanted to on battery power.
I'd suspect that the Transmeta chip would run circles around the intel chip (laughing the whole time) if they were benchmarked on battery power. And what's the use of running 20% longer if it takes you 120% more time to do anything?
I'd rather have five hours of good performance! (And I'd get a spare battery pack.)
Re:Sure (Score:1)
Intel stockholders would want that.
If the Pentium could converted to low power so easily, there is not much risk in doing so is there?
Re:Sure (Score:1)
Good one!
Re:Intel will kill it (Score:1)
I'd love to see an 8-way Cursoe server -- in a mid-tower case. This would never be possible with a Pentium.
I think this is what Transmeta has had in mind all-along.
Dom
Of course it didn't take much of a redesign (Score:2)
Intel won't compete w/ transmeta until... (Score:1)
Intel's Problem (Score:1)
Re:Not exactly powerful (Score:2)
Nice sig, too bad it goes both ways. Try these analogies:
If Windows was a beer...
it would be sealed shut so that you have no idea what they put in your drink.
On the side of the keg there is a warning saying that 1 in 5 kegs explode when you tap them.
Some of the beers you pour will be flat forcing you to dump out your beer and pour another one.
The beer tastes like shit, it makes you sick and lots of people laugh at you for drinking it, but you keep doing it (oh wait, that was my Budweiser analogy).
Re:Not exactly powerful (Score:1)
Server market (Score:1)
I say start shipping a bunch of these down to Silicon Valley to free up some electricity so we can read Slashdot all day long.
Re:Not exactly powerful (Score:3)
Actually, for CMOS, Power = C(Vdd^2)*f
with Power in Watts, C (Capacitance) in Farads, Vdd in Volts, and f in Hertz.
Basically, this equation says to lower power dissipation, we can lower Vdd (power supply), lower the capacitance, or reduce the clock frequency.
Intel will kill it (Score:1)
Re:Low Power - So Watt?... (Score:1)
The code morphing layer is indeed revolutionary, but what practical - profit-making - applications does it have right now? If people want x86 compatibility they buy an Intel or AMD CPU. One possibility I can think of is running native PowerPC (or even SPARC?) stuff - like a PC and a Mac in one - but surely that would require significantly more work on their code morphing.
Re:Not exactly powerful (Score:1)
"Where did you get this information? MHz and Watts have no direct, mathematical link between each other. Who did your math?"
I actually didn't mean to word it as an insult, but I just recently quit smoking. 8-).
I was honestly wondering what the Eqn was. I'll be the 8th to admit I was wrong. Thanks to everyone who wrote me personally to call me a flaming retard. 8-)
yikes (Score:2)
this really hurts transmeta, it's probable time they start expaditing the code morphing feature and not the power consumption. I would love to have a system capable of running multiple codes based on different architecture, but to be realistic whats the point. will we ever escape the evil clutches of intel ?
Does this strike anyone else as funny? (Score:4)
An Intel spokes-droid said today "Since we've utterly failed to compete with AMD, we've decided to set our sights lower. Transmeta, watch out! We're taking your nonexistant market share!"
Silly Intel, just make some decent chips.
Not Vaporware, just underclocked. (Score:1)
Well, it looks to me like this is just a Guyserville Pentium III (i.e. Mobile Pentium III with SpeedStep) that has been UNDERclocked to 300MHz. If you find a way to underclock the exisiting ones that far, I bet they'd only need half a watt, too...
Yeah. Head down to CompUSA or Fry's, and check out the Sony 'PictureBook' PCG-C1XN. It's powered by a Crusoe processor, and I've seen it in stores for over two months now.
No... More likely by Via. They've been the ones on a processor-company buying spree. AMD seems to want to do it the hard way.
Not exactly powerful (Score:4)
While it does have low power consumption, it's not exactly brimming with processor power. To get below 1 watt, they have to drop the speed of a PIII to just 300 MHz. Of course that's still plenty to run office programs or an mp3 player, but it's hardly revolutionary.
With Intel's Current Track Record... (Score:2)
*ahem* (Score:2)
--
MailOne [openone.com]
DAMMIT (Score:2)
--
MailOne [openone.com]
Re:Yeah but... (Score:2)
Pardon my ignorance, but how about the "moving parts" of a laptop? I can imagine the harddisk taking up quite a bit of power as well, which could become very frustrating as you frantically try to save your hard work before you're out of power...
Copper?? (Score:2)
I thought Intel has had horrible trouble making chips with copper? Is this true? If so, what can we expect from these chips? Isn't this why Intel is having so much trouble competing with AMD?
--
Re:Transmeta needn't worry. (Score:1)
Re:Not exactly powerful (Score:1)
Change that to read benchmarks instead of real applications and you might be right. When you start talking about real applications, where you actually do the same thing again and again rather than switching tasks as fast as possible, a lot of the performance disadvantages of the Crusoe go away.
