Apple to expand CPU design group beyond iPad A4 - Page 4

More to the point, Samsung would be in hot water trying to leach off the A4 as it's been patented for it's implementation and modified implementation licensed from ARM.

Look up Geekbench for iPhone and iPad. I think the CPU performance isn't as bad as many people seem to believe ...It depends on the software.


Check out CineXPlayer, VLC and OPlayer. They play MKV, XVID, etc. etc. and things have been gradually improving. CineXPlayer can do standard def XVID pretty damn well with sound in sync and everything. The difference between this and H.264 on iOS is that it is done purely with "standard CPU routines" without hardware acceleration. I haven't tried 720p XVIDs, I should give that a shot on my iPad. I think it's not too bad either, but not 100% sure on this.

CineXPlayer is an example of an app that makes me think, while Intel and X86 Win/OSX on Mac has at least a five-year future, ARM design and especially, software design for ARM is going to make some real leaps and bounds. When we have 2ghz dualcore ARM CPUs with 4x the graphic horsepower and maybe OpenCL support or something, you can imagine in 2-4 years ARM "MacBook Air"s being quite possible.

The killer feature for ARM is this: It would appear that it is easier to scale ARM *upwards* in performance while reaching real nice power and battery targets, rather than scaling *down* Intel and even AMD CPUs. The iPad is perfect for this middle ground between phone and laptop and desktop, and where ARM will shine over the next 5 years... However remember phone, tablet, laptop, desktop will all continue to be redefined.

I repeat, I don't think Intel is going to suffer because Sandy Bridge and a few cycles after that they will have some stellar performance per watt for "mainstream" computing. But a swath of computing "under" that mainstream is going to be increasingly ARM-based in the next five years.\\

And therein lies Apple's genius: Guess where Apple wants to be in the next five years? In "mainstream standard boring business computing", or sweeping up everything from under that.
;-)

Ahh that is was he was getting at. That is still a completely false interpretation of the situation. Apple purposefully designed the iPad to only play H.264 using hardware acceleration. Those other players are essentially software work arounds to play non-supported codecs with variable degrees of success. But that is not because of any inherent problem with the A4. Its because Apple did not design the iPad to play those codecs using hardware acceloration.

Does DivX even have HWA? HandBrake doesn’t even support AVI/DivX anymore. MKV containers often seem to use HD content and therefore use the H.264 codec, but those apps can’t accelerate it as far a I know. Maybe in iOS 5.0, but I doubt it.

From what I've read it does not.

No, lets not go down this road --- when one week after the last argument, you talked about how Apple is going to do a process shrink to the A4 to put it in the next iphone. By your original argument, Apple should be releasing their own dual core compatible core by then. That's a huge difference between the two.

It's actually closer to 5 years because PA Semi was founded in 2003 and their CPU wasn't launched until Q4 2007. And that's just to ship to customers who would have to spend another year to design their end product around the new CPU. You add an extra year into the time to design the actual phone around the new CPU, that closer to 6 years.

Qualcomm spent the same amount of time developing the Snapdragon chips. It is the norm for the industry --- nothing special about PA Semi.

Re: iPad Playing other codecs...

I read somewhere that one advantage of playing other codecs is that a device can read/play camera data directly.

I just tested my 7-year-old Cannon PowerShot and the iPad will import, then Play Video.

Not so for my 5-year-old Sony DCR HC42.

Nor my Panny HDC-SD1... the Panny uses AVCHD and that decode pegs the Mac's CPU cores.

It would be nice, if the next-gen iPad were robust enough (CPUs and GPUs) to handle import and conversion of video from external cameras.

An 8 GB HDSC card from the Panny converts to about 57 GB.

iMovie on the Mac lets you preview individual clips and then select what you want to import/convert.

I am hoping that the iPad Gen 2 will support iMovie -- and allow preview/import/conversion/play from a broad range of external cameras (and not be limited to any internal cameras)

External cameras usually have 10-12x optical zoom and much better picture quality than a camera could be built into an iPad.

I did some experimenting with twitter and investigation of vBulletin.

These won't meet my needs:

1) Twitter, on AI, for instance allows me to follow all AI articles -- I want to follow only selected AI articles that Interest me. This is the information overload I want to avoid.

2) Also, when an article no-longer interests me, I want to stop following it, while continuing to follow others.

3) the vBulletin, AFAICT, requires the site to be running its software -- without it, there is no way to follow individual forum threads.


