Apple iPad announced

[Ailuros]

Not in all performance aspects of course. I was mainly referring to this post from warmi regarding fillrate:

Under normal conditions all performance characteristics scale with higher frequencies on a chip. The SGX535 in the iPad is obviously clocked higher than the one in the 3GS and thus will inevitably also have a higher fill-rate amongst others.

So in regards to fillrate running at native resolution: iPhone = iPhone 3G = iPad = iPhone 4th-gen
Totally simplified of course (and assuming the iPhone 4th-gen uses an (slightly slower) A4 and has a 960x640 resolution).iPhone3GS = 153600 pixels
iPad = 786432 pixels
(your) iPhone"next" = 614400 pixels

Under that reasoning and assuming Apple will use the 535 again, they could clock it lower than in the iPad. I still don't see any "anomaly" or better don't understand what you really mean. I said a long time ago that IMHO the iPhone3GS has too much fill-rate. If you look at all other smartphones of the 3GS class containing SGX you'll see that the widest majority contains a SGX530 and not a 535.

And as most know already:

SGX530 = 1 TMU
SGX535 = 2 TMUs
SGX54x = 2 TMUs

In other words going above a 53x for Apple wouldn't had gotten them more texel fill-rate per MHz and in my mind confidentially the 535 in the 3GS has one TMU too much. One license though for multiple devices is always cheaper though ;)

In any case if you meant something along that line I totally agree. Other than that I don't see anything weird.

I think that the 65nm SGX535 in the iPhone 3GS is clocked at 200Mhz.

Which could mean that the frequency for the 535 in the iPad is quite a bit higher than 250MHz, which I don't think so.

iPhone3GS = 153600 pixels to fill = 105185 kTexels/s multi-textured (SGX535 = 2 pixels/clock)
iPad = 786432 pixels to fill = 236559 kTexels/s multi-textured (SGX535 = 2 pixels/clock)

iPhone3G = 153600 pixels to fill = 28508 kTexels/s multi-textured (MBX Lite = 1 pixel/2 clocks)

My personal estimate for the above is in that order 150, 250, 100MHz.

[warmi]

I said a long time ago that IMHO the iPhone3GS has too much fill-rate. If you look at all other smartphones of the 3GS class containing SGX you'll see that the widest majority contains a SGX530 and not a 535.


That's irrelevant... Considering a typical game developer who got used to being able to push the envelope in terms of fill-rate on a 3gs device, it will take significant adjustments in terms of managing visual quality to get the same application working on the iPad.

[Ailuros]

That's irrelevant... Considering a typical game developer who got used to being able to push the envelope in terms of fill-rate on a 3gs device, it will take significant adjustments in terms of managing visual quality to get the same application working on the iPad.

Smartphones in their majority don't have even remotely close usable fill-rate than a 3GS. I'm obviously no developer but I'd expect a developer not to ignore the iPhone3G/2G, Touch devices as possible lowest common denominators for game development and scale up from there. Eventually such a mobile game could work also on all other existing MBX Lite/MBX smartphones and have a quite wider userbase. Just developing with only one smartphone in mind sounds damn limited to me.

[rpg.314]

Eventually such a mobile game could work also on all other existing MBX Lite/MBX smartphones and have a quite wider userbase. Just developing with only one smartphone in mind sounds damn limited to me.

SJ: iForbid

:wink:

[warmi]

Just developing with only one smartphone in mind sounds damn limited to me.

Normally , I wouldn't disagree but .. this is Apple/iPhone/AppStore combo we are talking about here and there are a lot of iPhone-only games out there ...

[Entropy]

Let me rephrase/correct/edit. In the markets Apple has decided to compete, it has so far thrashed all others put together. In the >$1000 desktop/laptop market too, I seem to recall that Apple has >~50% revenue share. On the basis of track record, I doubt Apple's competition is any good. The only solace is that Apple will prolly not compete in the low margin category, leaving a window for the competition there.

NPD has pegged Apples market share in the >$1000 desktop/laptop market at 90%(+) for some time. It went from 88% to 91% going from May 2009 to June 2009, and as far as I've seen has remained at those levels.

It also gives some perspective on how tiny the market for "premium" Windows PCs really is compared to the whole.

[eastmen]

What's a "mobile Xbox360" exactly or better what is there integrated in a multitude of current devices that isn't a mobile XBox360 already?


a mobile 360 would be a deviced dedicated to gaming that has hardware buttons to play more complex games on it.


Think of the ds or psp. IT would be a modern version of those

[silent_guy]

I think for the iPhone internals to be that small Apple must surely use a SoC-based approach.
That's really not saying anything: all cell phone chips are SOCs... But if you mean specifically that the baseband is integrated into the main die: I doubt it.
There are many reasons not to do it: you're wedded to the schede of the base band, it's highly specialized technology that can't easily be bought (both in terms of IP and engineers), it requires a relatively low speed interface so you don't have a lot of integration benefits anyway etc.
I haven't picked up anything about Apple hiring dsp and telecom engineers (allround designers is a whole different story, they have been hiring like crazy for a year now.)

Maybe they integrated it into the same package, but that's also unlikely since it would hamper their ability to integrate memory, which has much higher benefits to reduce PCB complexity.

IMHO it will feature the exact same chip as the A4. Spinning smaller cost derivatives is rarely cost efficient.

