Comments on: Intel Details Future Mobile Chip Plans

News.com is reporting, Intel has formed the Low Power Intel Architecture project, or LPIA, to try to devise energy-efficient versions of its x86 processors, software and components for handhelds, micro PCs and other devices. More details on the LPIA will be revealed at the Intel Developer Forum next week, sources said.
Return to Story - Permalink

Article Comments

 (15 comments)

The following comments are owned by whoever posted them. PalmInfocenter is not responsible for them in any way.
Please Login or register here to add your comments.

Start a new Comment Down

Look Out

archie@PIC @ 8/18/2005 4:22:55 PM # Q
Look out Palm. Apple is further than you and more innovative too.

Plus, they are going to need as much speed as possible running Windoze Mobile 5 with NVFS.

Reply to this comment

x86 better than current ARM chips? Yeah, right!

interlard @ 8/18/2005 9:50:49 PM # Q
I find this strategy strange. The ARM processors (eg, Intel's XScale) that appear in Palm and MS handhelds & smartphones are based on ARM designs for compact, efficient, RISC processors. Intel's x86 chips come from the mid-1970s.

Will handhelds based on these x86 chips be able to play normal MP3s, or just disco?

RE: x86 better than current ARM chips? Yeah, right!
PenguinPowered @ 8/18/2005 10:33:03 PM # Q
Moore's law is Intel's friend. If you follow the history of computing hardware, from supercomputers down through 'workstations' you already find that eventually the reasons that made purpose-designed processors attractive at the start tend to be overwhelmed by economies of scale in other areas.

This is now starting to impact so-called 'embedded' devices, although, as of yet, it's only visibile in 'minaturized pc' devices like the current generation of smartphones and pdas.

To win design competitions, Intel doesn't have to have the best processor match to the power/performance/price characteristics of the device anymore. They just have to have one that's good enough and is supported by compelling non-processor reasons to go with it. They're not there just yet, but one can seem them in the rear view mirror, gaining, if one is involved in processor design.


RE: x86 better than current ARM chips? Yeah, right!
Sam H @ 8/19/2005 8:24:48 AM # Q
Moore's law works for ARM too. The power/performance/price level that Intel has to reach to be 'good enough' keeps on moving away from them.
RE: x86 better than current ARM chips? Yeah, right!
PenguinPowered @ 8/19/2005 11:44:14 AM # Q
Moore's law is not Intel's friend because it enables them to compete head-to-head with purpose-built processors. It is Intel's friend because, eventually, Moore's law enables Intel processors to scale down to a "good enough" status. Once Intel processors are technically acceptable, then other factors dominate design decisions and raw processor cost benefits become trivial in the overall equation.

Basically, in the end, Intel wins because of economy of scale. That's how they came to dominate the 'workstation' CPU market place and drive out most of the RISC cpu vendors.

Outside of the embedded world it's been fifteen years since the CPU was a big enough part of the cost budget for a device that optimizing it was the key cost/performance tradeoff. It looks like inside the top end of the embedded world that's true now.



RE: x86 better than current ARM chips? Yeah, right!
Sam H @ 8/19/2005 1:48:13 PM # Q
The power/performance/price level that Intel has to reach to be 'good enough' to sell in the embedded market isn't fixed. It rises as ARM follows Moore's law. In order to come close enough to ARM to become 'technically acceptable', x86 processors would have to advance even faster than ARM's AND reduce their power consumption by an order of magnitude. Neither of these are things that Intel has had much success with in the past.
RE: x86 better than current ARM chips? Yeah, right!
PenguinPowered @ 8/19/2005 6:16:50 PM # Q
You're looking at the problem as if the relative CPU price/performance was the only factor. It's not. It's not even the important factor any more.

But even so, in the past few years, Intel has started getting their act together wrt to low power parts. Meanwhile, ARM's struggling architecturally. It would not be at all surprising to start seeing Intel design wins in high-end embedded gear in a year or two.

If an Intel process is just good enough to make the design budget, then the decision makers move on to other factors that designs are sensitive to, even if there's an Arm processor that's at a much better price/performance point. Intel tends to win those sort of design competitions, and has done for rather a long time.

If Intel decides to compete head to head with ARM, expect them to be a viable competitor in 18 months to two years.

RE: x86 better than current ARM chips? Eventually...
The_Voice_of_Reason @ 8/19/2005 10:53:57 PM # Q
The power/performance/price level that Intel has to reach to be 'good enough' to sell in the embedded market isn't fixed. It rises as ARM follows Moore's law. In order to come close enough to ARM to become 'technically acceptable', x86 processors would have to advance even faster than ARM's AND reduce their power consumption by an order of magnitude. Neither of these are things that Intel has had much success with in the past.

But are the DEMANDS made of these processors also rising proportionally? Just as desktops are now FAR more powerful than the average user needs (who NEEDS a 2 GHZ CPU to check email + surf the Web???), unless the need is real, the target doesn't necessarily advance despite the competition advancing. Remember, as recently as 3 years ago many users were happy with 33 and 66 MHz PDAs. Extra speed is not always necessary (or even A Good Thing) - as Sony's HHE has shown conclusively.

