[cybermat]

With the release of the iPad using Apple's own custom chip, do you guys think all Apple computers will go through another change in Architecture soon? I think it's quite possible since this is a good way for Apple to lock down its operating system.

[turtle777]

For what product ?

Macs ? No way.
iPhone / iPod ? Most likely

-t

[mduell]

In addition to the iPod/iPhone it seems like an obvious fit for TV, paired with the Broadcom chip for real HD.

PA Semi's experience is in low power chips, not high performance chips. I don't see them doing their own Mac chip.

[imitchellg5]

For tv and iPods/iPhones, for sure! But they won't switch from Intel for a very very long time, if ever.

[Veltliner]

It's much too expensive to develop a processor for desktop computers to be handled by Apple alone.

At least for now.

[EndlessMac]

I agree with the others. I doubt it will happen for their Macs. As Velt has mentioned the expense alone will make the price gap between PCs and Macs worst. The computer market is a very competitive one with many manufacturers. Also with computers people compare technical specs a lot more than some other electronics so if your computer is slower than others people will complain. Using the same processors as everyone else means it's easier for Apple to stay competitive.

It seems to be an ingrained habit because even the average person compares CPU, RAM, etc when buying a computer even though all they do is browse the web and check emails. My roughly 5 year old PowerBook can still do that. On stuff like a phone I could really care less about how fast the processor is as long as it does it's job to my satisfaction. It's more about the experience and IMO the iPad follows the same formula which means it allows Apple to use their own architecture and worry less about customers comparing processor specs. For example, when people compare the iPhone with the Blackberry I don't hear people talking too much about the processor speed but rather how it operates and does what it's suppose to do.

You are also forgetting about the benefit it is to some people who want to run Windows for those few applications they can't get on Mac. That issue alone has allowed several groups of people to make the jump to Macs. I don't know about everyone else but I really don't want to go through the pain of updating all of my applications again to run on yet another architecture.

[OreoCookie]

I disagree: in a few years, ARM cpus will be fast enough to handle simple computer tasks and I would really like to have a lighter notebook with a 10+ hour battery life. The iPad's power adapter has 10 W, that's it!

I think Apple could switch the MacBook to some ARM platform if they wanted to. Apple knows how to build this machine today, all the technology is in place. However, I do see obvious problems with support of some third-party apps (Adobe, I'm thinking of you), but then again, if you compartmentalize quite clearly and market this machine as a lighter portable for lighter computing, I don't see why not! At one point, computers will be `fast enough' for the average user who is just browsing the net, watching videos, writing mails and texts. To be honest, if it weren't for my hobby (photography) and my LaTeX addiction, I wouldn't need much more.

BTW, analysts predict that ARM cpus will outsell Intel cpus in the netbook/tablet/UMPC segment[/url]. There is definitely growing momentum.

[Andy8]

May help Apple control jailbreaking easier if they are controlling more of the chips inside?

[The Final Dakar]

I agree that it may find its way into an Apple TV revision.

[osiris]

I think Apple will split into their architecture into two divisions - consumer and professional.

The consumer line will have a closed system - like the iphone/touch/pad and likely future Apple TVs. This makes sense in terms of profit, control, support, etc. This is for those who want to view content.

The professional line will continue with the Intel chips, and cater to geeks/video/graphics/audio pros who create content (or consume it at an unbelievable rate).

Who knows... I sense a disturbance in the Force. But it could be indigestion!

[The Final Dakar]

So... everything but the MacPros?

[osiris]

Everything but the MacPro lineup - MBP, MacPro. and I imagine we'll see a new iMacPro at some point.

[OreoCookie]

I don't think so. The main separation is not consumer vs. professional, but if the device is made for content creation or content consumption (I'm borrowing this line from turtle, I can't find the post at the moment).

Plus, the ARM architecture is optimized for low power, not high performance. I'm thinking more of a MacBook Air with larger screens, less weight, much longer battery life that do the basics and offer you some way to create content.

[Big Mac]

It has been foretold: PPC shall return to Apple desktops at the End of Days.

