I'm in a quandary. I dislike samsung quite a bit and would rather TSMC get the earnings. But being a Texan- the A8 would be made in Austin. I think I'm still rooting for samsung here. God help me.
But Samsung microprocessors are a joke just like their Exinnos or their 100$ phones, the SoC design by Apple that is what makes A8 chip a great microprocessor
That’s one of the big problems with throwing in a bunch of RAM. Until that new tech where only the active RAM is powered gets released, that is. End of the decade.
<span style="line-height:1.4em;">About 2 hours of battery life...minimum</span>
That’s one of the big problems with throwing in a bunch of RAM. Until that new tech where only the active RAM is powered gets released, that is. End of the decade.
The new Sammy DRAM is faster and lower power than other DRAM...
I am not up on current RAM tech but in the early days, many computers used SRAM (Static RAM) -- SRAM used lower power... But was slower..
Be that as it may, I believe that we'll see more RAM in iPads or headless companion devices which can support a larger battery.
At this point, Apple should be seriously considering producing its own chips. Apple is big enough and the mobile revolution is not a fad. This would be a worthwhile investment for the company and could give them access to leading edge fabrication techniques. Currently they are at the mercy of chip suppliers in other countries and are constantly playing catch up to Intel.
No one has caught up to Intel yet. The funny thing here is that they can be two process nodes behind with ARM CPUs and still do better than Intel.
But isn't the iPhone 1 GB Ram faster (or equal) than 3 GB Ram found in Samsung Galaxy nOTE 3? JUST like its Dual Core faster than their Quad Core. I think iPhone will have 4 GB of Ram this year.. that means Samsung premium phones must have around 8 GB to compete with iPhone
It isn't that simple. Android devices need room for the virtual machine (the ripoff of Java) which takes up a fixed amount of RAM, so in their base models they need more RAM simply to get the required performance due to the allocation of RAM to that VM. After the VM and other Android overhead is taken care of, RAMcan be used equally on both platforms. Well as about as equally as Java and native apps allow.
As to RAM speed that is an unknown at this time. Well unknown to me. I don't expect significant speed differences in the devices but Apple is known to use a wider data path in some of its iPad implementations. In a nut shell though more RAM is better. However that needs to be available RAM.
At this point, Apple should be seriously considering producing its own chips. Apple is big enough and the mobile revolution is not a fad. This would be a worthwhile investment for the company and could give them access to leading edge fabrication techniques. Currently they are at the mercy of chip suppliers in other countries and are constantly playing catch up to Intel.
No one has caught up to Intel yet. The funny thing here is that they can be two process nodes behind with ARM CPUs and still do better than Intel.
What if you combined Apple's design capability with Intel or IBM's manufacturing capability?
Are you sure that Apple acquired (got ownership) of those foundry facilities? AIR, Apple just contracted excess capacity for Sammy to use if/when the Sammy Texas capacity was exceeded.
Well, that's according to Charlie Demerjian of SemiAccurate. He has an exceptional track record in predicting Apple's future plans.
Now, it's obvious that Apple will not bet the whole company on that single fab right away. They will depend on TSMC and Samsung for a few more years, until they master building their own silicon.
I am not up on current RAM tech but in the early days, many computers used SRAM (Static RAM) -- SRAM used lower power... But was slower..
Nope, SRAM is the fastest memory available. 700GB/s peak performance in current Intel processors. And that's gigabytes, not gigabits. Unfortunately, it is way too expensive to produce, so it is used primarily in L1 cache (32KB-64KB).
Quote:
Originally Posted by wizard69
No one has caught up to Intel yet. The funny thing here is that they can be two process nodes behind with ARM CPUs and still do better than Intel.
Well, yes. Intel are the leader, but their processes are too expensive. With the 16nm FinFET and 14nm nodes TSMC and Samsung will catch up.
The decline in PC sales significantly harm Intel and their ability to invest new technologies as retooling is extremely expensive and is only feasible with large volume production.
