Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

GeekLawyer

macrumors 68020
Bah, your just old and Jaded ;). There is plenty of magic to be found when using one. It's the UI. It has drastically shrunk the gap in what you want it to do and how you do it. The larger screen makes it far more instinctual than an iPhone ever was. I have shown it to people that struggle to use an iPhone and they understand and use the iPad like a pro in seconds. I have been truly astounded at how quickly these people understand the device.
I agree with everything that you said. Except the parts about me being old and jaded. ;)

And still: they'll never find the magic. It results from the gestalt. It's in the experience. Fancy microscopes and x-rays and spec-sheets are useless in divining the revolution that we're now witnessing in micro-computer use. :cool:
 

Jakeoster

macrumors regular
Mar 12, 2010
153
80
It's fast enough for what it does. I wasn't expecting any major speed increases because that would start to cause fragmentation within the app store because then everyone would expect the 4th gen iPhone and iPod touch to match the ipad's capabilities. Apple never touted that the iPad was faster than it really is, that was everyone else blowing it up. The only way to combat fragmentation of the app store in the future would be for apps to detect what device it is running on and adjust itself automatically, which I could see happening and would be a smart move.
 

alent1234

macrumors 603
Jun 19, 2009
5,688
170
From wiki:

-Lower transaction costs
-Synchronization of supply and demand along the chain of products
-Lower uncertainty and higher investment
-Ability to monopolize market throughout the chain by market foreclosure

They can ask ARM to build them a custom Cortex A8 and be completely at Cortex's beck-and-call......

or they can buy the A8 chip design, mod them on their own, and then get TSMC to print them.

It removes a dependency on an external party. Look at how the iPad was delayed because of their screen manufacturer...

There are both money and control issues at play.

They want to rely on external parties as little as possible.

ARM doesn't build anything. they license their IP out, you customize it and build your own CPU. every year or so ARM releases the IP for a better CPU.

Since battery tech probably won't improve by much soon, Apple took the next logical step and customized the hardware to maximize battery life. the A4 chip is a normal ARM core CPU with less circuitry. ARM chips are found in a lot of devices so it's becoming like Intel now. they have to add a lot of features some won't need to sell to as many customers as possible
 

marksman

macrumors 603
Jun 4, 2007
5,764
5
I will wait for Rev2 of iFixit's teardown.

Recent history shows they are usually wrong about something, and then come back and fix it.
 

econgeek

macrumors 6502
Oct 8, 2009
337
0
Ok, lets be real about what this is and what it isn't: It's chip die porn.

It tells us that Apple is using hte POP assembly method, and that they made their own die. The only thing iFixit was able to identify from this is that the RAM dies were made by samsung.

This tells us nothing about which version of the ARM core Apple used, how many cores Apple used, or which GPU apple used.

Such analysis could reveal it if someone were to compare the patterns to pre-existing chips.

But the idea that Apple is using off the shelf parts here and that we need to identfy them is wrong.

Apple is making a custom chip here. The A4 is custom in every meaning of the word. They have licensed the ARM and Imagination IPs, but that doesn't mean they just slap "a cortext" and "a PowerVR GPU" in there.

That means that Apple can be as custom as they want without violating the rights of either of these parties. They can choose what they want from any of the cores, or partial cores or individual circuits that these people offer.

The A4 may not be fully custom, as a new effort would suggest prudence before going wild.

But it doesn't mean Apple can't go fully custom, and it certainly doesn't mean that this is any sort of a commodity.

The idea that this is a "Cortex" anything is fundamentally wrong. It is an A4. It is it's own thing.

People are focused on individual chips because of successful marketing of specific chips by intel. But that's not appropriate here.

The chip here is the A4.
 

cmaier

Suspended
Jul 25, 2007
25,405
33,471
California
Ok, lets be real about what this is and what it isn't: It's chip die porn.

It tells us that Apple is using hte POP assembly method, and that they made their own die. The only thing iFixit was able to identify from this is that the RAM dies were made by samsung.

Apple doesn't own a fab. The CPU die was also made by Samsung.

This tells us nothing about which version of the ARM core Apple used, how many cores Apple used, or which GPU apple used.

Such analysis could reveal it if someone were to compare the patterns to pre-existing chips.

That analysis has been done. It's cortex-a8. Also, one doesn't need to compare the patterns - if there are two cores, it would be obvious without needing to compare to anything else.