[mduell]

Originally Posted by OreoCookie 

I disagree: in a few years, ARM cpus will be fast enough to handle simple computer tasks

Originally Posted by OreoCookie 

At one point, computers will be `fast enough' for the average user who is just browsing the net, watching videos, writing mails and texts.

Except that software will continue to evolve to do more and consume all available computing resources.

Originally Posted by OreoCookie 

I think Apple could switch the MacBook to some ARM platform if they wanted to.

Not even close. It'd be a 90% performance hit.

[turtle777]

Originally Posted by mduell 

Except that software will continue to evolve to do more and consume all available computing resources.

Which is exactly why we'll not see a full-blown OS X on the iPhone / iPad.

Apple needs to force programmers to optimize for the mobile platform.

Apple is very smart. There is no way any PC manufacturer will be able to get anywhere close in battery runtime, performance and size as long as they try to offer an Windows compatible tablet.
Even better: for any PC manufacturer to start a new platform (like Apple did with iPhone OS) is almost impossible.

Apple's closed ecosystem is going to make them rule the market like they do with iPods and iPhones already.

-t

[P]

They won't change the ISA Macs use because they quite simply won't be able to bring MS and Adobe with them one more time. What might happen is that they expand the current iPhone OS upwards so it takes market share from Mac OS X, but then it will be without major 3rd party apps.

ARM is nowhere near x86 performance today. While it is certainly possible to make an ARM processor with performance comparable to x86 - say 4 A9 cores at 2 GHz, the memory controller is already integrated - you'd lose most of the power advantage by doing so, and then why should you do it? x86 is a terrible ISA in some ways, but as soon as you have space for a massive decoder like in the P6, that disadvantage shrinks to the size of that decoder hardware - and x86 has a massive compiler advantage to counter.

That doesn't mean that Apple won't make thinner, lighter Macbooks at some point. They might - especially if Intel makes a cheap CULV Arrandale - but they won't go to ARM to do it.

[Big Mac]

Apple can't switch a particular Mac line like the MacBook to something other than x86 and still call it a MacBook without completely confusing the marketplace.

x86 is here to stay on Macs unless and until x86 is leapfrogged.

[OreoCookie]

Originally Posted by mduell 

Except that software will continue to evolve to do more and consume all available computing resources.

I don't think so: if you restrict yourself and if the UI is centered around single applications, you could be fine with it. The point is not to add feature after feature, but to keep apps focussed. For the applications I have in mind (LaTeX, git, etc.), I could easily live with a more focussed UI on my mobile machine.

Originally Posted by mduell 

Not even close. It'd be a 90% performance hit.

I wasn't talking about making the architecture switch today, but in a few years. Of course, Intel's cpus will remain faster than ARMs, but my point was that ARM's offerings could, at that point, be simply fast enough for our mainstream computing needs. The rest is then a question of compromise: do you focus on portability and battery life or on performance. And in my opinion, once the performance is sufficiently high, I'd be willing to make that compromise for my mobile machine. I'd, of course, have a second machine that is optimized on performance.

Right now, I own a MacBook Pro, because I absolutely need (1) a portable machine and (2) all things in one place. If I could have it my way, I'd have an ARM-based machine to carry around which syncs all my documents automatically with my 27" iMac.

[OreoCookie]

@P
The iPad's power adapter is rated at 10W! That's a lot lower than anything you could do based on Intel (non-Atom) chips.

To my knowledge, the performance of the Cortex A9 is comparable to an Atom. Since all I've seen are some web rendering `benchmarks'. Even this comparison is problematic since it wasn't a real benchmark: the Cortex A9 had two cores, was clocked at only 500 MHz and there was no functioning graphics accelerator. But I think it's good enough to claim that a dual core Cortex A9 @ 1 GHz coupled to a graphics accelerator has an integer performance of the same order of magnitude than a single-core Atom. If I extrapolate, the Atom is about 4 times slower than a Core 2 Duo.

Does anyone know of any SPECmarks for the Cortex A9? I'd be really interested just to see how much slower it is.