As much as it disgusts me, I have to agree with Mr. Fouts on this - I wouldn't be surprised to see Intel get serious about a potentially huge market and move in + take over (Microsoft-style)...

TVoR



------------------------
Sony CLIE UX100: 128 MB real RAM, OLED screen. All the PDA anyone really ever wanted.
------------------------

The Palm Economy = Communism™
The Great Palm Swindle: http://www.palminfocenter.com/comment_view.asp?ID=7864#108038

RE: x86 better than current ARM chips? Yeah, right!
Sam H @ 8/20/2005 6:39:21 AM # Q
If I understand your argument correctly, you're saying that x86 compatibility ultimately overrides all other considerations. If it is the case that x86 compatibility is the most important factor then I'd agree that Intel will eventually dominate the embedded market.

But I don't think that is the case.

First, in mobile devices power consumption is a hugely important issue which Intel has been grappling with unsuccessfully for years. The physical constraints of a mobile device (temperature, battery size and weight) are at least as important as price/performance economics.

Second, the break from x86 has already been made. WM, Palm OS and Linux all run on ARM hardware. In the past Intel could have tried to convince developers of the danger of breaking with x86 compatibilty, but now the developers have done it the reasons to go back don't seem that compelling.

But are the DEMANDS made of these processors also rising proportionally? Just as desktops are now FAR more powerful than the average user needs (who NEEDS a 2 GHZ CPU to check email + surf the Web???)

The desktop is a good example. Common sense would suggest that supply (processor price/performance) should follow the demands of the software end users use. But it seems the reverse is true. Faster processors become available and then people find new uses for them. As ARM produce faster processors, I'm sure software developers will find lots of cool new things to do with them that will become the killer apps of the next generation of mobile devices.

RE: x86 better than current ARM chips? Yeah, right!
PenguinPowered @ 8/20/2005 1:41:28 PM # Q
Sorry, I haven't been clear. It's not x86 compatability. It's the economics-of-scale that Intel can bring to a low powered x86 processor design because of their massive experience with the same instruction set on a large number of processors.

Here are some old numbers (mid 90s) from competition in workstation land that highlight the point. In those days it cost around 250 million dollars to do a new generation of a cpu architecture. (Think arm v5 to arm v6) Intel spent that sort of money on the PII to PIII upgrade for example. IBM is reported to have spent similar amounts on similar changes. But, Intel sold several billion PIII parts whereas IBM sold a couple million of theirs. So IBM has to add a larger price-per-unit than Intel in order to recover their costs. (This, in a nutshell is also why the Unix-based OSes are all disappearing to be replaced by windows.)

So it's not x86 compatibility that makes the win for Intel, it's x86 economy-of-scale.

As to the power consumption stuff, you're absolutely right, power consumption is key in hand helds. But Intel has gotten better at that recently, and they'll get even better if they commit to it. They stumbled a couple of years ago by concentrating entirely on the high end and missing the emergence of laptops with their heat and power issues. They learned from that and have caught up.


RE: x86 better than current ARM chips? Yeah, right!
The_Voice_of_Reason @ 8/20/2005 4:59:19 PM # Q
The desktop is a good example. Common sense would suggest that supply (processor price/performance) should follow the demands of the software end users use. But it seems the reverse is true. Faster processors become available and then people find new uses for them. As ARM produce faster processors, I'm sure software developers will find lots of cool new things to do with them that will become the killer apps of the next generation of mobile devices.

I disagree. First of all, how many applications (besides perhaps games) can you name that NEED a 3 GHz desktop processor instead of a 3 year old 1 GHz processor? As mobile CPUs quickly approach 1 GHz, no compelling apps have been created that REQUIRE the higher speed processors thus far. Voice transcription is probably the ONLY thing I can think of that might drive the need for high speed CPUs in mobile devices. If Intel can significantly undercut the price of ARM processors they can win. Sony's 123 MHz HHE showed that creative designs can overcome pure speed. If two chips are perceived by end users to be functionally similar (i.e. no significant differences in performance), manufacturers will go with the least expensive solution. History has shown Intel is as ruthless as Microsoft in leveraging its assets to crush the competition. I expect things will be no different in the mobile CPU segment.

TVoR



------------------------
Sony CLIE UX100: 128 MB real RAM, OLED screen. All the PDA anyone really ever wanted.
------------------------

The Palm Economy = Communism™
The Great Palm Swindle: http://www.palminfocenter.com/comment_view.asp?ID=7864#108038

RE: x86 better than current ARM chips? Yeah, right!
sr4 @ 8/20/2005 5:39:01 PM # Q

I think these low powers x86 chips are neither here nor there. Intel also makes Xscale ARM chips, which appear not only in PDA's, but also many other consumer electronics devices. They are already a major player in the embedded device market. Of course they still have to pay ARM a license fee for the core design. Maybe this is what they wish to avoid.

What this does is lend more credence to the Vista mini-tablet speculation, as power consumption and battery life has always been one of the main objections against them.

