PDA

View Full Version : When, in the distant future will OS X run solely on ARM?




Mhkobe
May 21, 2010, 12:54 AM
What do you guys think? Intel are jerks, and AMD sucks, ARM can be completely controlled by Apple (for now).



Scottsdale
May 21, 2010, 01:25 PM
I definitely believe Apple intends to switch Mac's OS to ARM at some point maybe three to five years down the road.

I believe Apple could provide an SoC that really gives its Mac users a system perfect for running a Mac OS. However, I doubt it will be called OS X. I would have to believe this will happen after OS X and become some OS that's completely different. Meaning, I don't think Apple will take the OS X code and convert it to run on ARM.

What ARM provides for Apple is removing the Mac buyer from the Intel system of saying buy the new Core i7 CPU. This way Apple can sell you a Mac with an Apple A7 CPU. We won't know if it runs at 3 GHz with 1 Core or 2 GHz with 64 cores or whatever. We will have the option of 13" 15" 17" Mac notebooks that vary based on their display size, ports, drive size, and capabilities NOT based on RAM, CPU, GPU.

Think how Apple currently positions its iPad. Apple doesn't tell us the CPU is a 1 GHz SoC chip with graphics built in. It just says Apple A4 silicon. We don't see or get options for RAM. We just know from iFixit that it has 256 MB of RAM. These future Macs will be sold with the same strategy. All will probably be the exact same SoC with same capabilities. We will select based on what size display, drive, and port capabilities we need. If we think of it, this is the Apple way. It will get out of the game of competing with PC laptops completely.

When the Macs used to have PPC chips there was less comparison between Macs and PC laptops. Now that we can compare, we all see just how inferior the Macs are with low grade components. Apple has to love the idea of getting out of this system. The ARM SoC way is the Apple way. Apple has to have the strategy to move to this as soon as possible. In addition, the way we compute will probably change and be more like the iPad anyways. It will probably converge between an iPad and a Mac. One OS will surely dominate too. It will probably be somewhat iPhone OS 4 and partially OS X.

jimboutilier
May 21, 2010, 02:48 PM
Those that don't learn from history are doomed to repeat it.

Does the Power PC ring any bells? Apple once adopted an excellent technology that in theory outperformed its competitors at the time. But small market share means small R&D and the technology with larger market share and more R&D will almost always come out on top in the long run.

Thats why the Power PC is dead and Intel thrives. And you think its a good idea for Apple to do that again?

Scottsdale
May 21, 2010, 04:28 PM
ARM is completely different from PPC. A Cortex 9 is an amazing chip with tons of capabilities. What I see happening in the future is not just two, four, or eight cores in a CPU. I see Apple software using 32 cores (just example could be amazingly more or half as many cores) in an ARM SoC design at a couple of GHz performing incredible vs. anything we see today in Intel Core i-series CPUs or even the future of those CPUs.

The whole thing is today's computers have CPUs that aren't being fully utilized. It's easy to think of the human brain as a current Core i7 CPU. We use about 10% of our brain's capabilities (so the myth goes). It is the exact same thing with the current CPUs in Macs. What customers would benefit much more from is Apple taking advantage of and using what's already available with the CPUs. For example, the SSD technology is just becoming readily available and affordable, but this technology should have been pushed a long time ago instead of focusing on clock speed. The bottleneck in any computer sold today is the drive and drive controller. While the drive controller was less of a factor before the SSDs, it's quickly becoming more of the problem when SSDs can saturate the bandwidth capabilities of the drive controller. Let's look beyond to drives that operate 5x as fast as today's SSDs and our computing experience would be radically different using the exact same Core 2 Duo CPUs.