[P]

Technically you don't get SPEC for a CPU, you get it for a complete system, but I see your point. And no, haven't seen any. According to ARM, the quadcore Cortex A9 delivers 2.0 DMIPS/MHz, or 0.5 DMIPS per core. IF the A4 uses a A9 CPU with 4 cores - and we don't know that it does, just rumors at this point - then it would deliver 2000 DMIPS.

DMIPS are the old Dhrystone benchmark divided by 1757, the score of a 1 MIPS machine. DMIPS scores are generally considered obsolete today, so it was hard to find any app that would run the test. I found one that was compiled for 32-bit x86 and ran it on my iMac i7 - I don't have the dev tools installed here yet, and I dind't want to bother with that for a silly forum post. I started four instances at the same time to use all four cores, and each reported a score of just over 10000 DMIPS. With this unoptimized code, my iMac has roighly 41000 DMIPS. I then started 8 instances instead, to see if Hyperthreading could help. It could - I got a total DMIPS of 52356. I had Safari and Activity monitor open, but not too much else. Dhrystone is a pure integer benchmark, no floating point operations involved. I don't know if SSE was used.

I'd like to turn your question on its head, however: If we decided that a 2.8 GHz Lynnfield is all the CPU we will ever need, why should it be better to achieve that level of performance using the ARM ISA? The x86 "tax" that you have to pay is so tiny when you get to a chip the size and power of Lynnfield, that there is very little to gain there.

[Cold Warrior]

Would it be possible or practical for their notebooks to have both their own and an Intel CPU and activating the Intel CPU only as needed?

[The Godfather]

Originally Posted by Big Mac 

It has been foretold: PPC shall return to Apple desktops at the End of Days.

While PPC takes ARM to school any day of the week...
Mindstab.net � Blog Archive � Primes results for x86 vs. PPC vs. Arm
PPC manufacturers have been avoiding the handheld market for unknown reasons, thus becoming undesirable for phones, PDAs, etc. Laptops and desktops will demand x86 processors because they need to run existing proprietary software, unless an emulation solution comes along.

However, PPC chipmakers have their blinders aimed at the networking, telecom, avionics and automotive industries. ALL OF THEM. None of them are integrating PowerVR, or even LCD controllers, or even minimizing their footprint.

ARM Cortex is where it is at.

[The Godfather]

Originally Posted by mduell 

In addition to the iPod/iPhone it seems like an obvious fit for TV, paired with the Broadcom chip for real HD.

PA Semi's experience is in low power chips, not high performance chips. I don't see them doing their own Mac chip.

But PA Semi's own marketing pamphlet says "high efficiency, with high performace"! <tongue-in-cheek>

[The Godfather]

Originally Posted by Cold Warrior 

Would it be possible or practical for their notebooks to have both their own and an Intel CPU and activating the Intel CPU only as needed?

Hackers need to get these babies on their laps
Lenovo IdeaPad U1 Hybrid hands-on and impressions -- Engadget
http://www.engadget.com/2008/08/15/d...screenshooted/

[mduell]

Originally Posted by Cold Warrior 

Would it be possible or practical for their notebooks to have both their own and an Intel CPU and activating the Intel CPU only as needed?

Sure... this leads to some very interesting possibilities.

edit: You win, The Godfather.

[The Godfather]

Originally Posted by Andy8 

May help Apple control jailbreaking easier if they are controlling more of the chips inside?

The first few iPod generations were hacked to run iPodLinux, but Apple, as they got more experience, eventually started producing un-hackable iPods. This might become the case for the iPad (though no one will care, except for warez fans, because it is not locked to a wireless carrier).

[P]

Originally Posted by P 

Technically you don't get SPEC for a CPU, you get it for a complete system, but I see your point. And no, haven't seen any. According to ARM, the quadcore Cortex A9 delivers 2.0 DMIPS/MHz, or 0.5 DMIPS per core. IF the A4 uses a A9 CPU with 4 cores - and we don't know that it does, just rumors at this point - then it would deliver 2000 DMIPS.

