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.
Apple Set to Shift From Samsung to TSMC for 'A6' Chip Production Next Year?
- Thread starter MacRumors
- Start date
- Sort by reaction score
Arstechnica.com has a comment thread on this as well:
http://arstechnica.com/apple/news/2...12.ars?comments=1&p=21797561#comment-21797561
Rocketman
http://arstechnica.com/apple/news/2...12.ars?comments=1&p=21797561#comment-21797561
Rocketman
Oh it is possible to clock them high. You are kinda stuck trying to find the elusive Golden Sample though. Heat and power requirements tend to be an issue.
Increasing clock speed to increase performance is inefficient. Power consumption equals CV^2f. And higher f requires higher V, too, to keep the edges sharp. So you burn much more power. Adding pipes, threads, and cores is much more efficient.
It seems like it was a bad move on Samsungs part to allow their business relationship to suffer. There was too much upside potential to risk losing Apple as a customer, plus I doubt Samsung can compete with Apple in the tablet market.
The A4/5/6 are SoCs which also include the GPU in addition to the processor, which with the A6, might have one powerful enough to drive "retina"
Which is why Intel flipped back the other way with the Core Series. But make no mistake if someone were to challenge Intel in speed in the desktop world, they could re-bin their stuff for higher speeds (if that was all it took) to be on top.
They have a 5Ghz club.
I know what the A4/A5/A6 are, what makes you think the current ones aren't powerful enough ?
You do realise 1024x768 is paltry resolution and that the current SGX543mp2 is quite a bit more powerful than way older cards who could power much higher resolutions right ?
No on-air restriction makes the distinction pointless. I've seen intel cpus north of 7 GHz on nitrogen. So what?
It's an empty statement to begin with. Retina is an ambiguous term. Need a PPI, size to go with it.
Yeah most of the rigs are using water. Like I said heat and power are the issues. Most also seems to agree that running the CPUs hard like that shortens their lifespan.
ARM-based Macs? Maybe not. ARM-based Apple products that sell very near or lower than the price of low-end Macs, but at a much higher volume with good profit margins, to a huge percentage of their customers who use these products for many of the very same things for which they currently use Macs?
What the name of that theory that suitable cheaper stuff displaces older "better" stuff (mostly) out of the market?
Depends what you quantify as "hard". The part that eats into a CPU's lifetime is not the overclocking itself, it's the over-volting, which accelerates the aging process through electromigration. There's no real expectation of shortened lifespan if you don't over-volt.
You are assuming that Samsung doesn't have customer that they turn down. Additionally, some of TSMC are going to show up at Samsung's door because Apple's runs have kicked them out of TSMC foundries.
This could also be a "huge blow" to Apple also. It isn't like TSMC hasn't had process and yield problems before. Nor does TSMC necessarily have the memory which gets embedded into the A6 part. It is more than just an ARM chip that goes into the package.
Likewise all of the Flash production that Samsung does for Apple. If Apple soaks up some other supplier's production that just means there are more folks coming from other suppliers. As long as single foundries are billion $ investments there are is a "reshuffling the deck chairs" effect that play a contributing role.
The bigger deal is that Samsung would loose a potential financing partner. While other folks might have as large an order for ARM chips. Few will have the upfront $200-600M in case Samsung needs some seed cash to get a new factory built or some other large capital outline. Apple can do that out of the "spare change" drawer . Apple has more money to loan than many banks do at this point.
It depends. Higher frequency also has a linear effect on electromigration if the current is unidirectional, as it would be for any dynamic circuits on-chip.
Right, which is why all chips eventually have a non-infinite lifetime. Over-volting is a much more significant contributor to failure than anything else from what I've seen. I merely stated it the way I did because I've never heard anyone imply frequency increases alone can have an appreciable affect on lifetime.
They still will have lost Apple, and that is lost revenue. You say 3 or 4 companies without Apple, I say 4 or 5 companies with Apple. Get it? Losing Apple is not something Samsung wants. They're a big customer, and no matter how many people replace them, they still lose a big customer.
some trolls actually try to act like they are really mac regulars..this one
that troll didn't even bother to fake being a regular..join date, june 2011.
his/her content: meaningless.
time wasted: reading it: five seconds.
writing response: seven seconds.
realizing there are actually grown up people who have time to go to sites they have no involvement with or interest in, and make up sxxx: hardly priceless (another four seconds wasted..grrr)
that troll didn't even bother to fake being a regular..join date, june 2011.
his/her content: meaningless.
time wasted: reading it: five seconds.
writing response: seven seconds.
realizing there are actually grown up people who have time to go to sites they have no involvement with or interest in, and make up sxxx: hardly priceless (another four seconds wasted..grrr)
With a paltry $172.5 billion in revenue, how will Samsung survive with only 96% of its revenue intact?
The sky is falling....
Apple's license for ARM includes modifying the designs to their needs which includes a SoC design with the PowerVR.
Apple has patented several hardware designs for memory management separate from the standard ARM design they incorporate into their A series SoC combos.
There are also many other effects that result in "aging" of the circuits, including hot carrier effects, which are also frequency-dependent.
And voltage dependent
, which is why people are being forced to go high-k.I guess my point is that any frequency dependent characteristics can be compensated for (in a simplistic sense) by the amount of time the computer is loaded. Now, given that someone who over-volts likely loads his or her computer more often than other users, the same is not true for servers, which see practically 24/7 utilization but nominal operating voltages.
And voltage dependent
I guess my point is that any frequency dependent characteristics can be compensated for (in a simplistic sense) by the amount of time the computer is loaded. Now, given that someone who over-volts likely loads his or her computer more often than other users, the same is not true for servers, which see practically 24/7 utilization but nominal operating voltages.
yes. anything that ages the chip is going to do so because of current (and hence electric fields), and current is going to depend on frequency and voltage.
On air and with out increasing votlage I know I can get my old AMD 64 3000+ to run at 2.25 ghz before it gets unstable. 2.3 and it gets flaky and crashes on me.
Not to bad pulling about a 15% overclock.