The thing that I love about these teardowns, and which really stands out to me, is how well designed Apple's products are internally as well as externally. If you compare the teardowns between the iPad and the HP Touchpad ( http://www.techrepublic.com/photos/cracking-open-the-hp-touc... ), it's pretty clear why Apple has such a lead in tablet manufacturing efficiency and cost.
Yeah, I hear you - that's Tim Cook the supply chair & operational efficiency maestro. Or more specifically, Apple's unique ability to successfully coordinate design, engineering, and manufacturing for appeal, quality, and profitability.
They aren't superfluous.. Wall St uses the information to determine what manufacturers in Apple's supply chain are receiving the most revenue from those millions of units sold. Before Apple started releasing their supplier list, teardowns were the only way to figure out whose products were inside Apple's products.
I think they mean the model number, LTN097QL01-A02, is a samsung model number. The 1024x768 lcd in the linked document has model number LTN097XL01-A01.
Wow, beautiful photographs. They always do a great job.
Has anyone looked closer at the A5X package in these pictures? Is that a glimpse of PC board under that can? Is there some kind of multi-die package (and not just POP)?
Although the heat spreader is oddly shaped, the A5X isn't big enough to justify more than one die. I wonder why they unstacked the RAM; I would imagine either for thermal reasons or for more memory bandwidth.
There is an entire industry built around decapping and reverse-engineering ICs, usually to make cheap clones in China or somewhere else that'll turn a blind eye to IP infringement.
Annoyingly, I can't find the article I remember reading, which was a look at the debugging / test labs at AMD or Intel. They had all manner of depackaging, imaging, probing, and even in-situ modification tools. They could cut traces with incredibly precise laser pulses, and use bond-wires or additional metal deposits to test faulty designs. Incredible stuff (and a nightmare for anyone who wants "tamper resistance")
And 1GB of ram confirmed. Apple's policy of not mentioning the ram made sense to me when they started, because it did seem like a amount of ram in the initial model (256mb). But they've doubled the ram with each new generation and made no mention of it. I guess apple has just made the calculation that 1) it doesn't matter to most consumers (the ipad is supposed to be magic), and 2) they'll be outspeced by the competition over the lifetime of the model, anyway.
RAM on a tablet affects performance in a very straightforward way. Why mention any specs if it's just a "consumer electronics device"?
I think the reason they don't mention RAM is because they don't want to get in a silly numbers based arms race. They want to say "here's the whole package, it's pretty great, and better than the other guy's stuff, even if some of his specs are a little better on paper".
Hm... the battery numbers are surprising. The battery is 43.5Wh, which is huge. Over 9.5 hours us "use", that comes to 4.6W average. My Dell XPS 15 L501x (very much not a "light" machine!) has a 9-cell 91Wh battery and got about 5 hours continuous on it when it was new, for about 18.2W.
So the iPad 3 is drawing 25% the current of a beefy 2.9GHz Arrandale laptop? That's shockingly high, IMHO, especially considering that the screen area of the 15.6" laptop is twice that of the tablet.
Similar, the numbers they show aren't much higher than the range (30-50Wh, 6-9 hours) already achieved by a bunch of netbooks on the market.
With LED backlighting, isn't most of the screen's power draw driven by the switching? And the iPad has a lot of pixels to switch--almost certainly more than your Dell laptop.
Honestly I'm no expert. But I'd find it very surprising if there was much difference at the same lumens output. The backlight is not per-pixel in a LED-backlit display (maybe you're thinking of OLED displays?), though I understand it's often per-region to allow large dark areas to be under-lit.
And note that LED backlights are routinely put up as having overall lower power consumption than CCFL panels like the one in my laptop.
Really my point is this: that's a laptop-sized battery in that thing, and it's seeing only slightly super-laptop power draw.
Your point misses the mark: with the 1/2 of the battery of your notebook iPad moves 50% more pixels for twice as long:
It's 2 * 1.5 * 2 = 6 times more efficient than your notebook. Of course, usage patterns are also different, but it's still wonderful piece of engineering.
More efficient only in your oddball unit of pixels-joules/second. I'm not interested in arguing about features or the design tradeoffs of an Apple device, that always leads to a platform flame just like this one.
I'm saying that the actual power draw of the iPad is much, much closer to that of a laptop than most of us expected.
IIRC the transistor on each pixel colour in the display blocks a fixed amount of light, so the increased pixel density means less area for light to pass through, so a brighter backlight is required to achieve the same light output.
(It's possible of course that these high-DPI displays use a process technology with smaller transistors, anyone know?)
High-DPI displays have a lower pixel aperture ratio than traditional displays, so the backlight needs to be brighter to give the same perceived brightness. Supposedly the new iPad retina display uses twice as many backlighting elements as the iPad 2.
I don't believe switching is the major component in the display's power draw, since adjusting brightness has such a significant effect on battery life. If switching was the dominant factor, you would expect battery life to be roughly independent of brightness settings.
Your Dell XPS has 1920 * 1080 resolution, that is, around 2M pixels. The iPad has 2048 * 1536 which is 3M pixels. Your claim "the screen area of the 15.6" laptop is twice that of the tablet" is actually irrelevant considering the number of pixels and the processing power needed to move all these pixels around for all smooth animations.