Next let's look at the GPUs used in our Macs. The GPUs are far more capable than Apple is giving them credit for. We would benefit from applications that use OpenCL technology. But even that is a small part of the focus. Right now Apple is not fully utilizing the GPU in Macs. The 9400m has been an incredible GPU for the MBA, and it's getting better over time as Apple figures out how to write better drivers. The recent 10.6.3 update brought over 100% performance gains for OpenGL on the 9400m. Apple finally gave access to h.264 API acceleration for the 9400m for third-party developers. This means Flash and HD applications will start actually utilizing the GPU for acceleration instead of counting on the CPU. We still see much more amazing performance from our Macs with 9400m GPUs in Windows because of the drivers and system capabilities in Windows. Let's face it, Microsoft provides a better graphics system than OS X. These gains could benefit the users by continuing to provide the best drivers that allow the GPUs to do what they're made to do. Nvidia can write a better driver for its GPU than Apple can, so why doesn't Apple let Nvidia help with the drivers? Apple has proven that it's not as good at writing drivers for graphics, but at least it's getting better. We could have much better Macs if Apple spent less money on CPU upgrades and more money on integrating better GPUs. In addition, it could drop the optical drives from MB/MBPs and give the systems more space for a dedicated GPU and cooling the GPUs. We also need GPUs that have enough VRAM available. We need these discrete GPUs to have dedicated VRAM and other computer manufacturers understand this but MBPs have had 256 MB of VRAM for three years! The low-end MBPs, MB, MBA, Mm, and iMacs share 256 MB of their system RAM with the GPUs.

The next thing is RAM. We need faster RAM, and we need more RAM. This is a no brainer.

Finally, software. If Apple spent more money making better software that took advantage of the computers hardware we would have much faster Macs. We don't need a new CPU people. What we need is Apple to work on making better software. We are all conditioned to believe we need a new 3 GHz Core i7 to have a faster computer. This is the PROBLEM with the Intel way of business. Intel conditions us to believe we need a faster CPU and ONLY a faster CPU can speed up our computers. This is ridiculous. The CPUs are absolutely the best and most advanced component used in computers today. If there's one component that doesn't need upgrading IT IS THE CPU!

Why do you think Apple is buying up ARM technology? Just for the iPad and iPhone? Apple is quickly becoming a mobility focused company. The ARM SoC design Apple used in its A4 is barely touching the capabilities that are awaiting the potential in the not too distant future. These current SoC designs like the Apple A4 and the new Nvidia Tegra 2 used in the new tablet which uses a ARM Cortex 9 SoC design, are already powerful enough to give the common or average computer user the experience they would expect. It will require rewriting software, but Apple is already doing it with the iPhone OS. Within three to five years, I definitely HOPE we are using ARM SoC designs in Macs. I guarantee you that Apple can provide us with the absolute best computers ever that blow away the current systems requiring us to pair an Intel chipset with an Intel Core i-series CPU... then forcing us to use Intel's crap GMA or adding the cost of a GPU that isn't integrated into the system.

ARM is definitely the future. It is a matter of WHEN not if.

Mhkobe
May 21, 2010, 05:10 PM
ARM is definitely the future. It is a matter of WHEN not if.

I completely agree, SoCs are the best way to improve speed in CPUs (although most people don't really need faster CPUs), and ARM is Apples way to prove that. I believe apple understands that more cores is way better than high clocks, but currently there is still a fight for faster clocks. Apple needs to improve this thought though, by either introducing an easy way to distribute the operations across massive (32+) numbers of threads (GCD is an excellent start), or by creating their own SoC with automatic switching. I would prefer the first option because the latter would have no idea which processes are large until they are already divided.

Also, Apple do say on their website that the A4 in the iPad is 1Ghz (http://www.apple.com/ipad/specs/).

As for when, my opinion is that we won't see one until 4-6 years have gone by. I agree that Apple would probably be past OS X by then, either on to OS 11 or 12 based on whether or not apple is keeping the OS 11.xx strategy over OS "x" or not.

No ice please
May 21, 2010, 06:58 PM
I could see apple using some ARM processors in maybe the Apple Tv and Mac mini to compete with Nettops and HTPCs.

Winni
May 21, 2010, 08:19 PM
ARM is completely different from PPC. A Cortex 9 is an amazing chip with tons of capabilities. What I see happening in the future is not just two, four, or eight cores in a CPU. I see Apple software using 32 cores (just example could be amazingly more or half as many cores) in an ARM SoC design at a couple of GHz performing incredible vs. anything we see today in Intel Core i-series CPUs or even the future of those CPUs.


