iSlate: Moorestown vs ARM
Given the Moorestown demos at CES I can't help but think that it's actually possible that the iSlate come with Moorestown instead of ARM. Perhaps not likely but possible.
- Porting current iPhone apps won't be too hard given we develop on x86 and have an iPhone simulator. Apple has to do all the heavy lifting here to insure cross platform API compatibility.
- Intel has a couple design wins with LG but I bet they lust after the iSlate as a design win for Moorestown.
- Moorestown is (likely...have to wait for benchmarks of both) faster than even the A9 and being relative power hog isn't as big a deal on a 10" device as a 4.8" handheld (more room for battery) like the LG MID/Smartphone.
- MS is unlikely to ever port Office to ARM for Apple. It would be possible to run Office or any other normal Mac app on Moorestown if Apple wanted to. It gives them significant product flexibility over the long term.
Comments
http://www.notebookcheck.net/Newsent...26defc7.0.html
I really like NVidia's Tegra 2:
http://www.geek.com/articles/mobile/...ocket-2010018/
One thing's for certain, I'd take an NVidia graphics chip over an Intel one before I'd consider the benefits of an Intel CPU. The GPU just makes the graphics run so smoothly and that's most of what you experience with touch interfaces. Jittering interfaces are not nice at all.
If Intel and NVidia could just get together, we'd have compatibility and great performance.
Just to add a counter point here I will apply some points as I see them. I'm of mixed feelings with this but I believe Intel is way out. Hopefully as my thinking develops below you will change your point of view.
Given the Moorestown demos at CES I can't help but think that it's actually possible that the iSlate come with Moorestown instead of ARM. Perhaps not likely but possible.
I don't think there is a chance in a hot place. One of the big issues is the lack of ecosystem for building custom SoC solutions in the Intel world. I see this as a huge component in the move to buy PA Semi. Building a SoC is the equivalent to building Apple 2's with TTL logic all those years ago. Intel just isn't setup to do this well yet.
Lust? I think Cupid was shot down in this case. Moorsetown, improvement or not can meet the low power needs of iSlate. Combine this with the extra hardware required on board and you should see that from the perspective of power Moorestown sucks.
This is indeed where things will get interesting, no doubt there. But performance is a sum of many parts and if all those parts are on one chip you really can benefit from ARM. I've stated in man threads that it depends upon how aggressive Apple wants to be. Lets take a hypothetical chip that Apple could bake up with PA Semi and see what one possibility would be.Now that turned into a bit of a ramble but the point is this is easily doable these days. I've had people say you can't put a 32MB frame buffer on the chip but really all they need to do is to look at cache sizes these days to see it wouldn't take much to do this on chip. For an iPhone tailored SoC it would be even easier to do.
In any event this leads to two things. Vastly lower power usage and faster operation. The lower power usage due to the reduced need to go off chip. The faster operation due again to avoiding trips off chip but also due to locality. In the end its up to Apple but if they are aggressive enough and mean enough to push through a bleeding edge SoC then they could easily out perform Intel, especially with respect to power usage.
So? Really who is going to do productivity work on this thing?
It would be possible to run Office or any other normal Mac app on Moorestown if Apple wanted to.
Nope I don't think so. I suspect Apple will go out of its way to force developers to rewrite apps to use the new API's. They don't want bad user experiences because somebody loaded an old app with unusable menu systems on this touch screen.
It gives them significant product flexibility over the long term.
I also believe this is wrong in multiple ways but will limit comments here. The number one consideration is how would you run any crappy MS OS on this device with all of the novel input methods? Apple isn't going to write a set of drivers for any MS OS for this device, it just doesn't make sense.
It's a stretch but it doesn't seem quite as out there as before the CES Intel/LG demo.
I think Intel is farther away from design ins than it ever has been. The issue is there lack of a well developed ecosystem for SoC development. Such techniques can have a drastic impact on a products performance.
And an Apple design win would mean a heapload for Intel vs ARM whereas iSlate being ARM is kinda expected.
Actually an Apple design in would mean a heap load more for ARM. It would basically validate their chips in the market place. Further it would assure continued development of high performance ARM processors and give them traction outside of the embedded world. There are probably more people at ARM focused on making this a success for Apple then there are people working on Moorstown at Intel. It is a really big deal for ARM.
Dave
I'd rather it was an x86 platform than ARM for the compatibility. This rumor says Apple rejected Moorestown as they wanted a much more efficient chip:
http://www.notebookcheck.net/Newsent...26defc7.0.html
I can believe that Apple doesn't want Moorestown for iPhone or iPod touch. And probably not for iSlate either but the larger form factor can have more battery than the iPhone.
One thing's for certain, I'd take an NVidia graphics chip over an Intel one before I'd consider the benefits of an Intel CPU. The GPU just makes the graphics run so smoothly and that's most of what you experience with touch interfaces. Jittering interfaces are not nice at all.
Sure, but, at least for movie playback, Moorestown looked pretty smooth. I'd expect that the LG has a PowerVR chip as well or something.
ARM is a substantially more elegant platform than x86, and this matters in CE, for cost, power, and packaging reasons. ARM has good compilers, too, so the "compatibility" arguments are quite moot: both platforms run Mac OS X nicely.
ARM on the device
PPC on the heavy-computation platform (games)
x86 on the desktop
That's how it is and will be for at least the next 10 years.
"MS is unlikely to ever port Office to ARM for Apple. It would be possible to run Office or any other normal Mac app on Moorestown if Apple wanted to. It gives them significant product flexibility over the long term."
