Welcome to the MacNN Forums.

If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below.

You are here: MacNN Forums > Hardware - Troubleshooting and Discussion > Mac Notebooks > Apple eyeing a transition to ARM CPUs for notebooks?

Apple eyeing a transition to ARM CPUs for notebooks?
Thread Tools
Lateralus
Moderator Emeritus
Join Date: Sep 2001
Location: Arizona
Status: Offline
Reply With Quote
May 6, 2011, 05:02 PM
 
There's been a lot of development on the ARM front in the past year, the most noteworthy being NVIDIA's entry into the game in a big way. It pretty much instantaneously put ARM on the map as a mid-market contender to x86 within the next few years.

And given Apple's heavy involvement with ARM theretofore, many eyes pointed toward Apple to see what this might mean for the Mac or other endeavors down the road.

Now the credible-rumor machine seems to be rolling...

Apple dumps Intel from laptop lines | SemiAccurate ||| Apple to Move from Intel to ARM Processors in Future Laptops? - Mac Rumors

What are everybody's thoughts on this? I for one cheer at the possibility.
I like chicken
I like liver
Meow Mix, Meow Mix
Please de-liv-er
     
mduell
Posting Junkie
Join Date: Oct 2005
Location: Houston, TX
Status: Offline
Reply With Quote
May 13, 2011, 07:51 PM
 
I don't buy it.
     
lpkmckenna
Addicted to MacNN
Join Date: Jul 2004
Location: Toronto
Status: Offline
Reply With Quote
May 13, 2011, 07:55 PM
 
No chance. Is Adobe gonna compile apps for ARM?

Besides, there is no benefit to this switch.
     
ilovemymac
Fresh-Faced Recruit
Join Date: Oct 2006
Status: Offline
Reply With Quote
May 22, 2011, 02:41 PM
 
but then again, apple's A5 chips are based on ARM tech and design right ? seems like Apple is going down the IBM PowerPC custom chips route again
     
P
Moderator
Join Date: Apr 2000
Location: Gothenburg, Sweden
Status: Offline
Reply With Quote
May 22, 2011, 03:59 PM
 
Real world tech did a pretty good analysis of that particular rumor. Short version: no way.

The only source for that rumor is SemiAccurate (MacRumors is just linking to them), which has a history of these far.out rumors that don't quite come true.
The new Mac Pro has up to 30 MB of cache inside the processor itself. That's more than the HD in my first Mac. Somehow I'm still running out of space.
     
chabig
Addicted to MacNN
Join Date: Jun 1999
Location: Las Vegas, NV, USA
Status: Offline
Reply With Quote
May 22, 2011, 07:46 PM
 
Originally Posted by ilovemymac View Post
seems like Apple is going down the IBM PowerPC custom chips route again
What are you talking about?
     
Waragainstsleep
Posting Junkie
Join Date: Mar 2004
Location: UK
Status: Offline
Reply With Quote
May 22, 2011, 08:28 PM
 
Apple had a lot of input into the designs of the PPC chips they used from IBM and Motorola. What they get form Intel is off the shelf. Sometimes they get it early but pretty much everything to date has eventually been available to anyone. The only possible exception being the first gen MacBook Air chips. I forget if they went on general sale eventually or not.

Chips like the G5 were not widely used outside of the Apple product line, hence Apple had a big influence on how they were designed. As an ARM licensee, they are in a position to have even greater influence on the chip designs to give them fully custom chips. You did see some IBM servers running PPC chips like Apple was using but the Xbox 360 chips were completely different again. Same for Cell chips. The mostly widespread use of PPC is actually as embedded systems. Chances are the engine management chip in your car is based on a G3. Maybe a G4.
I have plenty of more important things to do, if only I could bring myself to do them....
     
Chinasaur
Senior User
Join Date: Dec 2002
Location: Out West Somewhere....
Status: Offline
Reply With Quote
May 22, 2011, 08:56 PM
 
And Jobs has never had a problem with a "scorched earth" policy. And given the state of his health, what better way to go with a "bang" than to switch CPU's again.