And even if you had one billion CPUs in your computer, that doesn't change a single bit of the fact that most of your daily tasks simply cannot be parallelized to take advantage of so many cores. Except for the typical number crunching tasks (which include things like DVD ripping or Photoshop filters), nothing will ever benefit from having 32 or more cores available. And an average human being also doesn't multi-task well enough to use dozens of applications simultaneously and there certainly is no need or economic justification to have all installed apps running at all times.

Especially games, which are usually the most demanding type of application, are traditionally single-threaded apps (which is the reason why Xbox 360 versions of games always kill PS3 ports of the same game: The PS3 has more cores, but the games CANNOT use them; the Xbox has fewer but FASTER cores). So even if the developers move physics calculations to one or two separate cores, the rest of your system will still be hibernating in idle mode.

The problem is that software designers just have to find ways to take advantage of multiple CPU cores, because there are natural limits to the clock rates of a CPU, and those limits have already been reached. Or doesn't it surprise anyone why we still only have 3 GHz CPUs? With the currently available silicon technology, they cannot run much faster. -- That -- is why everybody is adding more cores instead of increasing the GHz counter as they've done in the past.

The industry has reached an evolutionary dead end and is desperate to find a way out of it.

thinkdesign
May 21, 2010, 08:21 PM
Wirelessly posted (Mozilla/4.0 (compatible; MSIE 6.0; Windows CE; IEMobile 7.11) Sprint PPC6850SP)

Is Apple so interested in the phone and the iPad, because that's the ONLY place where they get free of problematic dependencies on other companies' cooperation, for both chip AND O/S? If my recent reading here has at least given me a big picture -- of the infinite tangle of why-they-cant-make-a-better-Air/laptop ... licensing, etc. ad nauseum ... maybe they SHOULD build UPWARDS from the iPhone O/S for most of the laptops. Maybe even buy a chip fabricator, too. It's not about what any part's specs are; it's what the whole package can DO, right?

Scottsdale
May 21, 2010, 09:01 PM
And even if you had one billion CPUs in your computer, that doesn't change a single bit of the fact that most of your daily tasks simply cannot be parallelized to take advantage of so many cores. Except for the typical number crunching tasks (which include things like DVD ripping or Photoshop filters), nothing will ever benefit from having 32 or more cores available. And an average human being also doesn't multi-task well enough to use dozens of applications simultaneously and there certainly is no need or economic justification to have all installed apps running at all times.

Especially games, which are usually the most demanding type of application, are traditionally single-threaded apps (which is the reason why Xbox 360 versions of games always kill PS3 ports of the same game: The PS3 has more cores, but the games CANNOT use them; the Xbox has fewer but FASTER cores). So even if the developers move physics calculations to one or two separate cores, the rest of your system will still be hibernating in idle mode.

The problem is that software designers just have to find ways to take advantage of multiple CPU cores, because there are natural limits to the clock rates of a CPU, and those limits have already been reached. Or doesn't it surprise anyone why we still only have 3 GHz CPUs? With the currently available silicon technology, they cannot run much faster. -- That -- is why everybody is adding more cores instead of increasing the GHz counter as they've done in the past.

The industry has reached an evolutionary dead end and is desperate to find a way out of it.

Your whole point of apps not taking advantage of multiple cores is based on x86 type applications. Obviously I said the applications and software needs to be written to take advantage of a system with many cores. But make no mistake about it, many cores is where we're going with ARM. And the software will take advantage of it.

The thing you cannot get is why it isn't being taken advantage of right now... but the thing is we don't have many cores to take advantage of right now. We have two cores in the average computer's CPU right now. Figure Grand Central Dispatch on ARM written for SoC ARM silicon that has 16, 24, 32 cores. It is in the future not right now.

The limitations you are thinking are limitations in your head based on the current situation. Go read a bit about ARM and realize many core ARM is the way of the future. You just have to believe that when 32 cores are available the Mac OS will take advantage of those cores differently than a current OS X Mac could take advantage of those Cores.

Make the leap in your head understanding where ARM is going and what software can do to take advantage of those types of SoC designs. ANYTHING is possible, especially when you have $40b in the bank to buy ARM technologies. It costs $1b to write an OS from scratch... it costs $1b to make a new ARM SoC from scratch... do you think Apple can afford it?

darkplanets
May 21, 2010, 09:20 PM
If we're looking at the distant future, as the title suggests; I'd say never.