But the idea that Apple is using off the shelf parts here and that we need to identfy them is wrong.

Apple is making a custom chip here. The A4 is custom in every meaning of the word. They have licensed the ARM and Imagination IPs, but that doesn't mean they just slap "a cortext" and "a PowerVR GPU" in there.

That means that Apple can be as custom as they want without violating the rights of either of these parties. They can choose what they want from any of the cores, or partial cores or individual circuits that these people offer.

The A4 may not be fully custom, as a new effort would suggest prudence before going wild.

One of these two conflicting statements is correct. Apple used hard blocks supplied by ARM for most of the design. They did not do full custom physical design at the core level.

But it doesn't mean Apple can't go fully custom, and it certainly doesn't mean that this is any sort of a commodity.

The idea that this is a "Cortex" anything is fundamentally wrong. It is an A4. It is it's own thing.

No it's not. It is an implementation of the licensed cortex-a8 ARM microarchitecture. Cortex refers to SoC IP, not to a specific chip.

People are focused on individual chips because of successful marketing of specific chips by intel. But that's not appropriate here.

The chip here is the A4.

Ok.
 

BruiserBear

macrumors 6502a
Jul 24, 2008
584
534
As stated in the opening article..."There's nothing revolutionary here"

What do you expect, it's made by Apple.

I don't give a damn if they call the chip the 'Mickeymouse 500". If it renders web pages as fast as the iPad does, and gives me 10 hours of battery life, it's cool with me.

Haters just seem to scan any article about the iPad right now just desperate to post something negative about it. Kinda sad. :)
 

lilo777

macrumors 603
Nov 25, 2009
5,144
0
Ok, lets be real about what this is ...
The chip here is the A4.

You are right. The chip is A4. It's based on Cortex-A8 micro-architecture (same as iPhone 3GS). Combined with the fact that it has 1MHz frequency and 256 MB RAM it should be a little faster than CPU in 3GS but probably slower than Nexus One (also Cortex A8 based, same 1Mhz but 512MB RAM).
 

NT1440

macrumors G5
May 18, 2008
14,693
21,235
You are right. The chip is A4. It's based on Cortex-A8 micro-architecture (same as iPhone 3GS). Combined with the fact that it has 1MHz frequency and 256 MB RAM it should be a little faster than CPU in 3GS but probably slower than Nexus One (also Cortex A8 based, same 1Mhz but 512MB RAM).

I don't know, if it clocks in at 1Mhz, it should be significantly slower than any phone out there.....
 

cmaier

Suspended
Jul 25, 2007
25,405
33,471
California
You are right. The chip is A4. It's based on Cortex-A8 micro-architecture (same as iPhone 3GS). Combined with the fact that it has 1MHz frequency and 256 MB RAM it should be a little faster than CPU in 3GS but probably slower than Nexus One (also Cortex A8 based, same 1Mhz but 512MB RAM).

In tests it is faster than the Nexus One. Not sure what they clock nexus one at, but I think the memory bus on the A4 is twice as wide as the Nexus One's, and I also think the memory controller operates at a higher frequency.
 

pubwvj

macrumors 68000
Oct 1, 2004
1,901
208
Mountains of Vermont
Revolutionary it is - in how everything is brought together. It is the synthesis, the whole. Nobody else had done it at all never mind as well. This is what Apple does and does well. This is why I've bought Apple products for over 30 years. They let me do my work.
 

NT1440

macrumors G5
May 18, 2008
14,693
21,235
In tests it is faster than the Nexus One. Not sure what they clock nexus one at, but I think the memory bus on the A4 is twice as wide as the Nexus One's, and I also think the memory controller operates at a higher frequency.

I'm pretty sure the Nexus is clocked at 1Ghz
 

iSee

macrumors 68040
Oct 25, 2004
3,539
272
What can they actually learn from x-raying a chip like that?
The packaging details are kind of interesting...
I was hoping they would somehow delve into the inner workings of the chip in at least some detail. That's were I would expect to find the good stuff...
 

lilo777

macrumors 603
Nov 25, 2009
5,144
0
In tests it is faster than the Nexus One. Not sure what they clock nexus one at, but I think the memory bus on the A4 is twice as wide as the Nexus One's, and I also think the memory controller operates at a higher frequency.