[Mike11]

I'm obviously no developer but I'd expect a developer not to ignore the iPhone3G/2G, Touch devices as possible lowest common denominators for game development and scale up from there.
This forward(?) compatibility issue is why I really really hope that Apple won't continue to offer two hardware generations at the same time after this summer (like right now with the iPhone 3G and 3GS). That just makes it worse. Just offer a $99 8GB version (besides the 32GB and 64GB) of the current high-end iPhone and accept a little lower margins for that entry model. IMHO that would benefit Apple in the long-run. Same goes for any future iPad generations. And same goes for iPod touch OS updates. Make them free so that customers are more likely to update (right now a lot of iPod touch user don't).

[MfA]

It also gives some perspective on how tiny the market for "premium" Windows PCs really is compared to the whole.
With a PC there is no point to spend that much money for most people. Mac Pro's configuration is into a range of hugely diminishing returns as far as component costs go ... it's just that the middle ground does not exist, they are the only way to get decent hardware which will (officially) run OS X.

If they sold i7+5870 macs in an expandable case at reasonable margins the market for Mac Pros would shrink drastically.

[wco81]

This forward(?) compatibility issue is why I really really hope that Apple won't continue to offer two hardware generations at the same time after this summer (like right now with the iPhone 3G and 3GS). That just makes it worse. Just offer a $99 8GB version (besides the 32GB and 64GB) of the current high-end iPhone and accept a little lower margins for that entry model. IMHO that would benefit Apple in the long-run. Same goes for any future iPad generations. And same goes for iPod touch OS updates. Make them free so that customers are more likely to update (right now a lot of iPod touch user don't).

Do other manufacturers sell different generations of phones? A big company like Nokia probably has a lot of previous gen inventory in the channel.

And Android supports phones of a wide range of hardware specs.

So it may be a competitive advantage to have as homogenous platform as possible. But the installed base is going to be heterogenous as people hang on to their phones for a couple of years. Developers will still have to target older SKUs or support them.

[Pressure]

That's really not saying anything: all cell phone chips are SOCs... But if you mean specifically that the baseband is integrated into the main die: I doubt it.
There are many reasons not to do it: you're wedded to the schede of the base band, it's highly specialized technology that can't easily be bought (both in terms of IP and engineers), it requires a relatively low speed interface so you don't have a lot of integration benefits anyway etc.
I haven't picked up anything about Apple hiring dsp and telecom engineers (allround designers is a whole different story, they have been hiring like crazy for a year now.)

Maybe they integrated it into the same package, but that's also unlikely since it would hamper their ability to integrate memory, which has much higher benefits to reduce PCB complexity.

IMHO it will feature the exact same chip as the A4. Spinning smaller cost derivatives is rarely cost efficient.

Yeah, I can see how it was pretty vague formulated. However, there is no denying that things have to have changed when going from something that uses this amount of space.

http://s1.guide-images.ifixit.com/igi/hsa3HpY6I2UNWjYJ.large

To this

http://cache.gawkerassets.com/assets/images/4/2010/04/open19.jpg

[rpg.314]

That's really not saying anything: all cell phone chips are SOCs... But if you mean specifically that the baseband is integrated into the main die: I doubt it.
There are many reasons not to do it: you're wedded to the schede of the base band, it's highly specialized technology that can't easily be bought (both in terms of IP and engineers), it requires a relatively low speed interface so you don't have a lot of integration benefits anyway etc.
I haven't picked up anything about Apple hiring dsp and telecom engineers (allround designers is a whole different story, they have been hiring like crazy for a year now.)

Maybe they integrated it into the same package, but that's also unlikely since it would hamper their ability to integrate memory, which has much higher benefits to reduce PCB complexity.

IMHO it will feature the exact same chip as the A4. Spinning smaller cost derivatives is rarely cost efficient.

AFAIK, the baseband is not on the same die so that the same SoC can be used in different markets with different carriers/frequencies etc. unlike Snapdragon.

[rpg.314]

Yeah, I can see how it was pretty vague formulated. However, there is no denying that things have to have changed when going from something that uses this amount of space.

http://s1.guide-images.ifixit.com/igi/hsa3HpY6I2UNWjYJ.large

To this

http://cache.gawkerassets.com/assets/images/4/2010/04/open19.jpg

Whoa, hadn't seen those before. What's the second one from again?

[Mike11]

Do other manufacturers sell different generations of phones? A big company like Nokia probably has a lot of previous gen inventory in the channel.

And Android supports phones of a wide range of hardware specs.
Yes, they do.

So it may be a competitive advantage to have as homogenous platform as possible. But the installed base is going to be heterogenous as people hang on to their phones for a couple of years. Developers will still have to target older SKUs or support them.
Yes, IMHO it would be a big competitive advantage to have an as homogeneous platform as possible. That means moving customers to newer, more powerful/capable SoCs as soon as possible/available, to make sure that when they upgrade, they upgrade to the new version, not the old one, extending the lifetime of that old model another 2-3 years (contract duration + buffer).
E.g. someone who bought the original iPhone in late 2007 renews his contract in late 2009 and chooses an iPhone 3G. That means he uses the 2007 SoC until 2011 instead of just until 2009, if the entry level iPhone in 2009 would have been an iPhone 3GS SKU.

Apple is in a unique position to do so because it's the sole supplier of iPhones and only brings out one new iPhone per year and is on a yearly release cycle. So why extend the lifetime of last years version by keep offering it even (for another year) after the successor is released. And I'm not talking about 1-2 months to clear inventory. But yes, it's probably cheaper and increases Apple's profits short term.