There are too many greater options on the horizon, and I doubt ARM will be here to stay even +10 years on... thus is the natural advancement of technology. I mostly see ARM chips retaining their mobile dominance, perhaps even in laptops, but not much more.

Rather, I see 3 solutions in the future of computing that will, eventually (the title does say distant future), run OS whatever with something much different from x86, x86_64, or ARM.

1) Germanium chips. These things are very nice. If you don't know about Ge or about it's properties, I suggest looking them up. Right now the focus has been on hybrid SiGe chips, as per this link (http://www.zdnet.co.uk/news/processors/2005/08/05/ibm-gives-silicon-germanium-chips-a-speed-boost-39212172/), but in the future it will be pure Ge as the cost comes down.

2) Light based chips. They're all over the news; here's (http://www.cnn.com/2003/TECH/ptech/10/31/israel.lenslet.reut/index.html) a snippet. This will be the next big thing. This will happen before the following, due to a plethora of problems.

3) Quantum processors. We're a long way off on this-- there are many technology hurdles present that will limit our use of this technology.

In short, I don't see ARM being used for main-purpose computing in the future, both due to instruction set limitations and advancing technologies. Thats just me though.

Scottsdale
May 21, 2010, 09:33 PM
If we're looking at the distant future, as the title suggests; I'd say never.

There are too many greater options on the horizon, and I doubt ARM will be here to stay even +10 years on... thus is the natural advancement of technology. I mostly see ARM chips retaining their mobile dominance, perhaps even in laptops, but not much more.

Rather, I see 3 solutions in the future of computing that will, eventually (the title does say distant future), run OS whatever with something much different from x86, x86_64, or ARM.

1) Germanium chips. These things are very nice. If you don't know about Ge or about it's properties, I suggest looking them up. Right now the focus has been on hybrid SiGe chips, as per this link (http://www.zdnet.co.uk/news/processors/2005/08/05/ibm-gives-silicon-germanium-chips-a-speed-boost-39212172/), but in the future it will be pure Ge as the cost comes down.

2) Light based chips. They're all over the news; here's (http://www.cnn.com/2003/TECH/ptech/10/31/israel.lenslet.reut/index.html) a snippet. This will be the next big thing. This will happen before the following, due to a plethora of problems.

3) Quantum processors. We're a long way off on this-- there are many technology hurdles present that will limit our use of this technology.

In short, I don't see ARM being used for main-purpose computing in the future, both due to instruction set limitations and advancing technologies. Thats just me though.

Nice info... no matter what, we're going places... it might start with ARM and end elsewhere, but ARM in the mobile world is a definite going to happen in my eyes. I see a three to five year timeline.

darkplanets
May 21, 2010, 10:23 PM
Nice info... no matter what, we're going places... it might start with ARM and end elsewhere, but ARM in the mobile world is a definite going to happen in my eyes. I see a three to five year timeline.

I can definitely see both sides of the argument.

IMO, if what I previously posted doesn't come to bear fast enough or with complete implementation in the next 3-5 years; quite possibly to an almost definitely. 3-5 years would be a reasonable time frame; I'd say the latter of the two, as investment -> R&D -> manufacturing -> product does take some time, especially with newer technologies, and even more so with fresh meat (new design/manufacturing companies). As you said, ARM in the mobile world is a definite going to happen, but not past that, IMO. I don't see ARM SOC chips being used in desktops and all-in-ones, or anything higher than that, as ARM SOC really benefits mobility and hand helds, but not more than that in terms of general processor need. In fact, I can see ARM SOC's becoming the C2D of future notebooks; lower end options available for lesser price points.

As an aside, I wonder how much of an impact SiGe or Ge will have on ARM...
After all, it is already low voltage and low heat... so perhaps we could see even lower voltage and/or higher clock speed? That's just food for thought.

Winni does have a point though-- redoing either the instruction set or rewriting the Kernel would indeed be unwieldy and a PITA.... perhaps there is some validity in the fact that an OSX type OS won't exist on ARM SOCs... but then I once again digress to my original post, which I see as the most likely "long" term outcome.