Not sure what you mean. Pixels don't take power by themselves (and of course screen area does take power all by itself). Moving them does, sure, but the GPU (GeForce 420M) in my laptop is signicantly faster (literally 5x the GFLOPs per spec) than the 4x SGX 54x cores in the A5X.
I'm not making a platform flame here or asking you to defend your favorite device. I'm just expressing surprise at the apparent power draw of a device that conventional wisdom claims is exceedingly thrifty.
I think you are confused on how displays/graphics work. An active display needs to be lit, and the more pixels you have the more lighting is required, so physical size matters less than resolution. As for the GPU, moving the pixels can require more power but you must constantly redraw as your scene changes (you could cache static backgrounds as an image, but this is often not done). GPU power needed is most definitely impacted by resolution and has absolutely no bearing on screen size.
Yeah, comparing the specs (CPU/GPU performance/display resolution/weight/battery capacity/disk/RAM) of the iPad vs. MacBook Air over time is pretty compelling evidence that the touch / keyboard & mouse paradigms are converging and the current dichotomy between the two types of devices/OSs won't last much longer
"Like the A5, the A5X system features a 1 GHz dual-core CPU. The upgrade that earns it an 'X' is the new GPU, which Apple claims outperforms even Nvidia's Tegra 3 processor."
Considering that even last year's A5 outperforms Tegra 3 on many benchmarks casting aspersions here seems meritless. You might quibble with Apple's 4x claim but that is not how this reads.
So for next year's iPad we can expect a tock architechture update (to A-15) for the SoC and a tick process shrink for the baseband chip. Not sure how many batteries they'll be able to throw out but hopefully some and it'll be thin again.
36 comments
[ 3.3 ms ] story [ 53.5 ms ] threadHas anyone looked closer at the A5X package in these pictures? Is that a glimpse of PC board under that can? Is there some kind of multi-die package (and not just POP)?
http://guide-images.ifixit.net/igi/i6ZlJCQfdaHyktRY.medium
See http://www.lps.umd.edu/MicroelectronicsIntegration/Microelec... and http://www.break-ic.com/topics/break-ic.asp for 2 overviews.
There is an entire industry built around decapping and reverse-engineering ICs, usually to make cheap clones in China or somewhere else that'll turn a blind eye to IP infringement.
There's also 'Dr. Decapitator' of ROM extraction fame: http://decap.mameworld.info/
Annoyingly, I can't find the article I remember reading, which was a look at the debugging / test labs at AMD or Intel. They had all manner of depackaging, imaging, probing, and even in-situ modification tools. They could cut traces with incredibly precise laser pulses, and use bond-wires or additional metal deposits to test faulty designs. Incredible stuff (and a nightmare for anyone who wants "tamper resistance")
I think the reason they don't mention RAM is because they don't want to get in a silly numbers based arms race. They want to say "here's the whole package, it's pretty great, and better than the other guy's stuff, even if some of his specs are a little better on paper".
So the iPad 3 is drawing 25% the current of a beefy 2.9GHz Arrandale laptop? That's shockingly high, IMHO, especially considering that the screen area of the 15.6" laptop is twice that of the tablet.
Similar, the numbers they show aren't much higher than the range (30-50Wh, 6-9 hours) already achieved by a bunch of netbooks on the market.
And note that LED backlights are routinely put up as having overall lower power consumption than CCFL panels like the one in my laptop.
Really my point is this: that's a laptop-sized battery in that thing, and it's seeing only slightly super-laptop power draw.
It's 2 * 1.5 * 2 = 6 times more efficient than your notebook. Of course, usage patterns are also different, but it's still wonderful piece of engineering.
I'm saying that the actual power draw of the iPad is much, much closer to that of a laptop than most of us expected.
(It's possible of course that these high-DPI displays use a process technology with smaller transistors, anyone know?)
I don't believe switching is the major component in the display's power draw, since adjusting brightness has such a significant effect on battery life. If switching was the dominant factor, you would expect battery life to be roughly independent of brightness settings.
iPad is definitely impressive engineering feat.
I'm not making a platform flame here or asking you to defend your favorite device. I'm just expressing surprise at the apparent power draw of a device that conventional wisdom claims is exceedingly thrifty.
"Like the A5, the A5X system features a 1 GHz dual-core CPU. The upgrade that earns it an 'X' is the new GPU, which Apple claims outperforms even Nvidia's Tegra 3 processor."
Considering that even last year's A5 outperforms Tegra 3 on many benchmarks casting aspersions here seems meritless. You might quibble with Apple's 4x claim but that is not how this reads.
http://www.anandtech.com/show/5681/apples-a5x-die-and-size-r...
Qualcomm's chip is confirmed as still being the 45 nm model.
http://www.macrumors.com/2012/03/08/new-ipad-appears-to-util...
So for next year's iPad we can expect a tock architechture update (to A-15) for the SoC and a tick process shrink for the baseband chip. Not sure how many batteries they'll be able to throw out but hopefully some and it'll be thin again.