I do not know if they use any type of memory virtualization in these phones (it looks like they do not) but if they do, then most likely A4 will be faster on apps that fit into 256MB and slower on apps with higher memory requirements.
 

BVerites

macrumors newbie
Apr 6, 2010
6
0
In reality Apple can't become independent in ARM chips since it doesn't hold the IP.

Not true, Apple has licensed the IP to design ARM chips and uses Samsung which has a license to manufacture them.

__________________
-as
Scrabble - Have you played it with your friends, your colleagues and your relatives? Why not?
 

deconstruct60

macrumors G5
Mar 10, 2009
12,309
3,900
I do not know if they use any type of memory virtualization in these phones (it looks like they do not) but if they do, then most likely A4 will be faster on apps that fit into 256MB and slower on apps with higher memory requirements.

Would not want to do much memory virtuallization since storage space is limited. And on the Nexus One some of the storage space is on a SD card. That's relatively super slow access times. SD cards are slower than hard drives in access times.

Also since don't cap the 3rd party apps, the extra memory on the Nexus one can disappear with multiple concurrent apps.
 

econgeek

macrumors 6502
Oct 8, 2009
337
0
Apple doesn't own a fab. The CPU die was also made by Samsung.

Obviously Apple doesn't have a fab, and it really doesn't matter who they used to fabricate the die.

We've seen no evidence that this particular die was made by Samsung. That you would assert this as if it were a fact is part of the problem. It may be likely, consistent, or reasonable to expect that the die was probably made by Samsung.... but it is not a known fact.

This may seem pendantic, but the problem here, and with much reporting about these things, is that people assert speculation as if it were fact, and eventually, since Apple doesn't respond to speculation, people start believing it to be true.

It's cortex-a8.

You do not know this. You are expressing an opinion based on ignorance of the nature of CPUs. ARM licenses designs, but they also license the right to make derivative designs. We do not know whether Apple has used an ARM off the shelf design or produced a derivative.

Also, one doesn't need to compare the patterns - if there are two cores, it would be obvious without needing to compare to anything else.

Again, your ignorance of this situation, leads you to assert as fact a supposition made by someone else who is also ignorant.

The logic goes something like "The cortex A9 has a two core version, therefore if this CPU has one core it must be a cortex A8."

There are numerous errors here. First off, there is a single core version of the cortex A9.

More importantly, though, is the presumption that Apple cannot make a derivative CPU, and that they must have chosen an A8 or an A9.

This is a presumption based on a lack of understanding on the part of the people making it.

They think of Apple as a computer maker, and so they presume that Apple can't produce it's own cores. (And really, they don't even know what a core actually is.)

Apple used hard blocks supplied by ARM for most of the design.

When you state facts that you do not know to be true, are you a liar? Or does one have to prove that those facts are false before one calls another a liar?\

At any rate, you are making assertions of fact here for which you do not have any knowledge.... unless you work for Apple and are divulging trade secrets.


They did not do full custom physical design at the core level.

Speaking of contradictions this is actually a nonsensical statement on one hand, and something you cannot, again, know on the other.

But it doesn't mean Apple can't go fully custom, and it certainly doesn't mean that this is any sort of a commodity.

When you talk about "Cortex-A8" you are talking about a commodity. The people who question whether this is an A8 or an A9 think that these are CPUs like the 386 or 486. They are not even aware of the fact that there were quite a few variations of the 386. That's far beyond their understanding.

This is why you said presumed that if it had 1 core it's an A8. This is as silly as presuming that because it has one core it's an Intel Strong Arm design from earlier in the decade.

Just because something has four wheels does not make it a truck, and preclude it from being a car.

No it's not. It is an implementation of the licensed cortex-a8 ARM microarchitecture

First off, again, you don't know this. You're just making assertions based on a lack of understanding. Secondly, even if it includes Cortex A8 technology, Apple has a derivatives license and thus they can do anything they want. Thirdly your assertion that it is A8 vs. A9 is based on a lack of understanding, and presumption that its impossible for Apple to have made their own derivative.

Cortex refers to SoC IP, not to a specific chip.

Do you know what IP is? Do you know what it means to license IP?

Your argument is that all AMD processors are actually intel processors because they contain the x86 microarchitecture.

