Chip strip reveals 'handmade' Apple A6 Electron microscope photos of an Apple A6 processor lifted out of an iPhone 5 confirm the presence of a dual-core ARM CPU within the system-on-a-chip plus a trio of Imagination Technologues PowerVR graphics cores. The SoC was pulled from the handset by the team at iFixit.com, and the die extracted from its ceramic package and …
The question is, does Apple have enough suppliers to drop Samsung, before Samsung decide it's not worth dealing with Apple, and taking the risk to not sign a new fab contract?
Considering that Samsung currently owe $1bn to Apple, (I know it's still in the courts, but if they do end up paying it), would it cheaper to pay fees for breaking a contract early, and let Apple stew in a channel of no parts, or use it to beat Apple round the head? (i.e. drop the shitty patent suits, or you get no more parts, and we'll pay the contract fees)?.. Anybody know what the default fees might be?
In no business, have I ever heard of a parts supplier, continue to supply parts, to a rival, that is doing it's best to bankrupt the supplier.. it just does not compute... and what if Apple win, and Samsung shuts down.. sure the different fabs and labs might be broken up, but those would go the prefered bidder..
Maybe Apple want to assimilate teh Samsung R&D and fabs, without the hassle of setting up their own.. but that would just expose their lack of internal investment or over-outsourcing in the first place...
If that was the case, I can't see the Samsung investors not selling to another rival like Google anyway.. just for the shits and giggles..
Plus one of the benefits of being a big conglomerate is you become like a hydra - you cut one head off and two more grow back. Even if Sammy Electronics were taken to bankruptcy, Sammy Group could close that down and continue with its other operations. Plus, their patent pool would be owned by the group, so there'd be nowt stopping them growing a 'Samsung Chips and Bits' and a 'Samsung Phones and Telly's' division in its place.
@Steve I
""I can't see the Samsung investors not selling to another rival like Google anyway.. just for the shits and giggles.."
With billions of $ at stake, they'd sell to the highest bidder."
Don't misquote me please, I said "If that was the case", just before that.. I never claimed it was the case, or was the current situation..
You may well be right, and probably are; I'm sort of being devils advocate here, especially with the current questions around juror's conduct behind closed doors.
However, even though I know pride comes before a fall; If I was a boardmember of Samsung, and any part of the company was looking like it was going to be sold to Apple, I'd fall on my sword and sell what I had to Google, before I'd let those fruity harpies near it.
Having done some chip layout myself, it is amazing how much performance you can gain by using the human brain rather than an algorithm. I would put my money on performance first with thermal being an extra benefit.
However, the algorithm generally does it while you get a cup of coffee (or with this size of design maybe lunch) but hand layout can take weeks!
> the spacing lowers hot spots.
It's been a while since I was anywhere near CPU design, but back then, whitespace was indicative of a rushed design; less exacting florplanning requirements meant you could get the chip our quicker, at the cost of a larger (= more expensive) die and the commensurate lower yield...
Vic.
#1 -- To me, a "haphazard" layout would actually indicate that it was laid out by a computer, which would have little regard to making sure that all the GPU cores were lined up in a nice row etc.
#2 -- Laying out a chip like this by hand might increase clock speed and/or reduce power consumption but layout doesn't affect the actual function of the chip and doesn't explain how a 1.2GHz core could run faster than a 1.5GHz version of the same core.
I would side with AnandTech on this and say the most interesting thing about the A6 is that it likely has an in-house ARM design and not that it (may have been) laid out by hand.
Samesong will probably copy everything they can from this design.
They describe themselves as a "quick follower". which for everyone else means "not from China cloner".
The next Galaxy S4 will have a dualcore "hand made" chip with a tricore graphics chip.
it will be long and thin with rounded corner instead of pebble, and will have an aluminium back instead of plastic.
Available in black and white.
You heard here first,
The SIII has 2GB or RAM if you get the 1.5Ghz dual core version. It's the quad core 1.4Ghz version that only has 1GB RAM. Either way it's a faster chip than the A6.
The SIII has an overclocked Mali 400 graphics core. In benchmarks it sits between the dual PowerVR cores of the iPhone 4S and the four PowerVR cores used in the latest iPad. Whilst it's not a direct comparison I'd guess that gives it similar performance to the three PowerVR cores in the iPhone5.
The iPhone 5 which has just been released appears to be slower than the 3 month old Galaxy SIII, at least on paper.
I'm guessing he got downvoted because the Anandtech benchmarks don't appear to resemble anyone else's. Unfortunately nearly every rag out there has decided to reference the Anandtech results as proof positive that the iPhone 5 is faster (Inquirer, Yahoo, PC Mag; I'm looking at you).
http://browser.primatelabs.com/android-benchmarks
http://browser.primatelabs.com/ios-benchmarks
Anandtech posted results from multiple benchmarks, all you're referencing is Geekbench which is well known to be unreliable for cross-platform comparisons and only good for testing a limited set of CPU features.
But if you really want to use Geekbench as your point of comparison, don't forget that the update was only released today to use the new armv7s instruction set of the A6, so we haven't seen what that will do to scores. See here: http://www.primatelabs.com/blog/2012/09/apple-a6/
Looks like library parts placed into a die to maximise spacing and reduce thermal issues. Only Intel with their insanely over complex designs needs to "fill" the chip.
Any layout is a mix of autorouting and manual expertise or else anyone could buy the tools and do it.
This post has been deleted by its author
"If Apple has eschewed well-established - and used in pretty much every chip design elsewhere - die layout software tools in favour of a more manual layout..."
This is nonsense......having done chip design/layout for the last 20 years, it is NOT uncommon for layout to be done by "hand". Invariably it is for speed issues and timing optimisation/closure.
Nothing special, original or out of the ordinary. Move along please.
Beer, because chip design is hard work and shit pay, but interesting. Should have been an accountant.
The article is referring to the layout of the top-level blocks. Obviously the blocks themselves are done automatically.
Yes it does look like there is some wasted area in the design. All this does really is make the chip more expensive to produce (fewer chips per wafer, you see) hence lower margins. But it lowers risk - if some blocks turn out bigger than planned, you don't have to rip-up and re-do the whole thing to make it all fit. Time is money - lots of it in Apples case.
If they are shipping 5 million in the first week it seems they can afford to re-spin it in a more area-optimized form before very long. Probably already on its way.
"The article is referring to the layout of the top-level blocks. Obviously the blocks themselves are done automatically."
In the iFixit article they theorize it was a full custom job:
"Generally, logic blocks are automagically[sic] laid out with the use of advanced computer software.
However, it looks like the ARM core blocks were laid out manually—as in, by hand."
Pure supposition and a big chance of being pure tosh.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
