Possibly more than that.
They want control over the design of their chips and devices, not so much a large manufacturing facility. Why pay and operate a manufacturing facility if you can have sweeping control what an other company produces for you.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Inside Apple's A7 Chip, M7 Motion Coprocessor, and More from the iPhone 5s
- Thread starter MacRumors
- Start date
- Sort by reaction score
Possibly more than that.
They want control over the design of their chips and devices, not so much a large manufacturing facility. Why pay and operate a manufacturing facility if you can have sweeping control what an other company produces for you.
BS. The entire chip design was made by Apple. Samsung only produces the chips because they have such large manufacturing capabilities. TSMC will soon take over hopefully.
It is with some irony Samsung manufactures chips which far exceed their own in technological advantage.
why "hopefully"? samsung produces in good quality (too). doesn't really matter who produces it in the end.
TSMC has a process lead. Their 20nm will be ready next year whereas Samsung's won't. Expect A8 on TSMC 20nm.
So TSMC says. How often often have they been correct? It will likely be some time in 2014, but doubtful the dates they were they were predicting earlier this year.
Samsung tends not to over promise and under deliver. That actually where TSMC has a clear "lead" over its competitors. TSMC will likely go "live" sooner with lower yields than Samsung. They may get to "good enough" yields faster than Samsung. But it isn't like they have a 12+ month lead on everyone else. As long as Apple keeps packing these iOS device updates into Q3-Q4 timeframes, TSMC's gap doesn't mean as much.
[e.g., if TSMC goes 20nm in Q1-Q2 2014 and Samsung in Q4 2014... by Q1 2015 they are both in the same place. ]
TSMC has been in risk production for 20nm since 1st quarter of this year. I think they'll be ready for next year as they say. Especially since it's only a single process they're offering.
TSMC was pushing out 28nm chips in late 2011. Samsung is almost 2 years behind with regards to that on 28nm. It shouldn't be a surprise that TSMC will also be first to 20nm.
They don't make any now, so it would be a long time.
If talking about designing.... The processor on a Mac Mini 2013 will blow the doors off this A7 on OS X like intensive workloads. If Intel stumbles then Apple could switch over perhaps in 3-4 years. That is also conditional on user workload demands going flat. [ That also presumes that Microsoft implodes also. ]
As long as Intel does a good job there is just as little reason for Apple to get into the "replace Core iX and Xeon E5 level of performance" business as it is for them to get into the actual chip making business. There are others who do it very good and it is far cheaper to spread manufacturing costs over a larger body of vendors.
Facts are is that in the larger, legacy PC form factors market Windows still dominates. Moving away from the same infrastructure that Windows is on doesn't buy Apple much if anything. As long as Apple is holding down just some sub 10% of that market the other 90+% have the manufacturing volume to drive down costs lower than Apples; even on similarly priced devices. In mobile space sharing R&D and manufacturing costs with other ARM (and Imagine Tech GPU ) implementers makes more sense then forking into something purely proprietary.
If A10 , A11 , A12 , etc chips get up into the "good enough" performance range for most customers and Windows implodes, then it isn't so much whether OS X would switch to a ARM foundation as much as get canceled.
I'm not surprising they are first. As I noted it is ability to deal with volume production and high yields early is the "surprising" part. If this round they have making more designs fit their process rather than try to adapt to a wider variety of design then perhaps the track record will change.
still, I mean anything you buy these days are what, 40 50 dollars at least? a chip like that cost as much as my plastic iPhone case? cmon...
Samsung has got 20nm in engineering and 14nm in their sights. Apple isn't going to switch over completely to TSMC, but diversify the supply chain. BTW, most of the wafers are made by Samsung in a US fab, whereas all the wafers are made overseas from TSMC; American jobs at stake.
Last edited:
What TSMC says may often not be directly related to product or mass-availability in the real world, however, since Apple signed with TSMC and also gave them good money beforehand for the ramp-up and production capacity at their cutting-edge fabs, I see no reason why I should doubt any delay happening.