Same goes for future iPad generations.

Whoa, hadn't seen those before. What's the second one from again?
Leaked iPhone 4th-gen prototype.

For size comparison, the first one has a (Mini-)SIM slot, the second one has a Micro-SIM slot (in the middle). As a reference point you can use the six SIM contacts that should have the same size on each board (since the SIM chip itself is identical).

I linked the image of the new board some posts above:
Assuming the iPhone 4th-gen also uses an A4, how much more empty (or totally battery dominated) would the iPad have looked on the inside if had used the 4th-gens logic board? That board is really super small...

http://cache-01.gawkerassets.com/assets/images/4/2010/04/open19_01.jpg
Gallery: http://gizmodo.com/5520877/open-gal//gallery/1

Article about the internals:
http://gizmodo.com/5520876/the-next-iphone-dissected

And the original article about the leak is here:
http://gizmodo.com/5520164/this-is-apples-next-iphone

[silent_guy]

http://cache.gawkerassets.com/assets/images/4/2010/04/open19.jpg
Yeah, saw that too.

Given that gizmodo had already ripped the thing apart, I had hoped that they'd gone all the way and tear this down too.

But we're left to speculate.

With the now proven high volume of the iPhone and iPad, there'll probably find very willing suppliers to bend to their needs.

Potential improvements, based on the annotated version here (http://www.phonewreck.com/wp-content/uploads/2009/06/Logic11.jpg).
- integrated GPS/WIFI/bluetooth chip. I believe a couple of companies have been working on this.
- baseband processor and baseband SRAM chips can be merged into a POP package
- PMIC and BB-PMIC can also be merged.

With just these 3 optimizations, you already go a long way. Still, it's a very impressive piece of engineering. Can't wait for the teardown of this.

[wco81]

There's a rumor that Apple is looking to buy ARM for $8 billion.

Is there any other competitive mobile CPU?

[Ailuros]

Normally , I wouldn't disagree but .. this is Apple/iPhone/AppStore combo we are talking about here and there are a lot of iPhone-only games out there ...

Are those iPhone-only games (spreading over 2G/3G/3GS) or iPhone3GS-only games (honest question as all the others)? Because if it's the first, there's quite a difference between 50+MTexels (MBX Lite@100+MHz) and 300MTexels/s (SGX535@150MHz) for the exact same target resolution.

a mobile 360 would be a deviced dedicated to gaming that has hardware buttons to play more complex games on it.
Think of the ds or psp. IT would be a modern version of those

I'm not sure but Microsoft's future plans sound to me more like a possible tablet device than a hand-held console.

This forward(?) compatibility issue is why I really really hope that Apple won't continue to offer two hardware generations at the same time after this summer (like right now with the iPhone 3G and 3GS). That just makes it worse. Just offer a $99 8GB version (besides the 32GB and 64GB) of the current high-end iPhone and accept a little lower margins for that entry model. IMHO that would benefit Apple in the long-run. Same goes for any future iPad generations. And same goes for iPod touch OS updates. Make them free so that customers are more likely to update (right now a lot of iPod touch user don't).

For the time being Apple sells the 3G starting from $99 and the 3GS starting from $199. In other words the 3G hasn't phased out yet and it still poses the lowest possible iPhone denominator for Apple's smart-phones. Considering how long the 3G has been available and its price today, it wouldn't be out of line to suggest that the amount of 3G devices sold or better in user hands today is quite a bit larger than 3GS owners.

Yes, IMHO it would be a big competitive advantage to have an as homogeneous platform as possible. That means moving customers to newer, more powerful/capable SoCs as soon as possible/available, to make sure that when they upgrade, they upgrade to the new version, not the old one, extending the lifetime of that old model another 2-3 years (contract duration + buffer).While there's no doubt that SGX is a far more capable and efficient GPU than MBX Lite, how many of the first's advanced capabilities are really used in today's iPhone games after all? Yes there's a sizable performance difference between those two, but why should the "rules" for game development in the embedded market be any different than in the PC space for example especially considering how humble in general game content still is?

Besides what's "homogenous" platform anyway considering that Apple will inevitably move to another GPU IP eventually (if it's not this year it'll be the next year) since they can't really forever use for all that long a SGX535 in the end? There's always going to be a more advanced component (be it CPU or GPU or whatever else) in future SoCs. Of course as always only Apple itself knows about its future plans and we mostly find out most of the aspects shortly before or after each devices release.

However on a purely speculative basis IMG has announced that SGX543 MP has been licensed by 3 of their lead partners so far. IF Apple is one of them and they'll release it next year if I'd only assume a 2MP at the very same frequency compared to a 535 you'll end up with over 8x times the FLOPs, twice the texel fill-rate and about 4x times triangle rate and z/stencil fill-rate. It doesn't have to be IMG IP, but there are always going to be upgrades where disparities will appear between current and former generation devices.

Only Apple knows what it wants or feels that it needs for the future. But just like the used the MBX Lite for iPhone2G/3G, iPod Touch, they'll also re-use the SGX535 for a higher amount of devices. One license as I said is always cheaper than having more than one; any IHV/OEM won't just keep it forever.

[rpg.314]

There's a rumor that Apple is looking to buy ARM for $8 billion.I am EXTREMELY doubtful if it will happen.