I have been thinking about It, and have an idea for an implementation similar to Flipboard -- where I could include:

1) sites to monitor and be notified when any new articles are posted the site

2) articles to monitor and be notified when any new posts are made to the forum thread.


I would use the app as follows:

3) Browse the app to see the sites I am monitoring, showing the articles I am monitoring (optionally showing all current articles -- including those not being monitored)

4) Browse the threads for the articles I am monitoring -- showing the last post I viewed, plus n prior posts, and all subsequent posts.


When new article is posted, or a new post is made to a thread -- the app would get a Push Notification and Badge.

A glance at the app's icon would indicate new activity.

The next time I use the app, I would have all the changes (new articles, new thread posts) at my fingertips (indicated by the badges) and could go directly to them.

I have some things to finish up, but I am going to play around with this in the next few weeks.

.

What you are talking about isn't directly related to the hardware in the iPad. Video codec incompatibilities are a whole mess that the camera manufacturers themselves create.

There have been efforts by the International Electronics Commission and International Telecommunications Union both to create standard interoperable video codecs that will work across all devices. In the professional and semi-professional world video manufacturers mostly follow the standard codecs.

In the consumer space they don't necessarily follow standard codecs at all. Sony, Panasonic, Canon, JVC all invent proprietary codecs that are mostly slight tweaks of standard codecs. For the purpose of locking the consumer into their products. These companies frequently create consumer codecs for products that fail in the market and abandon further support for the codec.

Apple makes much more of an effort to support the cacophony of consumer codecs in iMovie, but Final Cut Pro doesn't support them all.

When it comes down to playback on consumer devices, attempting to support all of these codecs is unnecessary and futile as old ones will be abandoned and there will always be new ones.

To simplify all of this Apple brings everything back to H.264. If your codec is a derivative of H.264, its likely to be playable on an Apple consumer device.

So the problem is not with Apple. Its with video camera manufactures who continuously create proprietary codecs to lock people into their system.

In the pro video world, camera manufacturers have begun to use Apple ProRes to create native QuickTime files. So that the video is ready for editing right out of the camera without conversion. That would be a nice trend industry wide.

That certainly would be interesting to see.

It makes me wonder if Apple will remain a cutting edge manufacturer right through the beginnings of the nano-fabrication age, and when might we actually see a completely solid-state device constructed atom-by-atom come from Apple?

Also, the dearth of information on whatever Apple is planning to do with their upcoming releases indicates to me that they have successfully fired or sussed out leakers and dealt with them accordingly. Where's the 'insider' info?

Making up your own non-existant interpretations rather than reading the words on the screen again I see. When you quit being such an intentional reader-with-his-own-twisted/added-contexter maybe we can believe some small part of what you write. Now for the correction, I opined that a nice tactic for Apple would be to shrink the A4 for an iPhone and release a dual core A4 follow on for the iPad to facilitate product differentiation and take advantage of additional power savings a shrunk die A4 would provide for the phone. Here's the text, I QFT myself!

I didn't say what Apple would do and I already addressed the supposed large missing piece of logic you tried to seize on. And even provided a nice motivation for why it might make sense. Basically you took a leap of ill will and forgot to strap on the parachute. oops.


Did you work for PA Semi? Were you party to EXACTLY the fact that they decided to start a company to do EXACTLY the single task you say it took them 4-5 years to accomplish? Unless you are a plankholder in PA Semi you don't know when they started work on any particular CPU project. By your logic they would have had to have the Motorola PPC licenses the day they opened shop, I find it pretty unlikely they did. It is far more likely they started their own projects a year or two AFTER they proved they had the chops to do that sort of work on contract. Thus the timeline would be more on the order of 3 years or so.


Again do you know exactly when what design goal decisions were made when? Do you really think Qualcomm decided one day "Hey, lets make a brand new CPU nobody ever saw before!" ? Then they decided to license ARM IP to do it? And then the day the ink was dry they began designing? If so you are ignoring the spool-up time in learning what goes where why. That can be a year or two before they could have the confidence to throw caution to the wind and start putting together something completely different.

Apple ASIC engineers have been playing with ARM-based hardware for decades. Apple has been extremely serious with ARM-based hardware for the iPhone line since at least 2003. Apple didn't start from a standing no-nothing start in 2007, so your ~gotta have at least 4 years to deliver an ARM related product~ don't really fit, and as many times as different folks have posted that, you have ignored it. When did the new chip development clock really start? When did Apple decide to strategically move from part consumer to part designer? How much clue did internal Apple ARM ASIC engineers already have and just needed I access to the IP and a few select PA Semi engineers to really get a project completed? We already know Apple has always looked at generating custom ASICS and does so with great regularity. And Apple has been helping manufacturers tweak CPU designs to meet Apple desires for several decades too. Apple wasn't designing a CPU from scratch anywhere in there, but they aren't strangers to the neighborhood either.

