TSMC reveals potential performance, power improvements in 'A14' chipset

Posted:
in General Discussion
Apple supplier TSMC has detailed some of the performance or power improvements from its new 5nm chip manufacturing process, expected to be used in the "iPhone 12."

Credit: Andrew O'Hara, AppleInsider
Credit: Andrew O'Hara, AppleInsider


The A14 chipset is rumored to be the first major Apple Silicon to be based on TSMC's 5nm process node, which will bring either performance or power efficiency improvements to Apple's flagship iPhones.

At its annual Technology Symposium, TSMC has itself given numbers for how the new 5nm chips could improve on Apple's past silicon. AnandTech has provided a graph of the improvements, based on those numbers.

Credit: AnandTech
Credit: AnandTech


The current A13 chipset is based on TSMC's 7nm process node, represented by the N7 designation. The N5 category represents TSMC's 5nm process node, which is expected to be used in Apple's A14 chipset.

Compared to the N7 A13 chip, an Apple A14 chip could either be 15% faster at the same amount of power or 30% more power efficient. It may also be somewhere in the middle. Apple, for its part, has historically focused on power efficiency in its iPhone processors. But, what Apple does with these gains in Apple Silicon Macs is yet to be seen.

According to AnandTech, TSMC is also mass producing 5nm chipsets at its production facilities. Apple's next-generation System-on-Chip (SoC) -- the A14 processor -- is expected to be among the first candidates to be made using the new process.

TSMC also gave a future look at its 3nm chipmaking process, which could be either 25% to 30% more power efficient or 10% to 15% more performant than the 5nm process. The chipmaker's 3nm silicon are expected to enter production by the second half of 2022.

Of course, these numbers purely represent hardware gains. Apple can further optimize performance or power efficiency in iOS, or through hardware and software optimization.

Apple is expected to debut four iPhone models in 2020: 5.4-inch and 6.1-inch "iPhone 12" models, and 6.1-inch and 6.7-inch "iPhone 12 Pro" devices. All are expected to sport a new squared-off design, support for 5G, and other upgrades.
«1

Comments

  • Reply 1 of 28
    When Apple presents a new A-series chip there is usually a chart showing the performance increase over the previous chip. Do we know how much of that increase is attributable to Apple’s chip design and how much is attributable to a new process node?
    rob53watto_cobra
  • Reply 2 of 28
    KITAKITA Posts: 375member


    Credit: AnandTech
    Credit: AnandTech


    The current A13 chipset is based on TSMC's 7nm process node, represented by the N7 designation. The N5 category represents TSMC's 5nm process node, which is expected to be used in Apple's A14 chipset.

    Compared to the N7 A13 chip, an Apple A14 chip could either be 15% faster at the same amount of power or 30% more power efficient. It may also be somewhere in the middle. Apple, for its part, has historically focused on power efficiency in its iPhone processors. But, what Apple does with these gains in Apple Silicon Macs is yet to be seen.

    The A13 Bionic uses N7P, not N7.
    FileMakerFeller
  • Reply 3 of 28
    thttht Posts: 4,072member
    Yup. Exciting. Will be interesting if they implement SVE2 or not. The phone chip may have 32 MB of system level cache. The desktop chips maybe in the 64 to 128 MB of levels!

    Also curious how the rumors on A14 are basically nil? There's been more rumors for Mac chips than the A14 or A14X.


    watto_cobra
  • Reply 4 of 28
    KITAKITA Posts: 375member
    tht said:
    Will be interesting if they implement SVE2 or not. 

    Arm has yet to announce anything regarding ARMv9. So I'm not sure if that's still to come or perhaps we might only see it in a 2021 chip.
  • Reply 5 of 28
    thttht Posts: 4,072member
    When Apple presents a new A-series chip there is usually a chart showing the performance increase over the previous chip. Do we know how much of that increase is attributable to Apple’s chip design and how much is attributable to a new process node?
    I'd say generally it is half and half. The new process node enables both higher clocks and more transistors. More transistors (chip design) mean more performance, it's just not easily made into a story in a news blurb like clock frequencies are.

    So, the 1.8x density improvement is just as huge, if not bigger, than the power consumption or performance rows. It's all related anyways.
    razorpitwatto_cobra
  • Reply 6 of 28
    bluefire1bluefire1 Posts: 1,176member
    The best news about the upcoming iPhone 12 models is that they’ll all haveQualcomm inside.
    caladanian
  • Reply 7 of 28
    MplsPMplsP Posts: 3,329member
    30% is a huge jump - when was the last time we saw anything near that with Intel processors? Even if the actual jump is smaller it is still quite significant. Along with everyone else, I’m very curious to see what the new Macs can do. 

    bluefire1 said:
    The best news about the upcoming iPhone 12 models is that they’ll all haveQualcomm inside.
    Why? I’ve yet to see what 5G will do for my iPhone beyond make it more expensive and power hungry. 
    rob53BeatsGG1watto_cobra
  • Reply 8 of 28
    KITAKITA Posts: 375member
    MplsP said:
    30% is a huge jump - when was the last time we saw anything near that with Intel processors? Even if the actual jump is smaller it is still quite significant. Along with everyone else, I’m very curious to see what the new Macs can do. 

    bluefire1 said:
    The best news about the upcoming iPhone 12 models is that they’ll all haveQualcomm inside.
    Why? I’ve yet to see what 5G will do for my iPhone beyond make it more expensive and power hungry. 
    30% decrease in power consumption while offering the same performance is going from N7 (A12) to N5.