Also, I don't think other CPU's are as perf/W efficient. I suppose x86 is the only alternative game in town if this happens, because of Intel's muscle.

[Ailuros]

I am EXTREMELY doubtful if it will happen.

Also, I don't think other CPU's are as perf/W efficient. I suppose x86 is the only alternative game in town if this happens, because of Intel's muscle.

While I doubt it too, it would be for sure a very interesting twist in the entire market picture if true. Albeit I've learned through the years never to say never again, I don't think Apple would want to take such a huge risk. What for anyway?

[rpg.314]

While I doubt it too, it would be for sure a very interesting twist in the entire market picture if true. Albeit I've learned through the years never to say never again, I don't think Apple would want to take such a huge risk. What for anyway?

Forget risk, I don't think Apple will be able to get it past EU competition commission. :razz:

And as you correctly point out, what is in it for them? They are hardly the pushing the CPU envelope. However, if they do, it'll be interesting to see of they stick with Mali or try to buy IMG as well. Intel may not care about ARM buyout, but they'll fight the IMG deal tooth and nail.

[Ailuros]

Forget risk, I don't think Apple will be able to get it past EU competition commission. :razz:

ROFL

And as you correctly point out, what is in it for them? They are hardly the pushing the CPU envelope. However, if they do, it'll be interesting to see of they stick with Mali or try to buy IMG as well. Intel may not care about ARM buyout, but they'll fight the IMG deal tooth and nail.Apple for sure is rumored to buy a whole damn LOT lately: http://www.macnewsworld.com/story/69816.html?wlc=1271843551&wlc=1271844581

Actually why doesn't Apple want Qualcolmm instead of all that, or does Apple have some kind of fetish with companies starting with an "A"? I'm sure some gentlemen in Finland would be delighted to change the office signs for the 4th time in a row :lol:

[Arun]

Oh well, I might as well stop lurking for one post so that this profoundly absurd discussion stops going entirely in the wrong direction :p

Here are, nearly certainly, the major changes:
- Infineon 65nm X-Gold 616 Baseband (integrated baseband PMU) with 130nm Infineon UE RF (many fewer external components) linked over DigRF 3G.
- No Baseband SRAM, uses the very rare Shared-DRAM interface (see X-Gold 616 datasheet) that allows the Baseband to share memory even standard LPDDR1. Complicates memory controller scheduling on the app processor quite a bit, but obviously saves space/cost.
- Infineon 65nm XPOSYS GPS chip instead of Infineon 130nm Hammerhead I/II (thse are actually the same silicon). Very small chance they went with the CSR SiRFStarIV instead.
- The Bluetooth/WiFi module is still from Taiwan-based USI but it's a new version, not a rectangle like the iPad. Presumably still using the BCM4329 though, but impossible to be certain.
- Unlike the iPad where the screen was too big for it, they are nearly certainly using the more space-efficient touchscreen solution from the 3GS.

The A4 is on 45nm, unlike the 3GS's SoC but uses the same CPU and GPU. That means it's probably *cheaper* (although there's the extra cost of the 64-bit memory bus) and there's absolutely no reason to speculate on what other processor Apple might be using. And as seems very clear here, footprint of modern phones even with discrete app processors can be very very small.

As silent_guy said, there's not much benefit from integration there. It still makes a lot of sense for Qualcomm or ST-Ericsson though, in which case it should be seen as a partial lock-in strategy: "if you want our application processor, you need to use our baseband", and pricing-wise it's probably fair to see it as mostly a mere bundling advantage going forward (there is a cost benefit, but it's not big).

More interesting, of course, is what Apple plans to do with the VXD375 in the A4 - presumably they couldn't do HD video on the 3GS because they lacked memory bandwidth but that's solved now. I do suspect a bit that VXD375 might be a 720p-only variant of the VXD370, in which case that's what we'll get like on the iPad. We'll see.

[Pressure]

Whoa, hadn't seen those before. What's the second one from again?

It's the iPhone 3GS compared to the prototype iPhone 4th generation bought by Gizmodo.

The difference is pretty wild.

Update: Totally ninja'd on that one heh.

[MfA]

There's a rumor that Apple is looking to buy ARM for $8 billion.
That seems like a dumb idea to me ... the competition would simply switch to a different architecture over time, making it neither a competitive advantage nor a good investment.

They'd do better buying IMG IMO.

[rpg.314]

They'd do better buying IMG IMO.

What will they gain by buying even IMG?

[Florin]

What will they gain by buying even IMG?

This one's easy: Rys :D

[rpg.314]

This one's easy: Rys :D

Why buy an entire company when you can just poach an employee? :lol:

[eastmen]

What will they gain by buying even IMG?

exclusive use of the chipset which seems very powerfull compared to what others are able to offer ?


Wonder what IMG can do linked up to the billions that Apple has

[MfA]

I think IMG's architecture and patent portfolio is more valuable as a differentiator than ARM. There are more viable alternatives to ARM than there are to IMG.

ARM's value lies in being the architecture which the market has chose to standardize on, Apple buying it destroys that value ... and merely reduces it to being a decent architecture to build their own processors on. One among many. For that 8 billion is a lot of money.