Additionally, 20nm is a borderline technology. Up until now, the prices for producing a single Integrated Circuit were falling down with each new shrink. 16/14nm will actually increase the cost per die compared to 20nm and the price increase will continue with 10nm and below.
That is one reason for Intel wanted to push the development of the 450mm wafers (300mm currently used). But, that's on pause for now.
Quote:
Originally Posted by wizard69
It isn't that simple. Android devices need room for the virtual machine (the ripoff of Java) which takes up a fixed amount of RAM, so in their base models they need more RAM simply to get the required performance due to the allocation of RAM to that VM. After the VM and other Android overhead is taken care of, RAMcan be used equally on both platforms. Well as about as equally as Java and native apps allow.
As to RAM speed that is an unknown at this time. Well unknown to me. I don't expect significant speed differences in the devices but Apple is known to use a wider data path in some of its iPad implementations. In a nut shell though more RAM is better. However that needs to be available RAM.
It would be perfect if only for the JVM overhead. What's worse is the automatic memory management based on garbage collection. That process is so ineffective, that applications easily consume 3-5 times the memory actually required, even when compared to unoptimised code C / Obj-C. That ratio increases significantly (10-15 times) when compared to well architectured and well crafted application. Those are my personal experiences with both Java and .NET. The one-size-fits-them-all memory management approach is idiotic.
Now, most Android libraries are native (written in C/C++), so the JVM overhead (CPU and memory) is not as bad as described above, where I compare pure Java/.NET vs pure C + Obj-C.
Nevertheless, 3GB in Android phone is a requirement to fairly compare to iPhone, as you already stated
Are you sure that Apple acquired (got ownership) of those foundry facilities? AIR, Apple just contracted excess capacity for Sammy to use if/when the Sammy Texas capacity was exceeded.
Well, that's according to Charlie Demerjian of SemiWiki. He has an exceptional track record in predicting Apple's future plans. http://semiaccurate.com/2013/07/12/apple-has-their-own-fab/
The full report requires professional membership, though.
Now, it's obvious that Apple will not bet the whole company on that single fab right away. They will depend on TSMC and Samsung for a few more years, until they master building their own silicon.
I am not up on current RAM tech but in the early days, many computers used SRAM (Static RAM) -- SRAM used lower power... But was slower..
Nope, SRAM is the fastest memory available. 700GB/s peak performance in current Intel processors. And that's gigabytes, not gigabits. Unfortunately, it is way too expensive to produce, so it is used primarily in L1 cache (32KB-64KB).
Well, it'll be extremely fast, have no doubt about it. Unfortunately, the price of the SoC will be exorbitant and production yields will be extremely low.
If Intel, or anyone else for that matter, were able to build CPU with 2MB L1 cache, they wouldn't have resorted to L2 and L3 caches.
Well, it'll be extremely fast, have no doubt about it. Unfortunately, the price of the SoC will be exorbitant and production yields will be extremely low.
If Intel, or anyone else for that matter, were able to build CPU with 2MB L1 cache, they wouldn't have resorted to L2 and L3 caches.
We have also been looking at the packaging; Apple’s suppliers have redesigned the package-on package (PoP) stack that has been used since the original iPhone way back in 2007. Specifically, the connections between the top memory package, and the lower section which houses the actual A7 chip have changed.
We are looking up at the bottom of the (top) memory part of the PoP. It’s a bit messy, because the solder balls have flowed a bit as we separated the two halves of the PoP.
The thing to note is the two rows of solder balls around the periphery of the package, which have a pitch of 0.4 mm. These are what connect the gigabyte of LPDDR2 memory in the top half of the PoP to the A6 chip itself.
It’s immediately obvious that we have three rows of interconnect, with many more connections. Close examination showed that the ball pitch is now 0.35 mm, which doesn’t appear much of a reduction, but is definitely of note to those in the business.
And, with three rows, we now have a solder ball count of 456, up from 272 in the A6. I make that a 68% increase.