To quote Spock "It's not logical...but it is often true.."
iMac - Late 2015 iMac, 32GB RAM
MacBook - 2010 MacBook, 1TB SSD, 16GB RAM
     
P
Moderator
Join Date: Apr 2000
Location: Gothenburg, Sweden
Status: Offline
Reply With Quote
May 23, 2011, 06:55 AM
 
It is true that Apple had some influence on how the G5 (PPC970, we should say) turned out, but it did so using IBM's building blocks. IBM does this for everyone: The chip is designed and developed by IBM, for a fee, and the chips sold at cost. The exception is the PPC970, which IBM developed at its own cost and then sold at a profit to Apple (to cover development cost), the same as the G3 and G4. The semi-official story from IBM is that IBM wanted to switch Apple to its standard model, and Apple refused, which lead to the Intel switch.

The lesson from this is that Apple DIDN'T want to design its own chips, and the one they did try to design (the G5 northbridge) was a dismal failure that IBM eventually had to bail them out from. If you look at the A5, it is a bog-standard Cortex A9 dualcore, another standard GPU (PowerVR SGX543MP2) and some RAM in one package. AFAIK, the parts aren't even on the same die. Apple chucked them together into one SOC to save space and manufacturing costs more than anything else.

It is correct that the main usage for PPC chips these days is in embedded usage. This is another point to remember - IBM and Motorola (then Freescale) competed with the PPC750 (the G3) versus the PPC 7450 (the G4) in the embedded space, and when Apple wanted features that didn't help either in that space (like a higher bandwidth FSB), neither could help for fear of alienating its much bigger market.
The new Mac Pro has up to 30 MB of cache inside the processor itself. That's more than the HD in my first Mac. Somehow I'm still running out of space.
     
Big Mac
Clinically Insane
Join Date: Oct 2000
Location: Los Angeles
Status: Offline
Reply With Quote
May 23, 2011, 09:15 AM
 
Great info P, particularly on PPC history.

I have a hard time believing ARM will ever be able to rival Intel and AMD on the desktop. The day that they do certainly hasn't arrived yet, and I don't see how it's possible they would be able to compete in that space. Maybe I'm missing some fundamental advantages of the ARM architecture that will make it zoom past the desktop market leaders in the future, but I don't see their competitive advantage outside of the mobile space.

Perhaps on the low-end one day Apple will replace the MacBook with an ARM based MacBook. But broadly across multiple and higher end product lines? Not going to happen unless something changes the performance equation massively in favor of ARM. Apple would be crazy to undertake another processor ISA transition just for the sake of doing it.
( Last edited by Big Mac; May 23, 2011 at 10:02 AM. )

"The natural progress of things is for liberty to yield and government to gain ground." TJ
     
OreoCookie
Moderator
Join Date: May 2001
Location: Hilbert space
Status: Online
Reply With Quote
May 23, 2011, 10:05 AM
 
I don't think Apple will switch within the next four~five years. Beyond that, I can't say.

Apple will only switch if it has a substantial advantage from doing so. It has experience with universal binaries and a two-platform strategy, but in any case, it's a painful transition for some. Right now, the advantage of ARM cpus is two-fold: (1) They're customizable and companies can pick and choose to design their own cpus. With Intel cpus, you have to take what you can get. While this is very important for smartphones and tablets and such, I think it will be less important for traditional computers and notebooks.

And (2), they use a lot less power. This will become less important in the future, I think, because ARM is pushing to extend its line-up to include high-performance cpus (see nVidia's project Denver which is very ambitious). But then, these high-performance cpus will be more power hungry. On the other hand, Intel has announced it will shift its mobile mainstream cpu target from making ~35 W TDP chips to ~17 W TDP chips.