That said, I'm not terribly worried either way. My workplace still uses PII 400 based desktops, and I have a PII 366 laptop, and they seem to be just fine for everyday use in a business setting. PIII 300 class performance is likely to be plenty for most applications anyway. It's not like you're going to be trying to set Seti@home records or do heavy-duty compilation work on your laptop.
Re:Not exactly powerful (Score:2)
From the article:
On average, the new Pentium III consumes about half a watt of power, or less than half the power of Intel's current notebook chips.
...and...
Like other mobile Pentium IIIs, the new chip contains SpeedStep technology, which lets the chip run at a slower rate when operating on battery power. The chip runs at 500MHz when plugged into a wall and at 300MHz on batteries.
...and...
Transmeta countered Intel's assertions, noting that Intel's new 500MHz chip runs at 300MHz when the notebook is unplugged.
"This is a very slow processor at 300MHz when mobile," a Transmeta spokesman said in an e-mail. "That's a big sacrifice to get to lower power levels. Transmeta is about double the megahertz at the same power levels."
Re:Yeah but... (Score:1)
do you think that many people outside the Linux and programming communities even know who Linus is nevermind buy one product over another because of him?
Jon
Re:Intel will kill it (OT - karma be damned) (Score:1)
The fundamental error in thinking was that the existing power infrastructure must be healthy since it reliably provided power for years. Well, the US system is overall fairly healthy, but the system in CA is not. Years of regulation and nimbyism caused a severe lack of generating capacity and led to reliance on the general health of the overall system.
Low Power - So Watt?... (Score:2)
What is interesting about their chips is the abstraction layer and the benefits it provides. No longer is it impossible to update chips to fix flaws in them and no longer is it necessary to have a single vendor for whatever chip the software was designed for. The fact that Transmeta's chips happen to draw very little power doing it is really just a nice byproduct. I think Transmeta is doing itself a disservice by harping on this fact.
Granted, low power consumption is an easier concept to communicate but I think it really misses the bigger picture. In the same way that higher level languages made software development easier than coding in assembly, Transmeta is taking the next step in abstraction. Easing software design barriers is what is truly interesting and revolutionary about what they are doing.
Re:No worries yet... (Score:1)
AFAIK Transmeta used something like two years perfecting their code morphing software for x86 emulation only. Emulating hardware like AMD's or Intel's 64 bit one isn't that easy to make effectively. However, I wouldn't be that surprised if Transmeta had SSE2 support in the near future because it should increase floating point performance greatly. AMD is doing that too. Though if there isn't good support for floating point calculations in hardware I wouldn't expect that much...
If their code morphing software is easily portable to other CPUs it would be interesting to see it's performance on something like Alpha or G4. Of course those don't have required hardware to make emulation effectively compared to Crusoe chips.
_________________________
not surprised... (Score:1)
give smart people alot of time and money and they replicate
something that already exists, shocking.
peace.
Re:Not exactly powerful (Score:2)
So, this is why they killed Tinma? (Score:1)
Re:Not exactly powerful (Score:2)
Then once Transmeta is safely in Chapter 11, Intel can stop revving these low power CPUs, and go back to 60-watts-in-a-laptop-set-specs-for-drive-and-disp
Re:Not exactly powerful (Score:2)
It's as if Microsoft coded the translation app with a lot of WaitAndPunishUserForUsingNonWordDocumentFormat() loops.
The Word window is just sitting there, hourglassing away. Fuckitall.
BAD reporting, interesting story (Score:2)
However, the writing in this story hit a new low, even for C|Net. It seems as though the reporter couldn't justify transcribing the entire press release, so he added a few sentences of his own. Aside from that, it was pure and utter pro-Intel, end of Transmeta, drivel.
C|Net has never had great reporting, but this _stunk_.
Re:Transmeta needn't worry. (Score:2)
Also, as Transmeta reduce feature sizes, they should be able to break into the embedded market at some stage over the next few years. I fear for Intels future, in this regard.
why do people keep gunning for Transmeta to go into the embedded market? why do you want antiquated x86-based chips in the embedded market anyhow?
x86 is still kicking because of backwards compatibility for PC applications. the embedded market doesn't need this kind of handicap! there are plenty of good embedded processors, such as PowerPC and MIPS, that are considerably better designed to meet the needs of true embedded applications.
now if you're talking about a Transmeta chip that's not simply x86 compatible, perhaps with more direct access to their "true" instruction set, that's different. but embedded devices shoudn't be cripped by a processor that's better designed for other tasks. use the right processor for the job!
- j
Re:No worries yet... (Score:3)
I don't have any inside info. on how portable the code morphing is, but implementing a new instruction set (not x86) or porting code is bound be easier the second time around.