Which leaves us with a question – if there’s still only a gigabyte of memory up top (albeit LPDDR3 instead of LPDDR2), and it’s still 2 x 32-bit interface, why do we need all that extra interconnect?
Question: Is it possible that Sammy, as the foundry for the A7, could manufacture and QA test the chip and not know it was 64-bit?
Wouldn't it be something if Apple could cut the cord with Sammy for APUs and RAM?
On the first question, I very much doubt that. Samsung, however, is more of a conglomerate than a corporation. So, having high enough business ethics, it is possible for Samsung Semiconductor to keep the secret from, say, Samsung Mobile.
According to SemiAccurate, Samsung Semiconductor personnel work on bringing Apple's New York Fab (the ex-Global Foundries one) up to speed.
While the A8 production is already split between TSMC and Samsung, the A9 will be mostly their own, as far as I read, again on SA.
RAM and flash memory are two topics I have read nothing about. But, if Apple manages to pull the trick with building those on their own, Samsung are in for a huge surprise as Apple orders currently represent between 20% and 25%, I think.
On a side note, I hope Apple move to Quantum dot displays soon, dropping Samsung Display as well and taking the lead in display quality once again.
Question: Is it possible that Sammy, as the foundry for the A7, could manufacture and QA test the chip and not know it was 64-bit?
Wouldn't it be something if Apple could cut the cord with Sammy for APUs and RAM?
On the first question, I very much doubt that. Samsung, however, is more of a conglomerate than a corporation. So, having high enough business ethics, it is possible for Samsung Semiconductor to keep the secret from, say, Samsung Mobile.
According to SemiAccurate, Samsung Semiconductor personnel work on bringing Apple's New York Fab (the ex-Global Foundries one) up to speed.
While the A8 production is already split between TSMC and Samsung, the A9 will be mostly their own, as far as I read, again on SA.
RAM and flash memory are two topics I have read nothing about. But, if Apple manages to pull the trick with building those on their own, Samsung are in for a huge surprise as Apple orders currently represent between 20% and 25%, I think.
On a side note, I hope Apple move to Quantum dot displays soon, dropping Samsung Display as well and taking the lead in display quality once again.
I did some surfing, and apparently Apple uses Micron for the LPDDR3 RAM in the iPhone 5S PoP SoC.
Apple uses [B][I][COLOR=blue]the APL0698 variant of the A7 chip in the iPhone 5S and the second-generation iPad Mini[/COLOR][/I][/B].[19] This A7 is manufactured by Samsung on a high-? metal gate (HKMG) 28 nm process[20][1] and the chip includes over 1 billion transistors on a die 102 mm2 in size.[2] According to ABI Research the A7 drew 1100 mA during fixed point operations and 520 mA during floating point operations, while its predecessor, the A6 processor in the iPhone 5, drew 485 mA and 320 mA.[21] [B][I][COLOR=blue]It is manufactured in a package on package (PoP) together with 1 GB of LPDDR3 DRAM with a 64-bit wide memory interface onto the package.[/COLOR][/I][/B][6][22]
Apple A7 (APL5698) Apple uses[B][I][COLOR=blue] the APL5698 variant of the A7 chip in the iPad Ai[/COLOR][/I][/B]r. Its die is identical in size and layout to that of the first A7 and is manufactured by Samsung.[23] However, unlike the first version of the A7,[B][I][COLOR=blue] the A7 used in the iPad Air is not a package-on-package (PoP), having no stacked RAM. Instead it uses a chip-on-board mounting, immediately adjacent DRAM,[/COLOR][/I][/B] and is covered by a metallic heat spreader, similar to the Apple A5X and A6X.[23][24] [/QUOTE]
I'm hoping this equates to some truly innovate features. I've been a Iphone user since the 3GS and upgraded to the 4S because at the time Siri was innovate and exciting. After years of marginal upgrades such as a longer screen and fingerprint id that I cared very little about, I grew weary of waiting for something new. After two months of visiting the Note 3 in a local Best Buy and being apprehensive about the transfer process of my data and the general usability I had grown accustomed to with Apple and Itunes, I traded my 4S in. I was using my phone as a universal remote with a free app and the built in IR receiver because my DirectV remote went missing. I have 96 gigs of space, which is great because my 64 gig Iphone was always full and apple hasn't committed to making a higher capacity phone yet. I'm able to drag and drop and file onto my phone without messing with the cumbersome Itunes or any other software, or having to sync for 10-20 minutes. All my computer's video files just play through the Note 3 without me having to convert anything.