You are right. The chip is A4. It's based on Cortex-A8 micro-architecture (same as iPhone 3GS). Combined with the fact that it has 1MHz frequency and 256 MB RAM it should be a little faster than CPU in 3GS but probably slower than Nexus One (also Cortex A8 based, same 1Mhz but 512MB RAM).

Yeah, you don't know anything about it, obviously. You're just repeating nonsense you read on the web from other people know know nothing about it.

Further, Apple hasn't shipped a computer based on a 1MHz CPU since the days of the Apple II, which, if I recall correctly, was 1.47MHz.

It used to be that people who were engineers would talk about things, and the rest of the world wouldn't.

Now people who know nothing about something read a site like ars technica, written by people who know marginally more than nothing about something, and take it as the gospel truth.

And then they go onto forums and argue with people who do know something about the subject.

But since they know nothing, they don't even comprehend what the person who does know something said, and so their arguments are nonsense.

Hell, actually making arguments would be a nice change.

You just reasserted a big string of terms whose meaning you can't fathom. But you've got the certainty of the ignorant to back it up, don't you?

In tests it is faster than the Nexus One. Not sure what they clock nexus one at, but I think the memory bus on the A4 is twice as wide as the Nexus One's, and I also think the memory controller operates at a higher frequency.

Yeah, the memory bus claim is pretty funny. I can't say it is not twice as wide, but I can say its absolutely hilarious to see people say that the CPU is "just a cortex A8" and then claim the "memory bus is twice as wide". Unless people are in the habit of making SoCs with a memory controller whose bus is half the width of the CPUs, it would be pretty pointless to double the width of the memory bus without redesigning the CPU, in which case it is not a "cortex A8", which while we're at it, is more like a confederation of designs than a single entity.

And evne then it is a design.

Reality is, nobody really knows, and likely the A4 is as different from the Cortex A8 as the Cortex A9 is.

I do not know if they use any type of memory virtualization in these phones (it looks like they do not) but if they do, then most likely A4 will be faster on apps that fit into 256MB and slower on apps with higher memory requirements.

In iPhone OS applications are constrained in the amount of memory they are allowed to use and there is a delegate method which is called to tell the application that it needs to free up some memory.

For a variety of reasons, the OS and Apps all fit in memory (which is not to say there's not some use of mass storage for things other than file storage.)
 

cmaier

Suspended
Jul 25, 2007
25,405
33,471
California
Obviously Apple doesn't have a fab, and it really doesn't matter who they used to fabricate the die.

We've seen no evidence that this particular die was made by Samsung. That you would assert this as if it were a fact is part of the problem. It may be likely, consistent, or reasonable to expect that the die was probably made by Samsung.... but it is not a known fact.

The part numbers on the chip package indicate that all three dies were fabricated by Samsung.


You do not know this. You are expressing an opinion based on ignorance of the nature of CPUs. ARM licenses designs, but they also license the right to make derivative designs. We do not know whether Apple has used an ARM off the shelf design or produced a derivative.

I was one of the team of 18 who designed Athlon 64 and Opteron. I designed CPUs for more than a decade for AMD, Sun, and Exponential Technology. I have a Ph.D. in electrical engineering and my dissertation involved research on a particular type of CPU.

I don't think I am "ignorant of the nature of CPUs."


Again, your ignorance of this situation, leads you to assert as fact a supposition made by someone else who is also ignorant.

The logic goes something like "The cortex A9 has a two core version, therefore if this CPU has one core it must be a cortex A8."

There are numerous errors here. First off, there is a single core version of the cortex A9.

More importantly, though, is the presumption that Apple cannot make a derivative CPU, and that they must have chosen an A8 or an A9.

Except that chipworks looked at it and it very closely matched the layout of the cortex A8 chip from samsung. Again, who's ignorant here?

They think of Apple as a computer maker, and so they presume that Apple can't produce it's own cores. (And really, they don't even know what a core actually is.)

I designed the first mass market 2-core CPU. How about you?


Do you know what IP is? Do you know what it means to license IP?

Now you're just being silly. I'm an IP attorney. I write, and litigate, IP licenses (among other things). What do you know about licensing IP?

Your argument is that all AMD processors are actually intel processors because they contain the x86 microarchitecture.

No it's not. My argument is that I can look at a die photo of an x86 chip and I can immediately tell you whether it is based on Intel's i-series or on AMD's Athlon 64 series. You, however, cannot.