[Mike11]

Besides what's "homogenous" platform anyway considering that Apple will inevitably move to another GPU IP eventually (if it's not this year it'll be the next year) since they can't really forever use for all that long a SGX535 in the end? There's always going to be a more advanced component (be it CPU or GPU or whatever else) in future SoCs. Of course as always only Apple itself knows about its future plans and we mostly find out most of the aspects shortly before or after each devices release.
I didn't mean "homogeneous platform" as in "only one hardware platform", I meant as in "as few as possible". Yes, there's always gonna be something newer and better. And yes, app developers probably have to support a hardware generation up to 3 years after it stops selling because of the install base. My point was that IMHO Apple shouldn't grow the install base of the old hardware further after the new generation is released.

More interesting, of course, is what Apple plans to do with the VXD375 in the A4 - presumably they couldn't do HD video on the 3GS because they lacked memory bandwidth but that's solved now. I do suspect a bit that VXD375 might be a 720p-only variant of the VXD370, in which case that's what we'll get like on the iPad. We'll see.
Since there's a good chance that we'll see 720p video recording and video chat in the 4th-gen iPhone, I wonder why Apple would not use VXE as well. Are the benefits not worth the extra die-space?

[Ailuros]

I didn't mean "homogeneous platform" as in "only one hardware platform", I meant as in "as few as possible". Yes, there's always gonna be something newer and better. And yes, app developers probably have to support a hardware generation up to 3 years after it stops selling because of the install base. My point was that IMHO Apple shouldn't grow the install base of the old hardware further after the new generation is released.

IMHO Apple should have picked a MBX and not MBX Lite in the first place for their first generation devices in order to have smaller fill-rate gap. Beyond that and with the introduction of programmable hw things will be a lot easier IMO. Differences between DX7 and DX9 for example are lot larger than DX9 to DX10 or even 11.

Since there's a good chance that we'll see 720p video recording and video chat in the 4th-gen iPhone, I wonder why Apple would not use VXE as well. Are the benefits not worth the extra die-space?No idea if they'll eventually get there, but dedicated ff hw is the best solution for such cases. However video recording on iPhones I've seen to date is anything but bad. I've seen other solutions dropping frames while recording like there's no tomorrow.

exclusive use of the chipset which seems very powerfull compared to what others are able to offer ?

Exclusive use sounds like a crappy investment to me. Granted IP companies are typically low risk/low revenue, but exclusivity would mean cutting off all other partners IMG has to date. At the moment they claim integration in 250M units for their IP and their next goal is half a billion units.

If IMG would be available for sale (which I still don't think it is) it would be anything but cheap. To acquire such a company at that price for just Apple devices sounds weird to me. Besides considering that Intel has a healthy stake in the company it doesn't sound that easy.

Wonder what IMG can do linked up to the billions that Apple has

Depends what hypothetically Apple would use them for.

[Mike11]

If IMG would be available for sale (which I still don't think it is) it would be anything but cheap. To acquire such a company at that price for just Apple devices sounds weird to me. Besides considering that Intel has a healthy stake in the company it doesn't sound that easy.
It wouldn't be PA Semi cheap ($278m), but also not ARM expensive ($8b). Market cap is £667m right now, and Apple already owns nearly 10% of it. Apple certainly could afford it without problems (Apple has $10b in cash and $13b in short term investments). And I think Apple could come to an agreement with Intel, since IMHO Intel isn't in it for the long run (they ultimately wanna use their own IP in their MID/smartphone SoCs, but just aren't there yet).

But I think it would only make sense if Apple planned to rely heavily on GPGPU/OpenCL in future iPhoneOS devices and was convinced that IMG was already way ahead of everyone else in the game and had a very strong roadmap.

[Ailuros]

It wouldn't be PA Semi cheap ($278m), but also not ARM expensive ($8b). Market cap is £667m right now, and Apple already owns nearly 10% of it. Apple certainly could afford it without problems (Apple has $10b in cash and $13b in short term investments). And I think Apple could come to an agreement with Intel, since IMHO Intel isn't in it for the long run (they ultimately wanna use their own IP in their MID/smartphone SoCs, but just aren't there yet).

But I think it would only make sense if Apple planned to rely heavily on GPGPU/OpenCL in future iPhoneOS devices and was convinced that IMG was already way ahead of everyone else in the game and had a very strong roadmap.

Bears still the question if IMG is available for sale. Besides Imagination Technologies doesn't start with an "A" :twisted:

Anyway for those interested IMG just added a bunch of videos to the PowerVR Insider section of their demo room: http://www.imgtec.com/demo_room/index.asp
Especially the UI demos are quite nice.

[Simon F]

Why buy an entire company when you can just poach an employee? :lol:

You'd need a big saucepan for Rys :razz:

[Mike11]

Bears still the question if IMG is available for sale.
I don't think they are either. This was more of a "if Apple absolutely wants to spend a few billion and buy a (fabless) chipmaker, it should buy... because of ..." thought experiment. And Apple already uses IMG IP in their iPhone OS devices and owns 10% of the company, so it's not totally far-fetched ;-)

Besides Imagination Technologies doesn't start with an "A" :twisted:
Well, maybe Apple got soft on that one. The name just has to have an "A" somewhere in it. Like PA Semi, LAla, QuAttro Wireless :cool:

[Ailuros]

Well, maybe Apple got soft on that one. The name just has to have an "A" somewhere in it. Like PA Semi, LAla, QuAttro Wireless :cool:

Or IMG intends to rename itself to Amagination Technologies :razz:

You'd need a big saucepan for Rys :razz:

Big guys can carry a LOT of tea pots ya know ;)

[Rys]

I'm going to buy IMG and rename it Rysmagination Techlolorys. RYS.L symbol on the LSE, the works.