Beyond the 4~5 years, I believe anything can happen. There is a nice piece written, saying that ARM could do to Intel what Intel has done to the RISC platforms workstations were based on. The argument is two-fold: (1) ARM cpus will remain cheaper and thus even if the cpus are slower, they will have an edge. (2) Intel's focus on cpu speed rather than price/performance goes in the wrong direction.
I don't suffer from insanity, I enjoy every minute of it.
     
Big Mac
Clinically Insane
Join Date: Oct 2000
Location: Los Angeles
Status: Offline
Reply With Quote
May 23, 2011, 11:21 AM
 
Very interesting observations, OC. I'm wondering though, what advantage does ARM have that makes it inherently cheaper than Intel? I know it has power advantages because that is ARM's core competency, but as you said the power savings come at performance costs. The customization factor interesting too - the notion that a microprocessor can be tailored to specific uses. But where does the inherent cost advantage come from? Doesn't Intel naturally enjoy much greater economics of scale?

"The natural progress of things is for liberty to yield and government to gain ground." TJ
     
OreoCookie
Moderator
Join Date: May 2001
Location: Hilbert space
Status: Online
Reply With Quote
May 23, 2011, 02:04 PM
 
Originally Posted by Big Mac View Post
I'm wondering though, what advantage does ARM have that makes it inherently cheaper than Intel?
Costs scale with die area and most ARM cpus have a very small die area. Also, Intel uses more expensive, but more sophisticated processes for its more expensive cpus (32 nm instead of 45 nm or 65 nm processes).

If you come to cpus of the same performance range (e. g. dual core Cortex A9s vs. a dual core Atom), the die size becomes comparable: Intel's Pineview dual core Atoms has a die size of roughly 90 mm^2 (this includes memory controller and graphics). This is comparable to Apple's A5 (~122 mm^2). The A4 was less than half the size (~50 mm^2). All of these are small compared to a modern Sandy bridge which is larger by a factor of 2~3 (depending on the model). The transistor count is substantially higher (~260 million vs. 995 million).
Originally Posted by Big Mac View Post
I know it has power advantages because that is ARM's core competency, but as you said the power savings come at performance costs.
If you use ARM cpus, you have a completely different mind set: you want to have a `sufficient' amount of cpu horse power. Sometimes, the lack of cpu power is compensated for by dedicated hardware (e. g. for hardware de- and encryption or de- and encoding). So you can buy a small cpu that is tailored to your needs. You can buy rather primitive ARM cpus that don't even have a Intel typically goes for broke with power. Also, it really shows that ARM cpus are built with a specific performance and W target. If you calculate the performance per milliwatt (DMIPS/mW), the Cortex A8 achieves 4~5 DMIPS/mW while a dual core Cortex A9 churns out 5.2~8 DMIPS/mW. Of course, as the clock speed increases, performance/Watt decreases.
Originally Posted by Big Mac View Post
The customization factor interesting too - the notion that a microprocessor can be tailored to specific uses.
IMO this is a major reason why the cell phone industry won't switch to x86 even if Intel manages to offer chips that consume as much power as rivaling ARM options. If you are a cell phone manufacturer, you have a multitude of options from many different companies that are built on top of the same architecture.
Originally Posted by Big Mac View Post
But where does the inherent cost advantage come from? Doesn't Intel naturally enjoy much greater economics of scale?
Not necessarily: per year, many more ARM cpus are sold than Intel cpus. It's kinda hard to come up with actual numbers (the sales figures I could find are in terms of US$ rather than number of cpus). Just to get a feel for how many ARM cpus are sold yearly: in 2005 it was at least 1 billion (I infer that from cell phone sales).

To be honest, though, this is split amongst many different kinds of cpus by many manufacturers. But as you can see, the ARM ecosystem is very, very large. That's why Intel is concerned, because a substantial part of the computer market is expected to move towards smartphones and tablets.
I don't suffer from insanity, I enjoy every minute of it.
     