I need to correct myself here. First up, the DMIPS number is being reported as either 2.5 DMIPS/MHz, or "over 2" DMIPS/MHz. Secondly, that is per core, so a 1 GHz quad model would be somewhere between 8000 and 10000 DMIPS. The quadcore model seems to exist only in theory, however, so the most likely configuration for the A4 is a single core@1 GHz, and still 2000-2500 DMIPS.

[OreoCookie]

Originally Posted by P 

Technically you don't get SPEC for a CPU, you get it for a complete system, but I see your point. And no, haven't seen any. According to ARM, the quadcore Cortex A9 delivers 2.0 DMIPS/MHz, or 0.5 DMIPS per core. IF the A4 uses a A9 CPU with 4 cores - and we don't know that it does, just rumors at this point - then it would deliver 2000 DMIPS.

While I think your overall argument is spot on, ARM cpus have much less cpu horsepower than current desktop cpus, I think your numbers are not correct: the 2.0 DMIPS/MHz are to my knowledge per core and not for four cores. Also, according to this anandtech article, the Cortex A9 is rated at 2.5 DMIPS/MHz while the A8 achieves 2.0 DMIPS/MHz.

Originally Posted by Anandtech

Cortex A9 has a shallower pipeline compared to A8, so it does more per clock. It also has an out of order execution engine, allowing it to also do more per clock. At the same clock speed, A9 should destroy A8. ARM estimates that the A8 can do up to 2 DMIPS per MHz (or 2000 DMIPS at 1GHz), whereas the A9 can do 2.5 DMIPS per MHz (2500 DMIPS at 1GHz). Given that most A8 implementations have been at or below 600MHz (1200 DMIPS), and TI's A9s are running at 750MHz or 1GHz (1875 DMIPS or 2500 DMIPS) I'd expect anywhere from a 30 - 100% performance improvement over existing Cortex A8 designs.

That's just for a single core though.

Hence, a quad core Cortex A9 running at ~1 GHz achieves roughly 8,000-10,000 DMIPS. While the difference is still roughly a factor 5, i. e. about 5 years of cpu development, it is in the same range as Intel's Atom -- which is commonly used in all sorts of appliances and netbooks and accepted as `suitable for light use.'

Originally Posted by P 

I'd like to turn your question on its head, however: If we decided that a 2.8 GHz Lynnfield is all the CPU we will ever need, why should it be better to achieve that level of performance using the ARM ISA? The x86 "tax" that you have to pay is so tiny when you get to a chip the size and power of Lynnfield, that there is very little to gain there.

To me the argument is that the theoretical maximum performance is of the same order of magnitude as an Intel Atom -- a cpu which is accepted as a cpu that is `fast enough for many.' Also, I personally don't care about architecture, but I do care about battery life. Since we Mac users are in the fortunate position that we are not tied to one particular architecture (can you imagine the market share of the Atom if Windows 7 ran on ARM cpus?), we could use a chip that is a lot more power efficient than an x86 cpu that is about as fast but much less frugal.

Disclaimer: I'm aware that comparing performance is very difficult in this segment as ARM cpus have co-processors for many tasks for which the cpu is commonly used on standard pcs. Plus, when you consider typical usages in the embedded segment, additional cores may not be very useful.

[OreoCookie]

Originally Posted by P 

I need to correct myself here. First up, the DMIPS number is being reported as either 2.5 DMIPS/MHz, or "over 2" DMIPS/MHz. Secondly, that is per core, so a 1 GHz quad model would be somewhere between 8000 and 10000 DMIPS. The quadcore model seems to exist only in theory, however, so the most likely configuration for the A4 is a single core@1 GHz, and still 2000-2500 DMIPS.

Agreed. I think the dual core configuration will be equally common since you can buy dual core Cortex A8 now and they're used in higher-end devices. However, due to the specific nature of the workload of embedded systems (less multitasking than on desktops), I think in most cases, you cannot really utilize more than 2 cores.

In any case, my point is that the performance of ARM cpus is encroaching the Atom territory.
Edit: wikipedia tells me that an Atom N270 churns out about 3,300 MIPS. That's a little more than one Cortex A9 core and certainly less than 2. The new Pinetrail Atoms will manage about 4000 DMIPS.