You ignore all of this and get absolutist in that if someone doesn't subscribe to your flawed timeline they are just plain wrong. Well as usual, absolutist makes for a fragile, brittle stance that just doesn't hold up to poking and prodding with the facts.

It is not a flawed timeline --- if EVERYBODY is taking that same amount of time, no matter how smart these engineers are.

However it is just plain wrong to expect Apple to HALF the time that the whole industry spent. From the time silicon companies start designing their own ARM cores to the time you get to see Motorola/Nokia actually shipping a phone with those ARM cpus --- takes about 6 years.

Can Apple do faster than the rest of the industry? Absolutely. But can Apple literally half the development time than the rest of the industry? Absolutely not. Nothing fragile or brittle with my stance.

You know... I am not qualified, in the slightest, to argue the technical aspects of this discussion.

But, as to the logical:

@Hiro refuted every one of your prior post's points!

Then you come back with "is not a flawed timeline --- if EVERYBODY is taking that same amount of time, no matter how smart these engineers are."

@Hiro illustrated that there is no way for you to prove that "EVERYBODY" ... "takes about 6 years".


Please provide links supporting your broad assertions.

Otherwise, respond in kind to the refutations provided.


From a logical perspective, you appear as a stubborn child blaring his own set of facts, while covering his eyes and ears -- oblivious to reality.

Further, this appears to be the MO you use in every thread you post... Is it working for you -- it doesn't seem to be very convincing to others.


It detracts from, rather than adds to the discussion!


Edit: Here's a quote from the CEO of ARM on 12/15/10.

If you read the article, he is talking about ARM hardware, estimated to be available in 2014 (4 years or less). The ARM hardware hasn't been announced yet.




“Work is under way: System designers are actively considering ARM architectures. We don’t want to raise expectations that next year there are going to be a lot of ARM servers. Of course, there aren’t," said Mr. East.

So, "designers are actively considering" in 2010, to use unannounced architecture, to be available within 4 years.


http://www.xbitlabs.com/news/cpu/dis...our_Years.html


Kinda knocks the wind out of your "EVERYBODY" ... "takes about 6 years".... doesn't it.


Who do you believe is better qualified to predict the timeline? You or the CEO of ARM?

.

No, he hasn't --- because he kept changing his arguments (while my arguments have been the same).

He was originally talking about Apple releasing their own dual core "compatible" core for ipad2/iphone5. I showed him how unreasonable his timeline was --- then he changed his argument by saying that Apple is going to process shrink the A4 and even dual core A4. You can't dual-core an A4 because Cortex A8 is designed to be single core only.

I don't know anything about PA Semi --- except that most of the top people left after Apple bought them, because Apple didn't price their stock options to their liking.

For the experts here:

According to what I've read the A4 is a relatively minor tweak of the Cortex A8 (with a GPU in the same package). With this approach, Apple was able to gain performance while minimizing risk and reducing time to market!

What would happen if Apple has taken the same approach to a dual-core Cortex A9 -- with a GPU capable of OpenCL?


NVIDIA announced the Tegra 2 A9 in Jan 2009, and was shipping product before Dec 2009.

I don't know when NVIDIA began work on the Tegra 2.


How long would it take Apple to do a A4-like tweak to the A9, for the A5?

When do you suppose Apple would have begun working on such a tweak?

When would you expect Apple could supply A5s in quantity to satisfy iPad demand?

.

It didn't take long for Intrinsity to make the Hummingbird for Samsung.

http://www.redorbit.com/news/technol...wer/index.html

It didnt occur to you that A4 could simply refer to Apples versioning of ARM CPUs as a whole because there is no other known term for Apples A4 marketing. For all you know Apple will keep A4 for ARMv7 chips they tweak.

It never occur to me because I speak in very precise terms and Hiro doesn't. Who knows what Hiro meant when he said things --- because he had zero idea what he was talking about. Then he had the balls to start throwing insults left and right saying that I knew nothing --- when he didn't even understand a simple term such as "architecture license".

