[The Godfather]

By a whole month if reports may be believed, our Chinese workers will work through the new year holidays to barely deliver on the pushed due date.
Report claims Asian suppliers scramble to meet special iPad 3 launch date: Jobs’ Feb. 24 birthday anniversary | 9to5Mac | Apple Intelligence

[Spheric Harlot]

So it's being rumored that some rumored, completely arbitrary date might not be held for the release of an unannounced, rumored product?

Oh noes!

(I'm sure Steve Jobs would have HATED the vanity of Apple releasing products on his birthday, and I'm sure everybody who works at Apple is sure of this, too. So I call bullshit. March/April release, as in the previous generations.)

[andi*pandi]

I can wait til then. My birthday is in May.

[glideslope]

Originally Posted by Spheric Harlot 

So it's being rumored that some rumored, completely arbitrary date might not be held for the release of an unannounced, rumored product?

Oh noes!

(I'm sure Steve Jobs would have HATED the vanity of Apple releasing products on his birthday, and I'm sure everybody who works at Apple is sure of this, too. So I call bullshit. March/April release, as in the previous generations.)

Agreed. Announce late March. Shipping begining April.

[Eriamjh]

Without a scheduled and announced launch date, there is no such thing as a delay.

[cgc]

My thoughts exactly...but how will people get by with that outdated and unSnappy home button on the iPad 2?

[Eug]

• My guess is the iPad 3 will get an A6.
• The A6 will need 28 nm production.
• 28 nm production is ramping up now, but I guessing not quite fast enough with high enough yields for Apple's liking.
• So, I'm guessing the iPad 3 may be a few months off yet.

I have no data or inside info to back that up.

[Eug]

Originally Posted by Eug 

• My guess is the iPad 3 will get an A6.
• The A6 will need 28 nm production.
• 28 nm production is ramping up now, but I guessing not quite fast enough with high enough yields for Apple's liking.
• So, I'm guessing the iPad 3 may be a few months off yet.

I have no data or inside info to back that up.

IPad 3 Said to Have High-Def Screen, LTE - Bloomberg

Apple Inc. (AAPL)’s next iPad, expected to go on sale in March, will sport a high-definition screen, run a faster processor and work with next-generation wireless networks, according to three people familiar with the product.

The company’s manufacturing partners in Asia started ramping up production of the iPad 3 this month and plan to reach full volumes by February, said one of the people, who asked not to be named because the details aren’t public. The tablet will use a quad-core chip, an enhancement that lets users jump more quickly between applications, two of the people said.

March seems a bit earlier than I was expecting, but I do think the new iPad will have LTE and A6 quad-core.

[P]

I don't get that. Why do you need a quadcore in an iPad - they're not even available on the 13" MBP? I'd rather they just make a version that turbos the clockspeed as needed and pioneer stacked memory.

Meanwhile, the Ivy Bridge graphics is so great that Intel's game demo turns out to be a recorded movie.

[Spheric Harlot]

I really don't think that LTE will be coming before 2013.

For one, it's not actually going into widespread service outside the US (and Japan?) anytime soon.

For another, I have my doubts that they'll improve power management on LTE chipsets enough within this year to make it worthwhile.

Finally, LTE in Europe will run on different frequencies, and there are no chips yet that will cover both spectra—meaning you'll need a separate LTE version for Europe for now.

[Eug]

Originally Posted by P 

I don't get that. Why do you need a quadcore in an iPad - they're not even available on the 13" MBP? I'd rather they just make a version that turbos the clockspeed as needed and pioneer stacked memory.

I dunno, but here are my uneducated 2¢ based on my readings. Please comment. My guess is that it's not a traditional quad-core like Tegra 3, but a big.LITTLE configuration.

For the others here who may not know what that is, that would mean two of the cores would be long pipeline OoO high clocked and thus significantly higher power. The other two cores are short pipeline in-order low clock, and much, much lower power. However, both sets of cores are compatible with the same instruction set. Low compute background process (iOS, notifications, and music playback) would run on the low power chips, and high compute applications (games etc.) would be on the high power chips. In a way it'd almost be like having a dual-core Atom and dual-core i5 in the same laptop.

The A6 is rumoured by everyone to require the 28 nm process, and I agree, and it's been said already that big.LITTLE configurations will essentially require the 32/28 nm process.

Here is a video describing ARM's approach here. It's extremely technical, but nonetheless you may find it interesting.

big.LITTLE: The right size for the right task.

Despite the complexity of that video, they messed up a little on the scaling for task vs. performance examples IMO, but you get the gist of it anyway.

My main question was the timing. I had predicted that a mature 28 nm process with good yields might have been a little later in 2012, but if 28 nm is already good enough that mass A6 production is already under way, then great.