[Ailuros]

I'm going to buy IMG and rename it Rysmagination Techlolorys. RYS.L symbol on the LSE, the works.

Do you accept investments then? I'd like to finance a high end GPU project :lol:

On a more serious note:

The head of ARM denied any talks regarding possible takeover by Apple, claiming thatit would be much cheaper to license ARM’s technologies than to acquire the whole company.

“Exciting though it is to have the share price pushed up by these rumours, common sense tells us that our standard business model is an excellent way for technology companies to gain access to our technology. Nobody has to buy the company,” said Warren East, chief executive officer of ARM, in a brief interview with the Guardian (http://www.guardian.co.uk/business/marketforceslive/2010/apr/22/armholdings-marketforces).

http://www.xbitlabs.com/news/cpu/display/20100422141657_Apple_Rumoured_to_Acquire_Microproc essor_Developer_ARM.html

http://www.guardian.co.uk/business/marketforceslive/2010/apr/22/armholdings-marketforces

Well I guess ARM can send the analyst who started the story a couple of beer cases ;)

[MfA]

I'm sure he made quite enough off that rumour himself already.

[Kurt]

Dude, where is my pinky?


http://www.youtube.com/watch?v=O005Hr-xQfc&feature=related


:shock:

[argor]

http://www.engadget.com/2010/04/27/apple-purchases-intrinsity-just-498-more-arm-licensees-to-go/
Apple purchases Intrinsity

[Simon F]

A colleague has just brought one in to the office. I'm not in the least bit covetous.

[rpg.314]

Just green with envy, eh? ;)

[darkblu]

Simon going NV, confirmed.

[eastmen]

Any reprts of problems with these. My friend got a DOA and while at the apple store with him yesterday someone had a unit that stoped charging.

Thankfully for them there were no refurbs avalible so they got new units from apple care.

Just a heads up for anyonethat owns them , make sure the recharge time isn't increasing each time you charge it. This si what the girl noticed before not being able to charge at all.

[Grall]

I'm going to buy IMG and rename it Rysmagination Techlolorys.
If you did that, your company would have the most advanced techlororys in the business! :)

[Mike11]

Sounds like Samsung is gonna start 32LP volume production in Q3/2010. Very impressive. AFAIK not quite early enough for the next iPad (in spring I assume), but theoretically possible for the 2011 iPhone. Apple would be pushing it, though.
"we have customers taping out on 32nm in Q3 on our high k, low power process for mobile applications. Risk production on 32nm will start in a few months."
http://www.electronicsweekly.com/Articles/2010/04/29/48537/samsung-taping-out-32nm-foundry-designs.htm

By the way, this is one of the times I wish I had an microprocessor report subscription:
Why Apple Wants Intrinsity
MPR believes the Intrinsity acquisition buys four things, well worth the money: Fast14 technology; exclusive access—or, at least, early access—to Intrinsity’s processor cores; a skilled processor-design team with expertise in balancing low power and high performance; and a multigigahertz Fast14 implementation of ARM’s Cortex-A9 dual-core processor.

That last item, in development for more than a year, could figure prominently in Apple’s plans for future iPhones, iPads, and other consumer gadgets.
http://www.mdronline.com/watch/watch_Issue.asp?Volname=Issue+%23042610&on=1#item2

[Lazy8s]

Apple has both a supposed ARM architectural license and Intrinsity's core tweaking skills now.

[darkblu]

I wonder what was the deal about pa semi, given how short-lived the employment was for some of those engineers.

[Rob Evans]

I wonder what was the deal about pa semi, given how short-lived the employment was for some of those engineers.

PA Semi was founded by Dan Dobberpuhl who was the lead designer at DEC who took an ARM architectural license and turned it into the StrongARM chip. I guess Apple thought he'd be just the chap to work some magic on their ARM architectural license, but they seem to have problems holding onto chip designers.

[silent_guy]

I wouldn't make to big of a deal of it. If you want to grow your chip design team quickly, buying an existing company is the way to go. It helps that PA semi had a lot of low power experience, but it's not essential to keep the lead designers. Designing for low power is now practiced by everyone, the knowledge has spread around. The real question is not how many of the principals left but how many of the rank and file has stayed. Given the not so great state of the economy, I'd say: the vast majority of them.

[wco81]

Maybe the Intrinsity guys will be able to use the PA Semi IP.

[silent_guy]

Maybe the Intrinsity guys will be able to use the PA Semi IP.
I don't even think there's that much reusable ip.

The IP around the CPU complex is not very large of these kind of chips. And PA didn't have anything in the field of audio, video, 3d, image processing etc. Apple has been hiring a lot of people with experience in those fields over the last two years.

[Florin]

TechCrunch murmurs about the possible premature death (http://techcrunch.com/2010/04/29/hewlett-packard-to-kill-windows-7-tablet-project/) of HP's Windows 7-based Slate project.

The iPad competitor was revealed by Steve Ballmer at this year's CES to much fanfare, but rumours claim HP is now unhappy about the touch experience in Windows 7 as well as the battery life of the Atom solution.

[Arwin]

TechCrunch murmurs about the possible premature death (http://techcrunch.com/2010/04/29/hewlett-packard-to-kill-windows-7-tablet-project/) of HP's Windows 7-based Slate project.