OK the way I read the timeline:


XXX 200-: Samsung/Intrinsity begin work on Hummingbird


Sep 2008: Samsung/Intrinsity Announce Hummingbird (your link)

http://www.redorbit.com/news/technol...wer/index.html


Jul 2009: Hummingbird Silicon Available

http://www.samsung.com/global/busine...o?news_id=1030


Oct 2010: Products with Hummingbird ship

http://www.fudzilla.com/mobiles/item...-a-porn-button



When do you think that they, Samsung/Intrinsity, began working on this?


Do you think that Apple was working with Intrinsity at the same time?


Recall that Apple bought Intrinsity in Apr 2010 -- and was rumored to worked with Intrinsity on the A4.

http://digitaldaily.allthingsd.com/2...ys-intrinsity/


OK, given all that, why do you continue to assert:

"EVERYBODY" ... "takes about 6 years"?


By the dates you provided, Samsung/Intrinsity took a little over 2 years for the Hummingbird.

I suspect, in actuality, it took a little longer (even 3 years) if they began work in 2007.


But that's a far cry from 6 years.

.

6 years in what it takes for designing a COMPATIBLE core and shipping a cell phone with that compatible core.

Hummingbird and the A4 are using cortex A8 core --- not a compatible core.

What is a "COMPATIBLE core"?

What does that have to do with what Apple and Samsung are doing?

.

Qualcomm and Marvell sells chips that have cores that they designed themselves. They don't use the cortex cores. And it generally takes 4-5 years to design your own compatible core and add another year for actual phones to show up with those cores.

If you don't know what a compatible core means --- why are you arguing with me in the first place.

I've changed nothing. I also haven't said you could dual core an A4.

You continue to poorly read the phrase, "follow on to the A4". I have used those specific words several times because I don't know if there ever will be an A5, or if Apple will pick up some other marketing term for the technology it wields as a follow on. A dual core follow on would obviously be, to anyone not trying to twist things, a CortexA9 design, but it could even be a CortexA15 (but I would bet on the A9 due to power considerations in the mobile space). Maybe Apple will shrink the A4 and call it an A5? If they do it's still one of the possible follow ons, but not a dual core follow on.

See, I'm using very precise language and terminology to delineate what I think and not say anything else. Yes, I believe Apple is working hard to power optimize an ARM design like the CortexA8 and/or CortexA9, lots of folks here think so too. Nothing new and hardly scandalous except in your world where somehow your personal view of "The Timeline" is being violated. So far the only timeline thats unacceptable to the larger audience seems to be yours.

My view of the timeline is that I don't know when it started, but I'm sure it's before your timeline says it does. And that the timeline is shorter than you say it is -- for everyone -- because you only picked isolated PR release dates to use as data-in. I'll take those and raise them by adding the context of business process and engineering spool-up time before the "start" decision of any timeline gets made.

And when you don't twist other folks context, you still fall into the trap of not analyzing what you read before you retort. You make the same error of superficial analysis many stock analysts made upon the announcement of more PA Semi management leaving. To wit

You: "I don't know anything about PA Semi --- except that most of the top people left after Apple bought them, because Apple didn't price their stock options to their liking."

Me: "just needed access to the IP and a few select PA Semi engineers to really get a project completed"

Notice I said nothing about the Founders/Senior Managers who left, or even the lions share of the engineering pool. Not because I didn't know about them, but because they weren't the talent Apple was after! The departed founders are successful entrepreneurs and idea men. The founders were very successful as engineers in the day, but their work at PA Semi had moved on to building business, not tinkering with chips on a daily basis anymore. Apple didn't purchase the businessman, it has plenty of it's own personnel to do that. Apple purchased the team of engineers they put together. And I'm sure when it gets all the way to brass tacks Apple found the critical core in that team that was the real target.

That half dozen to dozen engineers ARE the technical prowess of PA Semi, and everyone else was just operational support and overhead. Just the way it is for the rest of the ~150 engineers.

So sure, ton's of PA Semi folks left, but the thing we won't know for sure is did the right ones stay? My guess is those few folks DID get treated right and did stay.

I checked all the posts on this thread.

You are the only one talking about Apple building a compatible core.

You raised an argument to a point that no one was making -- a strawman...

It is all but proven that Apple uses ARM architecture in the A4.

It is assumed that Apple will continue to use ARM architecture in follow-on custom cpus for iOS and iDevices.


... You are right though -- it is stupid of me to try to engage in a reasoned discussion with you.

If you mean a dual-core cortex a9-based A5, then say it that way --- because that's how I would say it because I am precise.