P
Moderator
Join Date: Apr 2000
Location: Gothenburg, Sweden
Status: Offline
Reply With Quote
May 23, 2011, 04:37 PM
 
One point worth making: The processor with the highest performance per watt, right now, is probably Sandy Bridge. x86 doesn't necessarily mean that the performance per watt is bad, but the decoding logic that is required to translate x86 to something inserts a "floor" of base energy usage that means that a very low energy processor is hard to make on x86. Two things are happening to improve this over time:

* The decoder logic is basically fixed in number of transistors. Any energy saving from a new process translates into a lower floor, making this factor less of an issue
* Sandy Bridge seems to have finally managed to cache translated instructions without killing performance. Roughly 80% of instructions are cached in normal workload, meaning that the decoder logic can be turned off for 80% of the time. Including this feature in a future Atom (Silvermont/Airmont) will improve things drastically.

Taken together, this means that the problem x86 has had in getting to really low power designs will be significantly reduced. With Intel going to FinFET before anyone else and staying at least one node above the ARM pack means that they should be competitive by the time Airmont hits at least.
The new Mac Pro has up to 30 MB of cache inside the processor itself. That's more than the HD in my first Mac. Somehow I'm still running out of space.
     
OreoCookie
Moderator
Join Date: May 2001
Location: Hilbert space
Status: Online
Reply With Quote
May 24, 2011, 06:44 AM
 
Originally Posted by P View Post
One point worth making: The processor with the highest performance per watt, right now, is probably Sandy Bridge.
Interesting. Do you have any numbers, I'm curious?
Originally Posted by P View Post
* The decoder logic is basically fixed in number of transistors. Any energy saving from a new process translates into a lower floor, making this factor less of an issue
Doesn't the increase in transistor count make this point less and less of an issue?
Originally Posted by P View Post
Taken together, this means that the problem x86 has had in getting to really low power designs will be significantly reduced. With Intel going to FinFET before anyone else and staying at least one node above the ARM pack means that they should be competitive by the time Airmont hits at least.
I think Intel's advantage in manufacturing is quite significant. The die area per core of the A5 and Sandybridge is in the same ball park (I have divided the total die area by the number of cores which is a bit simplistic, I know), although the transistor counts are very different (the difference is roughly a factor of two).
I don't suffer from insanity, I enjoy every minute of it.
     
Veltliner
Mac Elite
Join Date: Nov 2006
Location: here
Status: Offline
Reply With Quote
May 27, 2011, 05:47 AM
 
(I don't think rewriting software just for laptops is in the cards. I see it as an experiment).

How many ARM processors would you need per laptop to get high end performance? A cloud of ARM? Lower end processors in large numbers like lower end PCs in clouds replace certain super computers?

Could this just be an experiment for the next generation of Intel chips, built in cooperation with Apple? And there was this one guy in that lab that was drunk and started the rumor of the switch to ARM?
     
P
Moderator
Join Date: Apr 2000
Location: Gothenburg, Sweden
Status: Offline
Reply With Quote
May 27, 2011, 07:00 AM
 
Originally Posted by OreoCookie View Post
Originally Posted by P
One point worth making: The processor with the highest performance per watt, right now, is probably Sandy Bridge.
Interesting. Do you have any numbers, I'm curious?
No, only reasoning. We know that Core 2 has a WAY higher performance per watt compared to Atom. We also know that Sandy Bridge is more efficient than the Nehalem/Westmere generations - the new icache would see to that alone - and AMD isn't really anywhere near at this point. The only remaining competitors are Itanium and Power7. Itanium is still stuck at 65nm and is missing many modern features. The next version adds quite a few and is projected to be a big boost, but only a boost up to Westmere levels.

So, Power7. Which is a monster, but it has a TDP of 200W. Even the low-end variants are 100W, while there are Sandy Bridge options at 17W TDP and even the quads can stay under 35W. Anandtech claims that even Westmere is more efficient than Power7. This is not evidence, but it seems likely.