The iPad competitor was revealed by Steve Ballmer at this year's CES to much fanfare, but rumours claim HP is now unhappy about the touch experience in Windows 7 as well as the battery life of the Atom solution.

It is interesting to see how many companies as well as 'people on the internet' are discovering the value of what Apple did with the iPhone OS by trial and error. ;)

I think this may well be tied to HP's purchase of Palm's OS technology. They now understand the value of it and I think we'll see a quick shift to Web OS.

[Florin]

I think this may well be tied to HP's purchase of Palm's OS technology. They now understand the value of it and I think we'll see a quick shift to Web OS.

It might well be tied to the purchase but so far webOS hasn't really managed to gain a lot of traction. Rumour has it that the main idea behind the Palm purchase was an inexpensive way to grab a nice patent portfolio.

Generally, I think HP might be wiser to ride the Android mindshare.

[wco81]

Well $1.2 billion is a lot for just a patent portfolio. Actually, Nokia and RIMM might have done better with the webOS than trying to catch up with their homegrown development.

And the value of iPhone OS now includes the app. ecosystem, something which webOS is far behind.

At their last quarterly report a few weeks ago, Palm said it had like 9 months inventory in the channel. They've been dumping that product, which by now has an uncompetitive SOC. HP might have to eat some inventory and move up to the fastest SOCs as soon as they can, otherwise the perception of webOS isn't going to improve.

Regarding the patent portfolio, Apple supposedly has patents on many of the multitouch gestures copied by Android and webOS. But yeah, those older Palm patents may have insulated them from litigation.

[Simon F]

as well as the battery life of the Atom solution.
Which Atom system does it use? That could make a lot difference to the power consumption.

[green.pixel]

TechCrunch murmurs about the possible premature death (http://techcrunch.com/2010/04/29/hewlett-packard-to-kill-windows-7-tablet-project/) of HP's Windows 7-based Slate project.

The iPad competitor was revealed by Steve Ballmer at this year's CES to much fanfare, but rumours claim HP is now unhappy about the touch experience in Windows 7 as well as the battery life of the Atom solution.

It surprises me that they didn't realise before in the design phase that putting a full-blown desktop OS on a device like this is major mistake.

[Florin]

Which Atom system does it use? That could make a lot difference to the power consumption.

Well supposedly it was going to use the Atom Z530 (http://www.engadget.com/2010/04/05/hp-slate-to-cost-549-have-1-6ghz-atom-z530-5-hour-battery/) so that'd be with Poulsbo, but the design apparently had only a 5 hour battery life.

[MfA]

Why the hell would they do that? Lincroft has been available for quite a while.

[wco81]

Well supposedly it was going to use the Atom Z530 (http://www.engadget.com/2010/04/05/hp-slate-to-cost-549-have-1-6ghz-atom-z530-5-hour-battery/) so that'd be with Poulsbo, but the design apparently had only a 5 hour battery life.

There was a supposedly leaked internal document posted a few weeks back, which was HP's comparison of their tablet with the iPad.

It pointed out the things it had which the iPad didn't such as webcam, USB.

But it acknowledged that the battery life was inferior, like 3-4 hours IIRC.

[Florin]

Why the hell would they do that? Lincroft has been available for quite a while.

I'm sure HP is aware of this, so there must be some reason. Anyone actually seen W7 running on Moorestown yet? I can imagine this being practically unusable. And probably still using at least twice as much power as a comparable ARM solution.

[Florin]

There was a supposedly leaked internal document posted a few weeks back, which was HP's comparison of their tablet with the iPad.

It pointed out the things it had which the iPad didn't such as webcam, USB.

But it acknowledged that the battery life was inferior, like 3-4 hours IIRC.

Always with the featuritis. Honestly, such people need to take an art of leaving things out class from Apple. Yes, techies will always hate it, but elegance is something any normal consumer can see.

[MfA]

If Win7 can't run at 1024*768 on Lincroft at decent clip then it's time for Microsoft to start firing and hiring. Getting a modern OS to get out of the way (CPU time wise) simply should not be that difficult. If they can't cut the crap in the OS down to a low enough level it's pure incompetence. They should have seen the need for win7 to run on MID/Netbooks/etc coming, even if they didn't see the tablet hype coming.

PS. when you take into account the backlight I really doubt a Moorestown tablet is going to use twice the juice of an Arm tablet.

[Florin]

PS. when you take into account the backlight I really doubt a Moorestown tablet is going to use twice the juice of an Arm tablet.

I didn't mean to imply that. Just the core logic.

[Mike11]

PS. when you take into account the backlight I really doubt a Moorestown tablet is going to use twice the juice of an Arm tablet.
What's Moorestown's power consumption? AFAIK the whole iPad with it's IPS panel needs between 2W (e.g. video playback or browsing via WLAN) and 5W (e.g. full brightness & WLAN activity & nonstop 3D intense gaming).

[MfA]

I don't think anyone has independently tested a Moorestown based design yet.

BTW, it should probably be downclocked to ~750 MHz of course (on non trivial stuff, barrel shifting is nice but not common, Atom is generally 50% faster per clock than an A8 ... so that's a conservative underclock for parity performance).

[Laurent06]

(on non trivial stuff, barrel shifting is nice but not common, Atom is generally 50% faster per clock than an A8 ... so that's a conservative underclock for parity performance).
Really? Can you provide a link that sustains this claim?