[P]

big.LITTLE is a brand for a specific design: a five-core (pentacore - yay, now we only have heptacore and nonacore left, and then we have all the numbers up to 10!) config, with one single low-power A7 core paired with 4 A15 cores. Neither of those core designs exist today in silicon on the market yet, so that is a little early. nVidia made something similar with Tegra 3, where they have 1 low-power A9 core (made with different transistors and running at a lower clock) and 4 "regular" high power A9 cores. Apple could of course do something like that, with a low-power core that runs when the phone is in standby or just handling voice traffic and powering up the others as needs require.

I think that this is a very interesting design, but I fail to see the use for more than 2 cores. Both the iPad and the iPhone have very limited background processing going on. I would rather see a better memory interface, as I said, and a more powerful GPU than 4 general purpose cores. Intel's recent Atom announcements (Medfield) also show that Intel has a huge advantage in the caching subsystem, being able to handle far more outstanding L2 cache misses. That's something ARM should work on rather than squeezing in more cores.

The A6 is rumoured by everyone to require the 28 nm process, and I agree, and it's been said already that big.LITTLE configurations will essentially require the 32/28 nm process.

Persistent rumors put Apple in bed with TSMC, and TSMC cancelled their 32nm process, so 28nm it is. Chips from the high performance 28nm node are out now (namely the Radeon 7970) but that doesn't mean that the low power node is ready. It should be around the corner, though. Apple needs to shrink - they're at 45nm with a comparatively large (122mm2) chip - but I always figured that they'd do a straight shrink now ("A5+"), maybe add more graphics cores to support the higher res, and then make improvements when the A15 was out. Just going to 4 A9 cores could be straightforward enough, and with the node-and-a-half step down they could do it in a chip smaller than the A5, but I just don't see why.

OTOH...Intel shows a Medfield Atom singlecore beating an A9 dualcore at the same power envelope - including binary translation of the ARM ISA. I know Medfield isn't ready to ship yet, but a dualcore of THAT I could see in an iPad. Too bad they picked a weaker GPU than what Apple already has in the A5.

[Eug]

Originally Posted by P 

Persistent rumors put Apple in bed with TSMC, and TSMC cancelled their 32nm process, so 28nm it is. Chips from the high performance 28nm node are out now (namely the Radeon 7970) but that doesn't mean that the low power node is ready. It should be around the corner, though.

Yeah, regarding the timing, I had originally been predicting the second half of 2012 for a new iPad release, based on rumours about 28 nm new-model ARM, but who knows, maybe Bloomberg is right and they've been able to ramp faster.

Whatever the case I don't expect an iPad 2s in March and an iPad 3 in October. I think there will be only one iPad release this year.

[P]

Correct, there is that production ramp to consider. The 7970 is a high-end low volume product - a new iPad would need a LOT of production capacity.

I've been thinking more about that asynchronousasymmetric design. It's an interesting idea, but it really needs an in-order design of the low-power core to keep the wattage truly low (OoO needs power to be kept on to the execution window at all times), but it also needs cache coherency protocols. The two likely options both fail one of those: The A9 is OoO, the A8 is a single core design without cache coherency. That last could conceivably be handled (by giving the A8 its own L2 which the A9 cores do not have access to) but that means that moving a thread to the A9 cores would be tricky.

[cgc]

Asynchronous computing was a big concept in the early 90's but never went anywhere.

[P]

Argh... I mean asymmetric, not asynchronous. Although asynchronous would be interesting as well, simply from a novelty standpoint.

[Eug]

Originally Posted by P 

Correct, there is that production ramp to consider. The 7970 is a high-end low volume product - a new iPad would need a LOT of production capacity.

I've been thinking more about that asynchronousasymmetric design. It's an interesting idea, but it really needs an in-order design of the low-power core to keep the wattage truly low (OoO needs power to be kept on to the execution window at all times), but it also needs cache coherency protocols. The two likely options both fail one of those: The A9 is OoO, the A8 is a single core design without cache coherency. That last could conceivably be handled (by giving the A8 its own L2 which the A9 cores do not have access to) but that means that moving a thread to the A9 cores would be tricky.

Perhaps Apple will be the first out with 2 x A7 (in-order) + 2 x A15 (OoO).

[cgc]

Originally Posted by P 

Argh... I mean asymmetric, not asynchronous. Although asynchronous would be interesting as well, simply from a novelty standpoint.

I thought it may get some attention nowadays as a potential way to conserve power.

[Eug]

I'm not buying it but...

Special ebento Feb. 3 for iPad 3?

Apple、２月上旬にiOS 5.1とiPad 3のSpecial イベントを開催？