Originally Posted by OreoCookie View Post
Doesn't the increase in transistor count make this point less and less of an issue?
If you keep adding transistors, the amount used by the more complicated decoding stage becomes less of an issue, yes. We don't know if we will keep adding transistors to mobile chips forever, though: there may not be many more node transitions out there, and we'd like to get more system logic onto the chip if we could.

Originally Posted by OreoCookie View Post
I think Intel's advantage in manufacturing is quite significant. The die area per core of the A5 and Sandybridge is in the same ball park (I have divided the total die area by the number of cores which is a bit simplistic, I know), although the transistor counts are very different (the difference is roughly a factor of two).
Every full node transition reduces the die area of any equivalent design by exactly half. The A5 is still at 45nm, while Sandy Bridge is at 32nm, so it makes sense that each core is that much bigger. Apple really needs to get to 32nm, or at least to some nice half node like 40 nm, sooner rather than later, and I wonder if the plan isn't to do a tick-tock maneuver and make an "A5+" at 32nm for the iPhone 5.
The new Mac Pro has up to 30 MB of cache inside the processor itself. That's more than the HD in my first Mac. Somehow I'm still running out of space.
     
ajprice
Professional Poster
Join Date: Dec 2000
Location: UK
Status: Offline
Reply With Quote
May 27, 2011, 07:41 AM
 

It'll be much easier if you just comply.
     
Koralatov
Senior User
Join Date: Jun 2007
Location: Aberdeen, UK
Status: Offline
Reply With Quote
May 27, 2011, 01:28 PM
 
Originally Posted by ajprice View Post
Apparently they've made an A5 Macbook Air test machine...
I have a feeling that the MacBook ARM probably won’t get past the prototype stage. Having two similar but distinct OSes running on different architectures is one thing, but having both running on ARM and one of them also running on Intel seems far too ambitious. It’s not like they could solve the problem by transitioning the whole Mac line onto ARM processors either; getting Mac Pro-level performance out of A5 chips is just untenable at present. Maybe when the A10 rolls around, but even then, I can’t see it — that would assume no progress at all from Intel over the next five years.
     
Waragainstsleep
Posting Junkie
Join Date: Mar 2004
Location: UK
Status: Offline
Reply With Quote
May 27, 2011, 01:43 PM
 
But how many A5 chips could you buy for the cost of a single Xeon?
I have plenty of more important things to do, if only I could bring myself to do them....
     
Koralatov
Senior User
Join Date: Jun 2007
Location: Aberdeen, UK
Status: Offline
Reply With Quote
May 27, 2011, 02:58 PM
 
Originally Posted by Waragainstsleep View Post
But how many A5 chips could you buy for the cost of a single Xeon?
About 35? (Based on the cost of an A5 found here — $14 — and an average price of $500 for a Xeon.)

Even if you could leverage all that processing power in a meaningful way (if memory serves, parallelism is really, really hard to program), don’t you basically end up using silly amounts of power to get a more complicated-to-program-for system? And to get roughly comparable performance in a MacBook Pro, how many A5s would you need? Four? Six? Eight? I think that many would negate the power-saving intention.
     
Waragainstsleep
Posting Junkie
Join Date: Mar 2004
Location: UK
Status: Offline
Reply With Quote
May 27, 2011, 03:30 PM
 
Makes you wonder though. Probably not suited to the Mac Pro (not yet at least) but how many average users really put their iMacs or MacBooks through their paces? Not many. Half a dozen A5s in an iMac could be a perfectly usable machine. Could get even thinner too if it moves to the Air-style SSDs and dumps the optical drives.

I suppose another big question is how does the PowerVR system stack up against the Intel HD GPU core? I assume it would be no big deal to pair a 6970 with an ARM CPU if needed, but in such a unit the heat sink on that GPU would be the determining factor as far as the overall thickness of the unit is concerned.

I was amazed how thin some of the new LED LCD TV and monitors are compared to my CCFL TV. Its a big difference.
I have plenty of more important things to do, if only I could bring myself to do them....
     