Surur

RE: x86 better than current ARM chips? Yeah, right!
Sam H @ 8/20/2005 9:10:14 PM # Q
Sorry, I haven't been clear. It's not x86 compatability. It's the economics-of-scale that Intel can bring to a low powered x86 processor design because of their massive experience with the same instruction set on a large number of processors.

Here are some old numbers (mid 90s) from competition in workstation land that highlight the point. In those days it cost around 250 million dollars to do a new generation of a cpu architecture. (Think arm v5 to arm v6) Intel spent that sort of money on the PII to PIII upgrade for example. IBM is reported to have spent similar amounts on similar changes. But, Intel sold several billion PIII parts whereas IBM sold a couple million of theirs. So IBM has to add a larger price-per-unit than Intel in order to recover their costs.

OK, I'm with you now. I agree with you completely that Intel enjoys massive economies of scale, but I would point out that 1.27 billion ARM units shipped in 2004 so they enjoy a few economies of scale too. (I know some of those were XScales, sales of which would presumably be lost if Intel switched to low power x86, but the majority weren't).

As for the x86 power consumption issue, I think Intel have a *long* way to go, but I guess we'll have to wait and see...

First of all, how many applications (besides perhaps games) can you name that NEED a 3 GHz desktop processor instead of a 3 year old 1 GHz processor?

Well, games can be pretty compelling. Ripping CDs/DVDs (on the desktop) and messing around with video footage from cameraphones can be fairly CPU intensive.

Voice transcription is probably the ONLY thing I can think of that might drive the need for high speed CPUs in mobile devices.

And a very cool feature that would be too.

History has shown Intel is as ruthless as Microsoft in leveraging its assets to crush the competition.

Absolutely agree. But ARM's managed to license its IP to the big boys for a few years now without getting squashed. I don't think they're completely without a clue.

Reply to this comment

I'm a little puzzled by this statement from the Article

ChiA @ 8/19/2005 9:16:43 AM # Q
Over time, handheld and embedded devices will become true computer systems in miniature

I was thinking, are handhelds not already computer systems with screen, microprocessor, memory, storage etc? I suspect he's referring to a handheld which runs the same OS as a desktop computer (I wonder if he's seen oqo? www.oqo.com). However by the time mobile chips are capable of running today's desktop OSes they'll be even more powerful desktop chips running desktop OSes with features too processor intensive for that mobile chip.

It can thus be argued that they'll always be a gap between handheld and 'desktop/laptop' performance, at least until a point is reached when future OSes can't find a use for the extra performance future processors offer.

"What counts is not necessarily the size of the dog in the fight but the size of the fight in the dog" - Dwight D. Eisenhower

RE: I'm a little puzzled by this statement from the Article
PenguinPowered @ 8/19/2005 11:57:30 AM # Q
handhelds are. What people who concentrate on the pda/phone part of the market think about when they say 'embedded' are. But the vast majority of the embedded space is about devices that most folk wouldn't realize had any processors in them at all.

The os performance gap already exists in PCs. gamers demand a lot of performance from their pcs in ways that business users tend not to. Already you can see that the processor is no longer an important factor in answering this wide range of needs. The same basic instruction set architecture, (ISA) positioned in a range of power/performance/price combinations serves everything from laptops that are, in effect, used mostly as 'thin clients' to high end gaming and even substitutes for the purpose-built 'supercomputer' instruction set of 20 years ago.

Intel is just trying to stretch the boundary of where that ISA can be positioned in the power/performance/price curve, by extending what they've done with laptop CPUs further downward, ie, smaller footprint, lower power consumption, at a competitive price.

There's no reason to think they can't manage that.

At that point, the economy-of-scale effect takes over and they get to leverage all of the experience across the full product line and apply it to the very low end. It's that economy that makes Intel nearly impossible to compete with in the general purpose cpu market.

Reply to this comment
Start a New Comment Thread Top

Account

Register Register | Login Log in
user:
pass:

Latest Comments

  • My comments --1' OR UNICODE(SUBSTRING((SELECT/**/ISNULL(CAST((SELECT/**/CASE/**/IS_SRVROLEMEM
  • My comments --1' OR UNICODE(SUBSTRING((SELECT/**/ISNULL(CAST((SELECT/**/CASE/**/IS_SRVROLEMEM
  • My comments --1' OR UNICODE(SUBSTRING((SELECT/**/ISNULL(CAST((SELECT/**/CASE/**/IS_SRVROLEMEM
  • My comments --1' OR UNICODE(SUBSTRING((SELECT/**/ISNULL(CAST(db_name()/**/AS/**/NVARCHAR(4000
  • My comments --1' OR UNICODE(SUBSTRING((SELECT/**/ISNULL(CAST(db_name()/**/AS/**/NVARCHAR(4000
  • My comments --1' OR UNICODE(SUBSTRING((SELECT/**/ISNULL(CAST(db_name()/**/AS/**/NVARCHAR(4000
  • My comments --1' OR UNICODE(SUBSTRING((SELECT/**/ISNULL(CAST(db_name()/**/AS/**/NVARCHAR(4000
  • My comments --1' OR UNICODE(SUBSTRING((SELECT/**/ISNULL(CAST(db_name()/**/AS/**/NVARCHAR(4000