[P]

Originally Posted by OreoCookie 

Hence, a quad core Cortex A9 running at ~1 GHz achieves roughly 8,000-10,000 DMIPS. While the difference is still roughly a factor 5, i. e. about 5 years of cpu development, it is in the same range as Intel's Atom -- which is commonly used in all sorts of appliances and netbooks and accepted as `suitable for light use.'

I was rather giving some background to roughly how fast it is than making argument, but OK. Yes, it is comparable to an Atom - potentially, anyway - but Jobs has derided netbooks publicly, and judging from the reaction of the man on the street (ie, people who ask me about their new netbooks), Atom isn't really fast enough. A lot of that is Flash, but still - not quite good enough. Pine Trail is a disappointment performance-wise, which has put Atom on the back foot for a while.

A factor 5 is roughly 5 years of development if you include process shrinks - but that's not fair here. Lynnfield is a 45nm chip, which is the same as or slight worse than A4 (A4 is either 45nm or 40nm, depending on the source). Intel achieves about 20% from one generation to the next without shrinks. (1.2)^9=5.1, so it's roughly 9 generations behind, if you put it that way. That is of course also not fair, as A9 is a tiny chip compared to Lynnfield, but...

Anyway. An A9 successor in 5 years with the same limitations on size and power ain't gonna be no Lynnfield.

Originally Posted by OreoCookie 

To me the argument is that the theoretical maximum performance is of the same order of magnitude as an Intel Atom -- a cpu which is accepted as a cpu that is `fast enough for many.' Also, I personally don't care about architecture, but I do care about battery life. Since we Mac users are in the fortunate position that we are not tied to one particular architecture (can you imagine the market share of the Atom if Windows 7 ran on ARM cpus?), we could use a chip that is a lot more power efficient than an x86 cpu that is about as fast but much less frugal.

Disclaimer: I'm aware that comparing performance is very difficult in this segment as ARM cpus have co-processors for many tasks for which the cpu is commonly used on standard pcs. Plus, when you consider typical usages in the embedded segment, additional cores may not be very useful.

So what you're saying is that if Apple should decided to make a Mac netbook, they would do well to put an A4 in it rather than an Atom? Sure, I agree with that. There are problems in that many big apps wouldn't run on it - no Adobe CS, no MS Office - but Apple can make good substitutes. It's just that

a) I don't think Apple will ever make any netbooks - iPad is their entry into that market.
b) There are other CPUs beside Atom and ARM - Celeron CULV, for instance - that would be an even better fit.
c) Intel can easily boost Atom to higher performance, if they'd like to. The current Pine Trail setup involves a memory controller on the chip, but with a high latency FSB connection to the cache subsystem simply because Intel couldn't be bothered to develop a new memory controller. There are many things like that they could do.

[The Godfather]

Originally Posted by OreoCookie 

Agreed. I think the dual core configuration will be equally common since you can buy dual core Cortex A8 now and they're used in higher-end devices. However, due to the specific nature of the workload of embedded systems (less multitasking than on desktops), I think in most cases, you cannot really utilize more than 2 cores.

In any case, my point is that the performance of ARM cpus is encroaching the Atom territory.
Edit: wikipedia tells me that an Atom N270 churns out about 3,300 MIPS. That's a little more than one Cortex A9 core and certainly less than 2. The new Pinetrail Atoms will manage about 4000 DMIPS.

Except that today's Atoms are handling 64 bit instructions, and a bloated OS, to say the least...

The day both CPUs are encumbered with the same tasks, we'll see who strongarms the other.

Has anyone loaded Android on their netbooks yet?

[P]

DMIPS shouldn't be affected by which OS is used, and 64-bit support is hardly a performance limitation. Besides, Atom N270 (Diamondville, what's usually in netbooks today) doesn't support 64-bit instructions, although N450 (Pinetrail, the new version ) does.

There are Android netbooks around, but I haven't seen any real tests on them.

[Spheric Harlot]

Originally Posted by Big Mac 

It has been foretold: PPC shall return to Apple desktops at the End of Days.

Does somebody still make 680x0 chips?