Ham Sandwich
Guest
Status:
Reply With Quote
May 28, 2011, 04:31 PM
 
I think the ARM rumor will be the biggest "failed rumor" of the year.
     
OreoCookie
Moderator
Join Date: May 2001
Location: Hilbert space
Status: Online
Reply With Quote
May 30, 2011, 04:53 AM
 
Originally Posted by Veltliner View Post
Lower end processors in large numbers like lower end PCs in clouds replace certain super computers?
They already have years ago: IBM makes super computers based on embedded PowerPC chips. The fastest one, JUBELE has 294,912 cores and was the fourth fastest super computer when it was released in 2009. This line of super computers sprang from research project which I read about years ago from The Register. The idea was to create a super computer the size of a washing machine.

Compared to the first generation, there have been a number of improvements: the latest generation of cpus used by IBM is a multicore design and integrates a lot of the necessary circuitry to link several cpus together. Without a doubt, the same could be done with ARM-based cores in the not too distant future.
I don't suffer from insanity, I enjoy every minute of it.
     
OreoCookie
Moderator
Join Date: May 2001
Location: Hilbert space
Status: Online
Reply With Quote
May 30, 2011, 05:00 AM
 
Originally Posted by Andrej View Post
I think the ARM rumor will be the biggest "failed rumor" of the year.
I don't think it's a failed rumor, I'm pretty certain the ARM-based MacBook Air is real. But it's a technology study rather than a prototype. Just like Project Marklar (OS X on x86) was real and existed from the start. (The purpose of Project Marklar was not just to keep the option of offering OS X on x86-based Macs, but rather to ensure OS X is not tied to any particular architecture. Indeed, OS X exists on three architectures right now, x86, PowerPC (deprecated) and ARM.)

I'm pretty sure Apple wants to see what real-world advantages switching to ARM cpus has: how much does the battery life really increase? What is the impact to performance? I think Apple is crazy enough to do the switch if they decide that the benefits outweigh the risks.

In any case, it'll be some time until Apple wants to switch away from x86 even for the AirBook: the current generation of ARM cpus just isn't beefy enough. But nVidia's next-gen Kal El is (according to them) significantly faster than Core 2 Duos and they've prototype silicon up and running. So they're catching up. I suppose in 3~5 years, the question of whether or not to switch is more realistic -- at least for parts of Apple's line-up.
I don't suffer from insanity, I enjoy every minute of it.
     
Waragainstsleep
Posting Junkie
Join Date: Mar 2004
Location: UK
Status: Offline
Reply With Quote
May 30, 2011, 05:47 PM
 
They'll look to switch the consumer portables first. The MacBook/Air. I think you'd have to keep an option for a heavy lifting version of the iMac and the Mac Mini though. Or maybe they'll use an ARM powered iMac range to boost sales of Mac Pros.
I have plenty of more important things to do, if only I could bring myself to do them....
     
mduell
Posting Junkie
Join Date: Oct 2005
Location: Houston, TX
Status: Offline
Reply With Quote
May 31, 2011, 12:45 AM
 
Originally Posted by Koralatov View Post
About 35? (Based on the cost of an A5 found here — $14 — and an average price of $500 for a Xeon.)
Price and cost are not the same thing.
     
Koralatov
Senior User
Join Date: Jun 2007
Location: Aberdeen, UK
Status: Offline
Reply With Quote
Jun 1, 2011, 01:22 PM
 
Originally Posted by mduell View Post
Price and cost are not the same thing.
You’re absolutely right, and I should know that from working with prices and costings all day long. A revised estimate based on the “Recommended Channel Price” (I’m assuming that’s cost-to-OEM) of $387: one Xeon would buy you approximately 27 A5s.

I still don’t think it’s likely to happen in the near future.
     
Big Mac
Clinically Insane
Join Date: Oct 2000
Location: Los Angeles
Status: Offline
Reply With Quote
Jun 1, 2011, 02:21 PM
 
What do you guys make of Intel's comment that it's willing to work with the likes of Apple to develop custom processors? It seems like a rather substantial admission that Intel sees a real and increasing threat from ARM.