[MfA]

http://www.eeejournal.com/2010/05/benchmarks-atom-vs-ipad-a4-vs-iphone.html

http://www.brightsideofnews.com/news/2010/4/7/the-coming-war-arm-versus-x86.aspx?pageid=0

[Laurent06]

http://www.eeejournal.com/2010/05/benchmarks-atom-vs-ipad-a4-vs-iphone.html
You can dismiss results for Atom with HT; comparing a single core vs a multi-threaded one is apple vs orange; and even with HT Atom isn't 50% faster.

The best Atom score is ~2.75 and the best A8 is ~2.6.

http://www.brightsideofnews.com/news/2010/4/7/the-coming-war-arm-versus-x86.aspx?pageid=0

Outside of poor FP performance, which is something that has been discussed here long ago, and poor memory bandwidth (which isn't necessarily a property of A8), I fail to see Atom 50% faster. The Firefox and JS scores are probably not comparable (for instance, v8 work on ARM has only recently received the attention it deserves for ARMv7 chips).

My own experiments make me think that Atom and A8 cores are about as "efficient" (again if you are not bound by IEEE FP computations).

[Ailuros]

Shouldn't it in general be hw X + sw implementation Y + target applications Z in the end?

[Florin]

Apple announced (http://www.apple.com/pr/library/2010/05/03ipad.html) they sold over a million of these.

“One million iPads in 28 days—that’s less than half of the 74 days it took to achieve this milestone with iPhone,” said Steve Jobs, Apple’s CEO. “Demand continues to exceed supply and we’re working hard to get this magical product into the hands of even more customers.”

I didn't expect this kind of appeal for the device.

[Ike Turner]

Apple announced (http://www.apple.com/pr/library/2010/05/03ipad.html) they sold over a million of these.



I didn't expect this kind of appeal for the device.

That's because it's a MAGICAL product.

[aaronspink]

I didn't expect this kind of appeal for the device.


Actually pretty impressive. The category might actually live this time.

[Simon F]

That's because it's a MAGICAL product.
Having just had a brief play with one, I think maybe they are.

[tangey]

Ipad conformancy for openGL es2.0 and 1.1 just added to Khronos site, confirms graphics driver is the same as the one in the iphone (although a later revision), and that the processor is ARMv7. Damn, I was just starting to beleive the PowerPC conspiracy :)

http://www.khronos.org/adopters/conformant-products/

[tangey]

analysis of A4:-
http://www.eetimes.com/news/latest/showArticle.jhtml?articleID=224701036&pgno=1

Interesting point is that the processor in the latest itouch is fabbed at 45nm, whereas the one in the latest iphone is 65nm. I always assumed the itouch/iphone processors where similar across generations.
http://www.eetimes.com/news/latest/showArticle.jhtml?articleID=224701036&pgno=1

[Mike11]

The A4 is on 45nm, unlike the 3GS's SoC but uses the same CPU and GPU. That means it's probably *cheaper* (although there's the extra cost of the 64-bit memory bus)
and there's absolutely no reason to speculate on what other processor Apple might be using. And as seems very clear here,
footprint of modern phones even with discrete app processors can be very very small.
Well, Arun's right as always. According to the new leak the iPhone 4th-gen's SoC seems to be almost identical to the iPad's, nothing fancy here.
From what I can tell, same APL0398 and K4X2G643GE (256MB RAM) part numbers as on the A4.
http://upload.ave.vn/sonpham1102/img/iphone4g/iPhone4g-Hardware-taoviet-2.jpg
http://taoviet.vn/showthread.php?t=16471

The site is pretty slow right now. Engadget's article about it:
http://www.engadget.com/2010/05/12/next-generation-iphone-escapes-in-vietnam/

Update: Obviously Engadget looked the numbers up as well by now and could make out one more number.
If we're not mistaken then we're seeing "339S0084" on that chip from today's fourth-generation iPhone teardown. Guess what? According to Chipworks, that's the Apple A4
microprocessor fabricated by Samsung and the presumed work of Apple's acquired PA Semi and Intrinsity engineers.
The "APL0398" text is also the same as that found on the iPad's A4 system-on-chip.
http://www.engadget.com/2010/05/12/fourth-generation-iphone-teardown-reveals-a4-microprocessor/
No word about the RAM just yet from engadget, but I guess it's only a matter of minutes now until they post an update with more information.

Update 2: Someone pointed out that this new prototype is the 16GB version and maybe the new
"low-end" $99 version of the iPhone with a slower SoC,
like the $99 8GB iPhone 3G right now. And that the high-end version with 80GB
could have more RAM or even a faster/better/newer SoC. IMHO that's unlikely.
Even more RAM seems like a long shot.

[wco81]

If they use the same SOC, then iPhone volumes help drive down the cost for iPad.

[Mike11]

If they use the same SOC, then iPhone volumes help drive down the cost for iPad.
Don't forget the iPod touch. Instead of 2m-3m A4 per quarter Apple can order 15m-20m. That's not bad. Let's see if they can really pull it of and sustain this development cycle for more than (this) one generation.

[green.pixel]

http://www.dodocase.com/


Pretty nice case. :)

[ban25]

http://www.dodocase.com/


Pretty nice case. :)

Still haven't been able to find a case. Thinking about a third party solution like this one....