There is NOTHING spectacular about Intrinsity's tweak --- we ain't seeing the Hummingbird capturing massive amount of market share. We ain't seeing news reports saying that Hummingbird is giving a massive battery life edge over its competitors.

My arguments with Hiro went back several weeks over several different threads.

A custom processor for an embedded device better not be your selling point to own a market. If so, you're dead on arrival.

It's and end-to-end solution that wins the market.

Intrinsity was dead the day it started, until Apple bought it.

Maybe if you stopped making up terminology others would be able to follow you. There is nothing "compatible" in a Snapdragon or a Marvell core. Compatible coloquially boils down to "different but works with". The Qualcomm and Marvel cores adhere explicitly to the target ARM architecture, not something different that is only compatible.

Qualcomm integrated their baseband processor and the I/O connects to a CortexA9 core to maximize efficiency in CPU-baseband processing. Then they put it into a SoC. When they were done it was no longer a technically a CortexA9 core, but for all intents and purposes to an application programmer that's what it looks like. To a systems programmer it looks a bit different. Simply it is a ARMv7-A ISA processor, not a different ISA that is compatible.

I don't know enough about Marvell to comment farther than the basic terminology.

A lot of what the A4 was for cost reduction. If you are not making a flip phone, then you don't need a second LCD output in the SoC. If you aren't planning to make an iphone with a HDMI output, then you can take that out as well.

What Intrinsity brought to Apple:


They are just a tool that when combined with the IP from PA-Semi and Apple's massive own IP in PowerPC, ARM, X86, etc., not to mention ImgTec and Apple's own extensove GPU background, created A4.

Right, and even then you were the only one with your viewpoint. And it wasn't just me, I'm just the one who didn't give up in exasperation. I remember something about you likening yourself to Neo from the Matrix and that you would never give in to the Agent Smith's.

That was your last post in that thread, you never had an answer to my reply. Some Neo.

Qualcomm designed its own ARMv7A ISA compatible core name "scorpion". It is a core that has performance in between the A8 and A9 core. The scorpion core has some limited out-of-order execution --- which A8 doesn't have (but not to the extent as the Cortex A9 has). The scorpion core has deeper pipeline than the A8 --- which allows Qualcomm to make CPU with higher frequencies.

And what did that have to do with any of the posts in this thread? It sure didn't fit as an answer to mdriftmeyer.

Oh well, off to dinner.

That has nothing to do with ISA or "compatability". It is implementation. Try again.

I want intelligent arguments --- and I don't mind me arguing against a dozen people who understand what the arguments are all about.

No it fits with the argument because all people talk about "custom" being faster and better features. But the A4 is mainly a cost reduction customization.


Why don't you try again? You don't even know what Qualcomm uses in their CPU.

On many of the iOS devices it is a hardware decoder.

We could debate that for a long time but the fact remains that VLC decoders do work well on other hardware. Besides that supporting old, or marginally used codecs should be one with solid stable code that minimizes support needs. Highly optimizing is not the smart thing if it takes development time away from codecs that get used in volume.

Are you really sure they can't do that now? but lets say the A4 can't do 1080P, doesn't that just support what I'm saying. You seem to be running in circles here.

I'm not sure why people get so tangled in their underwear when I state the obvious about the A4's CPU. The A4 falls flat on its face when ever an app becomes CPU bound just like the Intel processor on the new AIRs. That doesn't make the AIRs or the iPads bad machines it is just a statement of a characteristic of the machine.

Now if somebody wants to continue to argue against this point you are really going to have to come back with a lot more than just buts and ifs. I know all about the fast GP??U's and the flash memory making things appear fast but it isn't the type of fast usable buy all Mac users nor iPad users.

Nope, no hardware acceleration for Divx/Xvid on iOS devices. But like I said, you can still play these files smoothly on iPad and iPhone4 even though it's using standard CPU. I haven't tested 720p files yet, will do soon.

That's pretty interesting that Handbrake dropped Xvid, it's definitely an old codec now that is losing favour to H.264. Although Xvid is still de facto for pirated releases.

Ive tried HD files. Neither VLC nor CineXPlayer will play them. I havent tried in a few app updates so maybe its changed.


They did it awhile ago, and yet the commonality of free encoders for x264, the better compression of x264 with more options for containers, that x264 is used for HD in MKV containers, and that PCs to tablets to smartphones to PMPs all have HW decoders for H.264 its still the most common for pirated SD video, as you say.