"The natural progress of things is for liberty to yield and government to gain ground." TJ
     
imitchellg5
Posting Junkie
Join Date: Jan 2006
Location: Colorado
Status: Offline
Reply With Quote
Jun 1, 2011, 02:44 PM
 
I think an ARM-powered MacBook Air with iOS makes a lot of sense.
     
Big Mac
Clinically Insane
Join Date: Oct 2000
Location: Los Angeles
Status: Offline
Reply With Quote
Jun 1, 2011, 02:50 PM
 
But then it wouldn't be a MacBook Air it would be an iBook(?!)Air instead.

"The natural progress of things is for liberty to yield and government to gain ground." TJ
     
OreoCookie
Moderator
Join Date: May 2001
Location: Hilbert space
Status: Online
Reply With Quote
Jun 1, 2011, 03:03 PM
 
Intel is trying to get into a new market and the only thing they have to offer is superior manufacturing. As far as I understand, they would even be willing to fab ARM-based chips as well if the offer is juicy enough. As it stands, nobody really wants to shoehorn an x86-based cpu into a tablet or smartphone at this point, the technology isn't there -- yet (?). In three years, the story may be different.

It's a race from two directions: bottom-up for ARM and top-down for Intel. In a sense, ARM doesn't even have to make big inroads into the notebook or server business to turn a tidy profit, their business model is very different and they're much more content with smaller margins. Intel is used to rather big margins and it hurts to give them up during a race to the bottom.
I don't suffer from insanity, I enjoy every minute of it.
     
mduell
Posting Junkie
Join Date: Oct 2005
Location: Houston, TX
Status: Offline
Reply With Quote
Jun 2, 2011, 12:56 PM
 
Originally Posted by Koralatov View Post
You’re absolutely right, and I should know that from working with prices and costings all day long. A revised estimate based on the “Recommended Channel Price” (I’m assuming that’s cost-to-OEM) of $387: one Xeon would buy you approximately 27 A5s.
That's price to OEM. You can't compare ARM fab costs to Xeon prices.
     
Waragainstsleep
Posting Junkie
Join Date: Mar 2004
Location: UK
Status: Offline
Reply With Quote
Jun 2, 2011, 03:33 PM
 
Originally Posted by mduell View Post
That's price to OEM. You can't compare ARM fab costs to Xeon prices.
Why not?
In this instance you can. Apple can fab its own ARM chips and there is a cost per chip for that. Apple's other choice is to buy chips from Intel. If Intel choose to charge $387 a chip and Apple can fab A5s for $14, then thats the comparison Apple have to make when they design hardware.
Obviously you would need some fancy chipsets to run 27 ARM chips on one board. Would that cost more than the Intel chipsets though? Might have to settle for 16 or 24 A5s.

Realistically, if they were going to do this I suspect a certain number of A5s running the OS and then a bunch more on PCIe cards like those Cell based render node cards you can get to do the heavy lifting.
( Last edited by Waragainstsleep; Jun 2, 2011 at 03:47 PM. )
I have plenty of more important things to do, if only I could bring myself to do them....
     
Big Mac
Clinically Insane
Join Date: Oct 2000
Location: Los Angeles
Status: Offline
Reply With Quote
Jun 2, 2011, 03:50 PM
 
But can 20 small ARM chips really scale to the single thread performance of one Core i7? Sounds dubious unless ARM compilers are terrifically advanced at parallelization.

"The natural progress of things is for liberty to yield and government to gain ground." TJ
     
   
Thread Tools
 
Forum Links
Forum Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Top
Privacy Policy
All times are GMT -4. The time now is 05:52 AM.
All contents of these forums © 1995-2017 MacNN. All rights reserved.
Branding + Design: www.gesamtbild.com
vBulletin v.3.8.8 © 2000-2017, Jelsoft Enterprises Ltd.,