I feel about the Note 3 the same way I felt when I turned on my first Iphone. I truly do hope that Apple does use all this new speed to innovate and make the apps and ios truly something special again. Heres to hoping apple gives us more that a os facelift to have us fall in love with Iphones all over again.
Comments
Digitimes so we know it's fake but still:
I'm in a quandary. I dislike samsung quite a bit and would rather TSMC get the earnings. But being a Texan- the A8 would be made in Austin. I think I'm still rooting for samsung here. God help me.
But Samsung microprocessors are a joke just like their Exinnos or their 100$ phones, the SoC design by Apple that is what makes A8 chip a great microprocessor
About 2 hours of battery life...minimum
Edit: they could of course buy a tech company offering advanced memory controllers....oh wait...
That’s one of the big problems with throwing in a bunch of RAM. Until that new tech where only the active RAM is powered gets released, that is. End of the decade.
The new Sammy DRAM is faster and lower power than other DRAM...
I am not up on current RAM tech but in the early days, many computers used SRAM (Static RAM) -- SRAM used lower power... But was slower..
Be that as it may, I believe that we'll see more RAM in iPads or headless companion devices which can support a larger battery.
No one has caught up to Intel yet. The funny thing here is that they can be two process nodes behind with ARM CPUs and still do better than Intel.
It isn't that simple. Android devices need room for the virtual machine (the ripoff of Java) which takes up a fixed amount of RAM, so in their base models they need more RAM simply to get the required performance due to the allocation of RAM to that VM. After the VM and other Android overhead is taken care of, RAMcan be used equally on both platforms. Well as about as equally as Java and native apps allow.
As to RAM speed that is an unknown at this time. Well unknown to me. I don't expect significant speed differences in the devices but Apple is known to use a wider data path in some of its iPad implementations. In a nut shell though more RAM is better. However that needs to be available RAM.
What if you combined Apple's design capability with Intel or IBM's manufacturing capability?
Are you sure that Apple acquired (got ownership) of those foundry facilities? AIR, Apple just contracted excess capacity for Sammy to use if/when the Sammy Texas capacity was exceeded.
Well, that's according to Charlie Demerjian of SemiAccurate. He has an exceptional track record in predicting Apple's future plans.
http://semiaccurate.com/2013/07/12/apple-has-their-own-fab/
The full report requires professional membership, though.
Here's another interesting article from Jan 2, 2013:
http://semiaccurate.com/2013/01/02/apples-silicon-design-capabilities-increase/
Now, it's obvious that Apple will not bet the whole company on that single fab right away. They will depend on TSMC and Samsung for a few more years, until they master building their own silicon.
I am not up on current RAM tech but in the early days, many computers used SRAM (Static RAM) -- SRAM used lower power... But was slower..
Nope, SRAM is the fastest memory available. 700GB/s peak performance in current Intel processors. And that's gigabytes, not gigabits. Unfortunately, it is way too expensive to produce, so it is used primarily in L1 cache (32KB-64KB).
No one has caught up to Intel yet. The funny thing here is that they can be two process nodes behind with ARM CPUs and still do better than Intel.
Well, yes. Intel are the leader, but their processes are too expensive. With the 16nm FinFET and 14nm nodes TSMC and Samsung will catch up.