Jason Beck
May 21, 2010, 10:39 PM
It won't. Thats really all I have to say about that. Mamma always said life is like a box of chocolates.

pharmx
May 22, 2010, 01:03 AM
Well one thing is for sure, Apple needs to increase their pool of skilled developers. Whether it's to write better drivers, utilize multi-threading to its full potential, adapt their OS, or work with new tech...it all starts with competent people being able to innovate on the software side of things to complement any advances on the hardware side. The recent acquisitions were a good start to boost their talent pool, but depending on what Apple has in mind for long term goals, they might need to beef it up some more.

Mhkobe
May 22, 2010, 01:38 AM
And even if you had one billion CPUs in your computer, that doesn't change a single bit of the fact that most of your daily tasks simply cannot be parallelized to take advantage of so many cores.

A few (a lot) years back Bill gates said something along the lines of 'Nobody will ever need more than 512kb of RAM.' He was wrong, and so are you; nobody could have predicted that we would be able to someday create 3D renderings that look real, or that we could take parts of these renderings and stick them into movies and make it look real. Even average consumers have benefited from these performance increases, with powerful word processing and formatting options etc. In the not so distant future we will have machines that can render in real time, with area mapped shadows, and beautiful reflections. With massively multithreaded apps for CPUs with massive numbers of cores I could stop hitting Cmd-R, and just watch my models change in perfectly rendered views.

There are too many greater options on the horizon, and I doubt ARM will be here to stay even +10 years on... thus is the natural advancement of technology. I mostly see ARM chips retaining their mobile dominance, perhaps even in laptops, but not much more.

ARM is a 10 billion dollar company that is relied upon by the majority of mobile device makers, x86 is old, and made by Intel. Intel is the new IBM (to my eyes as IBM was to Steve's).

darkplanets
May 22, 2010, 10:51 AM
ARM is a 10 billion dollar company that is relied upon by the majority of mobile device makers, x86 is old, and made by Intel. Intel is the new IBM (to my eyes as IBM was to Steve's).

I wasn't even implying x86 or x86_64... if you read my post I'm assuming a new instruction set will come out to supplement the advancing technologies I listed. There's no doubt x86 is old, same with x86_64, but despite how cool ARM SOCs are, I don't see them taking the place of heavy lifting processors, especially in the future, as there are better advanced technologies out there. Once again I maintain my two original posts-- if it does happen, it will happen only on low end notebooks within 3-5 years, but more than likely not.

Mhkobe
May 22, 2010, 11:21 AM
I wasn't even implying x86 or x86_64... if you read my post I'm assuming a new instruction set will come out to supplement the advancing technologies I listed. There's no doubt x86 is old, same with x86_64, but despite how cool ARM SOCs are, I don't see them taking the place of heavy lifting processors, especially in the future, as there are better advanced technologies out there. Once again I maintain my two original posts-- if it does happen, it will happen only on low end notebooks within 3-5 years, but more than likely not.

Ok, that's feasible, I was just a little shocked that you said in your other post that ARM would die or something; however, I completely agree with you that it is possible that ARM will never become the standard for heavy processing.

gnasher729
May 22, 2010, 11:38 AM
ARM is completely different from PPC. A Cortex 9 is an amazing chip with tons of capabilities. What I see happening in the future is not just two, four, or eight cores in a CPU. I see Apple software using 32 cores (just example could be amazingly more or half as many cores) in an ARM SoC design at a couple of GHz performing incredible vs. anything we see today in Intel Core i-series CPUs or even the future of those CPUs.

I don't know where you've been in the past, but the G4 was an excellent chip. Its only disadvantage was that Motorola aimed it at the embedded system market (the iPhone would have been right straight in its target market, but this was in 2001) and had no interest to go into the higher power higher clock speed market that Apple needed to compete with PCs.

With the advances in technology from 2001 to 2010, a G4 could easily compete with an ARM chip at same clock speed and the same power consumption. A G4-like processor with 16 cores + hyperthreading, 64 integer units, 32 FPUs, 32 Altivec units, would be easily possible.

But the excellent thing about Apple's software today is that nobody really cares what processor it runs on. Of the 100,000 iPhone developers, how many do you think have checked that the code executing on the iPhone is actually ARM code? If Apple had been lying all the time and there was a super-fast, super-low power 68040 processor inside every iPhone, how many developers would actually have noticed? Not many. And how many customers? Not one.