Yeah, the memory bus claim is pretty funny. I can't say it is not twice as wide, but I can say its absolutely hilarious to see people say that the CPU is "just a cortex A8" and then claim the "memory bus is twice as wide". Unless people are in the habit of making SoCs with a memory controller whose bus is half the width of the CPUs, it would be pretty pointless to double the width of the memory bus without redesigning the CPU, in which case it is not a "cortex A8", which while we're at it, is more like a confederation of designs than a single entity.

And evne then it is a design.

Reality is, nobody really knows, and likely the A4 is as different from the Cortex A8 as the Cortex A9 is.

The memory bus is decoupled from the core. It is fairly trivial to re-design the memory controller (which occupies a tiny sliver of the silicon) and not touch the core at all.
 

econgeek

macrumors 6502
Oct 8, 2009
337
0
Not true, Apple has licensed the IP to design ARM chips and uses Samsung which has a license to manufacture them.

I think the misconception that many of the people spreading misinformation are operating under is that this is just a license to drop pre-made cores onto an SOC.

Apple's License with ARM's specific terms are secret, which actually implies that Apple probably has special rights.

But even regular licensees have the choice of buying a license to produce derivative products.

PA Semiconductor's specialty was improving performance of stock designs by reworking at a very low level their power and speed.

Intrinsity is in the business of reworking ARM designs at a fairly low level.

Thus, Apple is clearly signaled an intent of reworking ARM designs at a fairly low level.
 

cmaier

Suspended
Jul 25, 2007
25,405
33,471
California
I think the misconception that many of the people spreading misinformation are operating under is that this is just a license to drop pre-made cores onto an SOC.

Apple's License with ARM's specific terms are secret, which actually implies that Apple probably has special rights.

But even regular licensees have the choice of buying a license to produce derivative products.

PA Semiconductor's specialty was improving performance of stock designs by reworking at a very low level their power and speed.

Intrinsity is in the business of reworking ARM designs at a fairly low level.

Thus, Apple is clearly signaled an intent of reworking ARM designs at a fairly low level.

No it wasn't.
 

econgeek

macrumors 6502
Oct 8, 2009
337
0
TExcept that chipworks looked at it and it very closely matched the layout of the cortex A8 chip from samsung.

Got a citation for that? The ifixit article states that they are making this presumption based on the number of cores, not any detailed analysis of the layout.

I designed the first mass market 2-core CPU.
Now you're just being silly. I'm an IP attorney. I write, and litigate, IP licenses (among other things). What do you know about licensing IP?

While you're pulling degrees out of your ass, you might as well assert you're a software developer and you wrote the original task scheduler for NeXT.

At any rate, if you were an attorney and familiar with the ARM licensing scheme, you'd know that ARM sells derivative licenses, meaning that Apple can modify the cores as much as they like.

If you were familiar with the history of Apple over recent years, you'd know that Apple has publicly revealed their intentions to use PA Semiconductor to design chips for iPhone OS devices, and that they are likely the purchasers of Intrensity.

Thus, your assertion that the A4 has no changes from Apple from the Cortex A8 design is silly on the face of it. Further, there is no reason to believe that Apple *must* have used a particular design as-is and not changed aspects of it which result in feature changes too small to see with the naked eye, but still result in improved performance or battery life.

No it's not. My argument is that I can look at a die photo of an x86 chip and I can immediately tell you whether it is based on Intel's i-series or on AMD's Athlon 64 series. You, however, cannot.

Any high school kid whose seen die photos of the two chips can tell the difference.

Your argument is that you've seen the die of the A4 and that it is exactly the die of the Cortex A8. This is nonsense on multiple levels, and unless you're privy to die photos the rest of us haven't seen, an obvious lie.

But I love how you're an attorney and chip designer. That's choice!
 

econgeek

macrumors 6502
Oct 8, 2009
337
0
The memory bus is decoupled from the core. It is fairly trivial to re-design the memory controller (which occupies a tiny sliver of the silicon) and not touch the core at all.

I love it, a chip designer that thinks a 64bit word memory controller with a 32bit word core will result in better performance!

Of course, the fact that you aren't responding to the argument I was making, and instead are knocking down a strawman to attempt to characterize me.... after your orgy of argument from false authority, is pretty funny.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.