I'm having none of that PowerPC devil's offspring in my machines!

[OreoCookie]

Very nice post, P.

Originally Posted by P 

I was rather giving some background to roughly how fast it is than making argument, but OK. Yes, it is comparable to an Atom - potentially, anyway - but Jobs has derided netbooks publicly, and judging from the reaction of the man on the street (ie, people who ask me about their new netbooks), Atom isn't really fast enough.

You're absolutely right. Hence, I don't think Apple will sell an ARM-based `netbook' now, we absolutely agree on that. What I can fathom is that more and more desktop-typ apps are ported to the iPhone OS (mainly a UI problem) -- and then you can run apps on your iPad that you'd traditionally use a normal computer for. So you have more and more functionality that you can get only on desktops at the moment and move it to something like the iPad.

I can really see myself using a lowly text editor on that thing to type a lot of texts. That'd be plenty for my type of work when I'm on the road.

Originally Posted by P 

Pine Trail is a disappointment performance-wise, which has put Atom on the back foot for a while.

Agreed, I couldn't believe when I read the first benchmarks. I thought to myself: that's it? That's how they want to compete?

Originally Posted by P 

Anyway. An A9 successor in 5 years with the same limitations on size and power ain't gonna be no Lynnfield.

No, certainly not. But this is where coprocessor and the GPU come into play that are utilized more and more to accelerate specific tasks. Of course, the approach is very different here: general purpose computing vs. a device for some specific tasks.

Originally Posted by P 

So what you're saying is that if Apple should decided to make a Mac netbook, they would do well to put an A4 in it rather than an Atom? Sure, I agree with that. There are problems in that many big apps wouldn't run on it - no Adobe CS, no MS Office - but Apple can make good substitutes.

Yup.
I think those substitutes will run on the iPhone OS and Apple will simply not necessarily use the traditional version of OS X. However, I think it's good to keep in mind that in terms of RAW processing power, new ARM offerings are on par with netbooks.

Originally Posted by P 

a) I don't think Apple will ever make any netbooks - iPad is their entry into that market.

Agreed.

Originally Posted by The Godfather 

Except that today's Atoms are handling 64 bit instructions, and a bloated OS, to say the least...

Atom being 64 bit says nothing, Atom's are being limited to 1 or 2 GB RAM and the most important reason to switch to 64 bit, namely being able to address more than 4 GB RAM, is moot.

The second advantage of being 64 bit applies only to x86 cpus: in 64 bit mode, there are additional registers available that speed up some tasks. But for other architectures, e. g. PowerPC, there is no such boost.

64 bit is not really a hot topic in the field of embedded devices and netbooks.

[P]

Originally Posted by Spheric Harlot 

Does somebody still make 680x0 chips?

I'm having none of that PowerPC devil's offspring in my machines!

I think TI is using them in their calculators (TI-89, TI-92 etc). Some of those are still sold, so presumably they are.

[P]

Originally Posted by OreoCookie 

Very nice post, P.

You're absolutely right. Hence, I don't think Apple will sell an ARM-based `netbook' now, we absolutely agree on that. What I can fathom is that more and more desktop-typ apps are ported to the iPhone OS (mainly a UI problem) -- and then you can run apps on your iPad that you'd traditionally use a normal computer for. So you have more and more functionality that you can get only on desktops at the moment and move it to something like the iPad.

I can really see myself using a lowly text editor on that thing to type a lot of texts. That'd be plenty for my type of work when I'm on the road.

I think more desktop apps are being ported - Omni is porting everything they make. I could see myself using an iPad as well, but then the AppStore has to go - or rather, I must be permitted to load whatever apps I want onto the device. If they want to restrict access to the cell phone radio and stuff like that, then OK, but the current setup is to restrictive for me for that "second computer" niche.

It really is a very nice idea otherwise - an iMac at home with all the stuff I need, and a small portable pad for what I might want to use on the road.

Originally Posted by OreoCookie 

Agreed, I couldn't believe when I read the first benchmarks. I thought to myself: that's it? That's how they want to compete?