The decline in PC sales significantly harm Intel and their ability to invest new technologies as retooling is extremely expensive and is only feasible with large volume production.
Additionally, 20nm is a borderline technology. Up until now, the prices for producing a single Integrated Circuit were falling down with each new shrink. 16/14nm will actually increase the cost per die compared to 20nm and the price increase will continue with 10nm and below.
That is one reason for Intel wanted to push the development of the 450mm wafers (300mm currently used). But, that's on pause for now.
It isn't that simple. Android devices need room for the virtual machine (the ripoff of Java) which takes up a fixed amount of RAM, so in their base models they need more RAM simply to get the required performance due to the allocation of RAM to that VM. After the VM and other Android overhead is taken care of, RAMcan be used equally on both platforms. Well as about as equally as Java and native apps allow.
As to RAM speed that is an unknown at this time. Well unknown to me. I don't expect significant speed differences in the devices but Apple is known to use a wider data path in some of its iPad implementations. In a nut shell though more RAM is better. However that needs to be available RAM.
It would be perfect if only for the JVM overhead. What's worse is the automatic memory management based on garbage collection. That process is so ineffective, that applications easily consume 3-5 times the memory actually required, even when compared to unoptimised code C / Obj-C. That ratio increases significantly (10-15 times) when compared to well architectured and well crafted application. Those are my personal experiences with both Java and .NET. The one-size-fits-them-all memory management approach is idiotic.
Now, most Android libraries are native (written in C/C++), so the JVM overhead (CPU and memory) is not as bad as described above, where I compare pure Java/.NET vs pure C + Obj-C.
Nevertheless, 3GB in Android phone is a requirement to fairly compare to iPhone, as you already stated
Hmm...
Interesting reads!
Where did/does Papermaster fit in to all of this?
Seems like Apple has executed a 3-card monte.
OK, humor me,,,
What if the A8 were to use SRAM instead of DRAM?
OK, humor me,,,
What if the A8 were to use SRAM instead of DRAM?
Well, it'll be extremely fast, have no doubt about it. Unfortunately, the price of the SoC will be exorbitant and production yields will be extremely low.
If Intel, or anyone else for that matter, were able to build CPU with 2MB L1 cache, they wouldn't have resorted to L2 and L3 caches.
What about this?
http://www.mosys.com/m/high-density-memory.php
Edit:
Check out the backgrounds of MoSys top mgmt:
http://www.mosys.com/m/company-management.php
What about this?
http://www.mosys.com/m/high-density-memory.php
The related technology patent is from December, 2006. A lot of foundries (including TSMC) are already capable of building the circuitry.
We will see in less than an year what Apple engineers pulled off with the A8. Up to now A-series CPUs simply kick ass (c:
We have also been looking at the packaging; Apple’s suppliers have redesigned the package-on package (PoP) stack that has been used since the original iPhone way back in 2007. Specifically, the connections between the top memory package, and the lower section which houses the actual A7 chip have changed.
Here we have the memory package from an iPhone 5:
[IMG ALT=""]http://forums.appleinsider.com/content/type/61/id/37838/width/500/height/1000[/IMG]
Memory Package from the iPhone 5
We are looking up at the bottom of the (top) memory part of the PoP. It’s a bit messy, because the solder balls have flowed a bit as we separated the two halves of the PoP.
The thing to note is the two rows of solder balls around the periphery of the package, which have a pitch of 0.4 mm. These are what connect the gigabyte of LPDDR2 memory in the top half of the PoP to the A6 chip itself.
Now if we look at the A7 memory package:
[IMG ALT=""]http://forums.appleinsider.com/content/type/61/id/37839/width/500/height/1000[/IMG]
Memory Package from the iPhone 5s (A7)
It’s immediately obvious that we have three rows of interconnect, with many more connections. Close examination showed that the ball pitch is now 0.35 mm, which doesn’t appear much of a reduction, but is definitely of note to those in the business.