And I would bet money on that. I´m that sure. No way there´s gonna be an A8 still on 28nm HKMG.
What I am more concerned about is actual production capacity and yield issues with their processes. Because that would eat into their gross margins and available iPhones a lot. I am not an investor though, so that´s not directly affecting me anyways.
Probably the worst post i've seen on here.
You seem like a great guy.
More evidence for Rogue (or an Apple variant of it), as you've said. Only G6400 and G6430 have 4 clusters according to this wikipedia table (BTW: what's the significance of the "30"? Is it something that can be determined from the delayered die? ...Is it cache?)
This suggests G6630 in the iPad 5, which doubles the GFLOPS - but it only increases the pixel and texture figures by 50% (according to the above linked table). Perhaps Apple might also clock it 33% higher, to double those figures?
EDIT According to this, the "30" seems to be a licensing, not technical, difference, in that you're allowed to dual/quad core it:
Last edited:
It wasn't until it reached our shores, if memory serves, that people found out that the Galaxy didn't have Samsung's exynos in it, rather a snapdragon.
The _30 have frame buffer compression logic. They are performance optimized rather than area optimized.
FWIW, ImgTec does not refer to the designs as having cores anymore. They envision the cluster architecture as fully scalable. There's less need to repeat logic as the front-end and back-end stay the same and aren't married to each core.
My impression after talking to several people is that Apple will not push the clocks higher. In fact, the iPad 4 GPU had a lower clock than the iPhone 5 GPU. I suspect we might see an 8 cluster design with a similar GPU clock.
Very intriguing piece of technology... it will be interesting to see if they tie into the rumored iWatch.
Wow, looking at those pics reminds me that all this truly is, well, magic.
I do not understand how exactly all this works. I find it incredibly amazing that our devices can do what they do.
What a wonderful age we live in: one in which we can literally see into each other's dreams and witness the impossible through a little handheld non-living thing we can talk to!
The next great revolution will be the Matrix; virtual reality pumped into the brain directly. I hope I'm alive to see when (not if) it happens!
I do not understand how exactly all this works. I find it incredibly amazing that our devices can do what they do.
What a wonderful age we live in: one in which we can literally see into each other's dreams and witness the impossible through a little handheld non-living thing we can talk to!
The next great revolution will be the Matrix; virtual reality pumped into the brain directly. I hope I'm alive to see when (not if) it happens!
Just as with similar sensor processors on Samsung and Motorola smartphones, the M7 watches for orientation and movement while using very low power.
In other words the power hungry main CPU can sleep while the sensor processor watches for things like shaking the device, being put into a pocket, etc... depending on what sensors are hooked up.
One cool thing that Apple has assigned to their processor, is using the accelerometer to count steps in the background. The software API lets any fitness app query how many steps have taken place during any time frame.
Another thing it does is watch to see how fast you're going and, for example, signal the main CPU to not try to connect to open WiFi hotspots while you're in a car.
(I wonder if the M7 even had a public name before Motorola made a big deal about theirs in the Moto X. Usually no one cares about this under-the-hood stuff.)
That seems the most logical scenario, but they only go up to 6 clusters - the G6630. Though that's according to at-times-inaccurate Wikipedia; and Apple might scale it internally. BTW: Have you found a better reference for Rogue specs? I'd love to get on top of this.
I still have hopes they'll unhistorically make it x4 the iPhone 5S (16 clusters) - suddenly the most powerful console!
Til November anyway.
That seems the most logical scenario, but they only go up to 6 clusters - the G6630. Though that's according to at-times-inaccurate Wikipedia; and Apple might scale it internally. BTW: Have you found a better reference for Rogue specs? I'd love to get on top of this.
I still have hopes they'll unhistorically make it x4 the iPhone 5S (16 clusters) - suddenly the most powerful console!
It's been designed to scale pretty high, so I wouldn't look at Apple as being hindered by announced products.
There are no good sources, and ImgTec has been somewhat tight-lipped on how to break down the performance parameters.