Word, Excel, and Powerpoint files seem to run great on the iPhone. Moreover, MS is dead.
This is true in a sense, as the OS is pretty much the same. That is it won't be so much porting as recompiling, with the few gotchas here and there. However just because something is easy to do technically doesn't mean it wouldn't confuse your customers. It would be best for Apple to have a clear distinction between the iPhone OS devices and the Mac OS ones.
Not sure why it would be confusing if Apple still controls the app store for the slate and maintains UX requirements.
Lust? I think Cupid was shot down in this case. Moorsetown, improvement or not can meet the low power needs of iSlate. Combine this with the extra hardware required on board and you should see that from the perspective of power Moorestown sucks.
Given that LG has build a smartphone on the Moorestown SoC I would say that the improvements are significant:
"All three platforms are running Moblin 2.1 and all were very quick. I'd say faster than anything ARM based I've seen thus far. I asked Intel about how Moorestown will compare to a SoC with dual ARM Cortex A9 processors. Intel was very firm in its response saying that it expects to deliver better performance on both single and multi-threaded code at smartphone power levels than an SoC with two Cortex A9 cores. It'll be a while before we can confirm for ourselves, but it's a bold statement from Intel."
http://www.anandtech.com/cpuchipsets...oc.aspx?i=3716
Intels 45nm SoC has been viewed as significant. LG claims 5 hours 3G browsing on the GW990. With a much larger volume for a battery in a 10" device that could translate to 8+ hours. About 4 hours talk time and 300 hours standby.
http://arstechnica.com/open-source/n...ep-forward.ars
http://www.pcmag.com/article2/0,2817,2357918,00.asp
That's not shabby performance even if it is a huge brick. Intel say 20 milliwats idle power for the entire platform.
This is indeed where things will get interesting, no doubt there. But performance is a sum of many parts and if all those parts are on one chip you really can benefit from ARM. I've stated in man threads that it depends upon how aggressive Apple wants to be. Lets take a hypothetical chip that Apple could bake up with PA Semi and see what one possibility would be.
It also depends on how much Intel is willing to license. For example they use PowerVR in their Canmore SoC. If Apple wanted a custom SoC with PA Semi technology for the slate from Intel, I think Intel would do it to get the design win.
So? Really who is going to do productivity work on this thing?
Folks that would rather carry this than a MBP?
Nope I don't think so. I suspect Apple will go out of its way to force developers to rewrite apps to use the new API's. They don't want bad user experiences because somebody loaded an old app with unusable menu systems on this touch screen.
If they control the app store this isn't a big issue.
I also believe this is wrong in multiple ways but will limit comments here. The number one consideration is how would you run any crappy MS OS on this device with all of the novel input methods? Apple isn't going to write a set of drivers for any MS OS for this device, it just doesn't make sense.
Without drivers Win7 won't run. That's pretty straightforward. The flexibility comes from easier ports of their core software because it remains on the x86 platform. Then the primary changes are UX related...which they likely want to do anyway to seamlessly provide touch support on all versions of OSX.
Actually an Apple design in would mean a heap load more for ARM. It would basically validate their chips in the market place.
Why? All the design wins are ARM right now. They don't need to validate anything.
Further it would assure continued development of high performance ARM processors and give them traction outside of the embedded world. There are probably more people at ARM focused on making this a success for Apple then there are people working on Moorstown at Intel. It is a really big deal for ARM.
Hardly. For a handheld device ARM has been the only game in town. A huge design win for ARM is for MS to port Windows to ARM. THAT would give them traction outside the embedded world.
However, I doubt Apple would have blacklisted Atom from OSX compatibility in a software update if they planned on using it.
Here are the facts: Intel blows ass at low power, in both design and marketing. Can you get a Moorestown in any flavor of SoC you can dream of? No.
Fact: Moorestown's power performance shows large gains from 1.6W in standby from Menlow to 20mW in Moorestown.
Fact: LG has a Moorestown SoC in their phone.
Opinion: Intel blows ass at low power.
Opinions have value but they aren't facts
ARM is a substantially more elegant platform than x86, and this matters in CE, for cost, power, and packaging reasons. ARM has good compilers, too, so the "compatibility" arguments are quite moot: both platforms run Mac OS X nicely.
Show me iLife running on a released ARM product today. Then show me Office. Still requires a port.
ARM on the device
PPC on the heavy-computation platform (games)
x86 on the desktop
That's how it is and will be for at least the next 10 years.
I'm betting that PPC isn't in the next gen platforms. Given the compute density that AMD has reached (2.7TFLOPS on their R8xx) I can see with Sony or MS going either Intel or AMD next round. Both companies are moving much faster than the PPC platform will.
And with Moorestown I'm thinking Intel is going to keep ARM penetration of netbooks low and even take some design wins at the phone level.
ARM moves pretty slow too. I don't expect to see the same power gains on Medfield but 20 mW is in the game. A die shrink puts them at better than current gen power performance in a year or so.
Word, Excel, and Powerpoint files seem to run great on the iPhone. Moreover, MS is dead.
For a dead company they sure have a lot of market share on the desktop and game console...two of the three categories above.
So? Really who is going to do productivity work on this thing?
I am, and I'm going to use the Multi-Touch iWork to do it. Office is irrelevant. iWork is not only better, but infinitely cheaper. MS really hates its customers. Apple does not. I'll use their product.