And, with three rows, we now have a solder ball count of 456, up from 272 in the A6. I make that a 68% increase.
Which leaves us with a question – if there’s still only a gigabyte of memory up top (albeit LPDDR3 instead of LPDDR2), and it’s still 2 x 32-bit interface, why do we need all that extra interconnect?
http://www.chipworks.com/en/technical-competitive-analysis/resources/blog/inside-the-a7/
Question: Is it possible that Sammy, as the foundry for the A7, could manufacture and QA test the chip and not know it was 64-bit?
Wouldn't it be something if Apple could cut the cord with Sammy for APUs and RAM?
Question: Is it possible that Sammy, as the foundry for the A7, could manufacture and QA test the chip and not know it was 64-bit?
Wouldn't it be something if Apple could cut the cord with Sammy for APUs and RAM?
On the first question, I very much doubt that. Samsung, however, is more of a conglomerate than a corporation. So, having high enough business ethics, it is possible for Samsung Semiconductor to keep the secret from, say, Samsung Mobile.
According to SemiAccurate, Samsung Semiconductor personnel work on bringing Apple's New York Fab (the ex-Global Foundries one) up to speed.
While the A8 production is already split between TSMC and Samsung, the A9 will be mostly their own, as far as I read, again on SA.
RAM and flash memory are two topics I have read nothing about. But, if Apple manages to pull the trick with building those on their own, Samsung are in for a huge surprise as Apple orders currently represent between 20% and 25%, I think.
On a side note, I hope Apple move to Quantum dot displays soon, dropping Samsung Display as well and taking the lead in display quality once again.
I did some surfing, and apparently Apple uses Micron for the LPDDR3 RAM in the iPhone 5S PoP SoC.
Toshiba makes the Flash...
http://www.isuppli.com/Teardowns/News/Pages/Groundbreaking-iPhone5s-Carries-199-BOM-and-Manufacturing-Cost-IHS-Teardown-Reveals.aspx
Unclear about the iPhone 5S Display manufacturer...
Other than that, Qualcom components are the major costs in the BOM.
[QUOTE]Apple A7 (APL0698)
Apple uses [B][I][COLOR=blue]the APL0698 variant of the A7 chip in the iPhone 5S and the second-generation iPad Mini[/COLOR][/I][/B].[19] This A7 is manufactured by Samsung on a high-? metal gate (HKMG) 28 nm process[20][1] and the chip includes over 1 billion transistors on a die 102 mm2 in size.[2] According to ABI Research the A7 drew 1100 mA during fixed point operations and 520 mA during floating point operations, while its predecessor, the A6 processor in the iPhone 5, drew 485 mA and 320 mA.[21] [B][I][COLOR=blue]It is manufactured in a package on package (PoP) together with 1 GB of LPDDR3 DRAM with a 64-bit wide memory interface onto the package.[/COLOR][/I][/B][6][22]
Apple A7 (APL5698)
Apple uses[B][I][COLOR=blue] the APL5698 variant of the A7 chip in the iPad Ai[/COLOR][/I][/B]r. Its die is identical in size and layout to that of the first A7 and is manufactured by Samsung.[23] However, unlike the first version of the A7,[B][I][COLOR=blue] the A7 used in the iPad Air is not a package-on-package (PoP), having no stacked RAM. Instead it uses a chip-on-board mounting, immediately adjacent DRAM,[/COLOR][/I][/B] and is covered by a metallic heat spreader, similar to the Apple A5X and A6X.[23][24]
[/QUOTE]
http://en.wikipedia.org/wiki/Apple_A7
This is so Apple -- using [practical] variations on a [technology] theme!
It also suggests that Apple can easily up the RAM on a full-size (or larger) iPad, AppleTV... whatever!
I feel about the Note 3 the same way I felt when I turned on my first Iphone. I truly do hope that Apple does use all this new speed to innovate and make the apps and ios truly something special again. Heres to hoping apple gives us more that a os facelift to have us fall in love with Iphones all over again.