They're worried about cannibalization, I think. The new Pine Trail platform must be significantly cheaper to produce, yet the prices haven't come down - if anything, they go up.

Originally Posted by OreoCookie 

Atom being 64 bit says nothing, Atom's are being limited to 1 or 2 GB RAM and the most important reason to switch to 64 bit, namely being able to address more than 4 GB RAM, is moot.

The second advantage of being 64 bit applies only to x86 cpus: in 64 bit mode, there are additional registers available that speed up some tasks. But for other architectures, e. g. PowerPC, there is no such boost.

64 bit is not really a hot topic in the field of embedded devices and netbooks.

64 bit should help Atom more than other Intel CPUs though, now that I think about it. The reason for the minor speedup on x86 in general is that there are twice as many programmatically accessible (named) registers - 16+16, instead of 8+8. When Intel developed P6 (Pentium Pro, ancestor to Core) they decided to compensate for this by adding more registers than required by the ISA and then "move the names around". This means that 64 bit mode doesn't help that much on modern CPUs, but Atom is a P5 (original Pentium) at heart. P5 never had register renaming before, and I hadn't heard that Intel added it to Atom. If going 64-bit means a doubling of the number of registers from 8 to 16, that could mean a big speedup in some cases. Hm. Maybe a Pinetrail netbook with Win 7 x64 would perform better than that meager 5% speedup?

[OreoCookie]

Originally Posted by P 

I think more desktop apps are being ported - Omni is porting everything they make. I could see myself using an iPad as well, but then the AppStore has to go - or rather, I must be permitted to load whatever apps I want onto the device. If they want to restrict access to the cell phone radio and stuff like that, then OK, but the current setup is to restrictive for me for that "second computer" niche.

I'm thinking the same thing. I'm very tempted to get an iPad, a Bluetooth keyboard and use it for LaTeX. However, the problem is that LaTeX is a unix app and as such you need to have access to the innards of the system. Perhaps someone finds a workaround, I'm not sure.

I've had laptops since 1998. As a matter of fact, my first laptop was my first Mac. I'm contemplating on getting a nice iMac and an iPad instead of my Macbook Pro. Then I'd use Dropbox to sync my work projects and some other files across devices. Other than a browser, a text editor, an e-mail client, iTunes, Aperture and OmniFocus, I don't use that many apps. (I know I wouldn't want to run Aperture on the iPad anyway )

I'm glad I'm not the only crazy person here

Originally Posted by P 

They're worried about cannibalization, I think. The new Pine Trail platform must be significantly cheaper to produce, yet the prices haven't come down - if anything, they go up.

Yes, but it's stupid: it's a growing market and the `new' platform doesn't offer anything exciting. Even worse, nVidia is moving out of the chipset market I've heard so that you have to couple the Atom to some mediocre integrated graphics by Intel.

I wonder how they plan on cracking the embedded market with such a strategy.

Originally Posted by P 

64 bit should help Atom more than other Intel CPUs though, now that I think about it. ... This means that 64 bit mode doesn't help that much on modern CPUs, but Atom is a P5 (original Pentium) at heart. P5 never had register renaming before, and I hadn't heard that Intel added it to Atom. If going 64-bit means a doubling of the number of registers from 8 to 16, that could mean a big speedup in some cases. Hm. Maybe a Pinetrail netbook with Win 7 x64 would perform better than that meager 5% speedup?

Good point, I didn't think about that. But even if there is a speedup of ~15-20 %, it is not earth shatteringly faster.

[olePigeon]

I suppose Apple could license x86 from Intel and make A5 and A6 chips for portables and desktops respectively.

[OreoCookie]

Intel has (to my knowledge) never licensed its x86 cpus to anyone. And even if they did, what is Apple supposed to do with it? For general purpose computing, there is less need to make custom chips while for embedded systems it's desirable to choose the chip that suits your need, i. e. a combination of power requirements, performance, co-processors and price.