They've got the experience from the first attempt, and that makes a huge difference.
gotta love the burnmaker (Score:1)
----
Re:Intel vs Transmeta vs AMD: forgot PowerPC! (Score:2)
AMD for now has a huge performance lead, at the sacrifice of power. Intel has a huge marketing and manufacturing lead, with slightly less power and slightly less performance.
Transmeta wipes the floor with power, at the sacrifice of performance.
How bout something in between?
PowerPC, with good power, and good performance?
Geek dating! [bunnyhop.com]
Re:Not exactly powerful (Score:2)
It's just fucked.
Re:With Intel's Current Track Record... (Score:1)
What the hell are you talking about? The K6-2 was an excellent product (I'm still using a K6-2 system and have never had any problems with it). Oh, and let's not forget that it was the product that prompted Intel to release the original Celeron.
You might also want to think about the present -- AMD's CPUs are the fastest x86 CPUs available and they cost far less than Intel's. If any company has made a bad name for itself, it's Intel (paper launches, P4, 1.13GHz PIII recall, rambus, bugs, bugs, more bugs... anything I've forgotten?).
Re:Not exactly powerful (Score:1)
Not true, it's always easier to implement established architetures.
Intel can stop revving these low power CPUs, and go back to 60-watts
Not so sure. There's a good possibility that Intel may have made a bad move by tipping it's hand, now that it really CAN offer an excellent power/performance part. Of course, Intel could always flip the bird to the OEMs who established these low power designs, but not if there's a foothold from parasitic competition from AMD and Cyrix. We'll see.
Don't you just LOVE the free market?
Of course I do. What are you suggesting?
---
Re:Transmeta needn't worry. (Score:1)
Re:Sure (Score:1)
Re:Intel colluding with IBM? (Score:1)
Just keep in mine the relationship between MSFT and Intel. Now picture them in a little street gang that is waiting to initiate big blue into their posie. Transmeta is just becoming a victim of MSFT's "class bully" tactics. I just hope they can hang on.
Who knows? Remember that "AMD to acquire Transmeta" thingy?
Whats the TYPICAL USE real numbers? (Score:1)
But what's the real average power consumption for the average websurfing user of the 2 chips?
Anyone in a position to test them.
Transmeta needn't worry. (Score:5)
Also, as Transmeta reduce feature sizes, they should be able to break into the embedded market at some stage over the next few years. I fear for Intels future, in this regard.
Transmeta have a bright future, despite this news.
You know exactly what to do-
Your kiss, your fingers on my thigh-
Re:not surprised either. Not (Score:1)
So it's true (Score:2)
Re:Not exactly powerful (Score:2)
This shows the benefits of competition working in a market driven society: Transmeta forced Intel to offer a low-power chip, which they really had no reason to do until now. Of course, Intel's offering will basically demolish Transmeta b/c OEM sol'ns for PIII are off-the-shelf simple, not to mention a barrage of other concerns (reliability, market proven, brand equity...)
---
Re:Not exactly powerful (Score:2)
To get below 1 watt, they have to drop the speed of a PIII to just 300 MHz.
Well, keep in mind that this is very possibly about the same performance as the Crusoe will end up being in real applications.
--
Transmeta's effect (Score:3)
-Chris
...More Powerful than Otto Preminger...
Re:I've always suspected ... (Score:3)
"Oh, look, the smaller feature size also made the chip smaller! Damn those mathematicians, they were right again!"
----------
Re:No, but THIS is: (Score:2)
More info in The Register (Score:2)
No worries yet... (Score:3)
Transmeta should be able to improve on their initial products as well. We're used to this kind of one-upmanship with AMD, I see no reason why Transmeta can't do the same thing.
But aside from that, I doubt that Transmeta's entire business plan is based on low-power. The strength of their technology is flexibility. It would be cool, for example, if they could produce a laptop version of AMD's 64-bit chip, or Itanium for that matter without a major re-design. While Intel needs years to ramp up a new, expensive hardware architecture, Transmeta can copy it in a few months cheaply.
Re:Hard times for Transmeta... (Score:3)
As for whether this is the end of Transmeta, I guess it depends what they do with their product line from here - The concept is really great, and should allow transmeta to bring out improved models fairly quickly. The question is
I think that is what will define whether Transmeta survives or not.
Re:Copper?? (Score:2)
Re:Transmeta needn't worry. (Score:2)
I don't want a x86 in my embedded product, but lots of people who design them do. Either because lots of tools work with x86s, or it is easy to find people who can do x86 code, or x86 drivers exist for devices they want, or because that is what their boss wants.
Transmeta shouldn't argue with those suckers, it should take their money.
I've always suspected ... (Score:2)
question: is control controlled by its need to control?
answer: yes