[P]

Correct, Intel doesn't license x86. The only entity with a x86 license is AMD, and AMD was licensed because IBM wouldn't buy 8086 CPUs for the original PC if they didn't have a second potential source for it. That doesn't prevent anyone from reverse engineering it, but if you do that, Intel won't like you. Apple cannot afford to do that - in fact, it's probably cheaper to just buy AMD if they wanted to make their own CPU/SOC for desktops.

The idea of making your own SOC based on Intel's CPU designs is intriguing, but it ain't happening now. Intel outsourced production of Atom to TSMC precisely so customers could combine it with their own hardware on one chip - the implication being that they wouldn't be permitted inside Intel's plants. Current desktop CPUs are also borderline too large already, without a lot of SOC circuitry, but sometime around Ivy Bridge there should be space for some more stuff. Not sure what the gain is in doing that, though - sure if you could ditch Intel's graphics, but I don't think they'll let you do that.

I

'm thinking the same thing. I'm very tempted to get an iPad, a Bluetooth keyboard and use it for LaTeX. However, the problem is that LaTeX is a unix app and as such you need to have access to the innards of the system. Perhaps someone finds a workaround, I'm not sure.

I've had laptops since 1998. As a matter of fact, my first laptop was my first Mac. I'm contemplating on getting a nice iMac and an iPad instead of my Macbook Pro. Then I'd use Dropbox to sync my work projects and some other files across devices. Other than a browser, a text editor, an e-mail client, iTunes, Aperture and OmniFocus, I don't use that many apps. (I know I wouldn't want to run Aperture on the iPad anyway )

I'm glad I'm not the only crazy person here

We're not crazy, we're part of the next wave...

I need a browser and I need to be able to at least read Office documents. Editing them would be nice, but the important thing is that I have to be able to read them. I have a work laptop that I lug around everywhere, but it's locked down pretty tight and the incessant virus scanning makes it slow. Playing movies on it would be nice, and I don't own a GPS which might be useful once in a while.

[The Godfather]

LaTEX depends on a zillion libraries, but not on access to the innards of the host it is running in.

Besides, being a unix app doesn't mean you need access to the innards of the system. Any hardware drivers are needed to run a LaTEX?

[OreoCookie]

No, hardware drivers are not necessary.
Since there have been LaTeX distributions for lowly OS 8, I am confident that one could make an `app package of it.' The problem is file handling: latex acts as a compiler that is fed html-like code. Code which you have to generate with some form of text editor.

AFAIK there is no way to exchange files across applications in the iPhone OS.

[P]

There is on the iPad, though - there is an area that you can access from the Mac as a USB disk, and as I understand it all apps can access that area.

The common process for doing it on a UNIX system isn't necessarily the best, however: A text editor, ideally with some sort of syntax coloring, and then a button that compiles it and shows the result in a PDF-viewer - assuming you compile into PDF these days. Back when I briefly used LaTeX, you compiled into .ps files, but hopefully that has changed by now.

Alternatively I guess you could use .dvi and write a DVI-viewer. It should probably be one app, though, to work around the lack of multitasking.

[Simon]

For years now there has been a thing called PDFLaTeX. Works like a charm. No more need to go the DVI-PS-PDF route anymore. Used in combination with a decent editor/viewer app like TeXShop it's pure awesomeness.

That said, I have zero interest to use it on something like an iPad. IMHO two things are required for a good LaTeX experience: a good KB and a huge screen. The iPad has neither. Sure I can get a BT KB for the iPad, but then I might as well just use my MBP for LaTeX. And that's what I'm already doing.

[OreoCookie]

Wait until you try xelatex … that's like pdflatex that allows you to use all of your fonts And then, to top things off, you switch to TextMate + Skim. TextMate is just amazing and third-most used app (right after iTunes and Safari).

The thing is, I'd like something more portable than my MacBook Pro (I'd be ok with a MacBook Air, I guess, but I'd like longer battery life), but on the other hand, I don't want to split my data in two.

I can see all sorts of problems associated to working with my LaTeX Projects on an iPad: I very much rely on quite a few tools such as git and tex that are most easily accessible from the command line. I need a sort of file management: I have a very clean structure of my latex projects (separation of content into quite a few files) that needs to be handled properly. I doubt Apple will give the user access to a Terminal