    So if we do N7P (A13) to N5 (assumed A14), then we have:

    ~22% power consumption decrease with the same performance

    -OR-

    ~7.5% performance increase with the same power consumption
  • Reply 9 of 28
    tmaytmay Posts: 5,465member
    KITA said:
    MplsP said:
    30% is a huge jump - when was the last time we saw anything near that with Intel processors? Even if the actual jump is smaller it is still quite significant. Along with everyone else, I’m very curious to see what the new Macs can do. 

    bluefire1 said:
    The best news about the upcoming iPhone 12 models is that they’ll all haveQualcomm inside.
    Why? I’ve yet to see what 5G will do for my iPhone beyond make it more expensive and power hungry. 
    30% decrease in power consumption while offering the same performance is going from N7 (A12) to N5.

    So if we do N7P (A13) to N5 (assumed A14), then we have:

    ~22% power consumption decrease with the same performance

    -OR-

    ~7.5% performance increase with the same power consumption
    I'm interested in the die size, transistor count, and the inevitable x-ray shot, which will give a good indication of the complexity and cost of the A14.

    Not that it really matter, since I'm going to be ordering/preordering an iPhone 12 Pro Max when available, with the only unknown for me being the screen refresh rate; 60 Hz or 120 Hz.
    watto_cobra
  • Reply 10 of 28
    thttht Posts: 4,072member
    KITA said:
    tht said:
    Will be interesting if they implement SVE2 or not. 
    Arm has yet to announce anything regarding ARMv9. So I'm not sure if that's still to come or perhaps we might only see it in a 2021 chip.
    SVE2 is already shipping in a ARMv8 core (Fujitsu A64FX), and is not contingent on ARMv9.

    It may come to pass that SVE2 will be a Mac only Apple Silicon due to power and chip size concerns, kind of mirroring Intel's AVX512 segmentation between Core and Xeon on Intel 14nm. That's all converging now with Ice Lake and follow-ons, but I can see Apple implementing SVE2 only for Mac silicon. And, scientific computing in a phone isn't something that is done regularly, unless there is some AR process that could use it.

    The A13 has 8.5b transistors. An A14 could have 10b to 15b transistors. They could do a lot of things with an extra 2b to 6b transistors.
    watto_cobra
  • Reply 11 of 28
    tmay said:
    KITA said:
    MplsP said:
    30% is a huge jump - when was the last time we saw anything near that with Intel processors? Even if the actual jump is smaller it is still quite significant. Along with everyone else, I’m very curious to see what the new Macs can do. 

    bluefire1 said:
    The best news about the upcoming iPhone 12 models is that they’ll all haveQualcomm inside.
    Why? I’ve yet to see what 5G will do for my iPhone beyond make it more expensive and power hungry. 
    30% decrease in power consumption while offering the same performance is going from N7 (A12) to N5.

    So if we do N7P (A13) to N5 (assumed A14), then we have:

    ~22% power consumption decrease with the same performance

    -OR-

    ~7.5% performance increase with the same power consumption
    I'm interested in the die size, transistor count, and the inevitable x-ray shot, which will give a good indication of the complexity and cost of the A14.

    Not that it really matter, since I'm going to be ordering/preordering an iPhone 12 Pro Max when available, with the only unknown for me being the screen refresh rate; 60 Hz or 120 Hz.

    60 Hz display in iPhone 12 generation would be a bummer, with even $200 Android phones coming up with 90Hz displays (when they shouldn't be, with SoCs lacking enough power to drive the display. Not an issue for Apple's SoCs for last 3 years). And it wasn't Android OEMs who showcased the benefits of higher refresh rates to the world. It was Apple themselves implementing it in iPad pros more than 2 years ago, with Android OEMs choosing to follow Apple's direction in this aspect. Apple themselves not implementing it in iPhones even after 2 years is unacceptable. If quantity is the issue, then it would lead to questions on Tim Cook's mastery over the supply chain management.

    edited August 2020 caladanian
  • Reply 12 of 28
    KITAKITA Posts: 375member
    tht said:

    SVE2 is already shipping in a ARMv8 core (Fujitsu A64FX), and is not contingent on ARMv9.
    Is it? No mention of SVE2, just v8.2-A and SVE.


  • Reply 13 of 28
    thttht Posts: 4,072member
    KITA said:
    tht said:

    SVE2 is already shipping in a ARMv8 core (Fujitsu A64FX), and is not contingent on ARMv9.
    Is it? No mention of SVE2, just v8.2-A and SVE.
    You are right, it's SVE, the first version. Whether SVE2 is contingent about ARM ISA v9, I don't know about then.


    watto_cobra
  • Reply 14 of 28
    rob53rob53 Posts: 2,676member
    With the die size being half what it is now, do you see Apple using more than one SoC for the Macs or simply keep the die the same size and add twice the amount of components to it? The die size only refers to how ARM would supply their typical CPU, correct? It doesn't necessarily mean Apple's A-series or M-series SoCs would scale the same amount when you add all the extra stuff Apple crams into their SoC. (yes/no?)

    I'm still hoping Apple will play around with some M-series blades that could be used in the Mac Pro or a configurable Mac maxi. Build some CPU blades, GPU blades, and combination blades on a nice backplane for compact servers and a supercomputer like the old days.
    edited August 2020 watto_cobra
  • Reply 15 of 28
    MplsPMplsP Posts: 3,329member
    KITA said:
    MplsP said:
    30% is a huge jump - when was the last time we saw anything near that with Intel processors? Even if the actual jump is smaller it is still quite significant. Along with everyone else, I’m very curious to see what the new Macs can do. 

    bluefire1 said:
    The best news about the upcoming iPhone 12 models is that they’ll all haveQualcomm inside.
    Why? I’ve yet to see what 5G will do for my iPhone beyond make it more expensive and power hungry. 
    30% decrease in power consumption while offering the same performance is going from N7 (A12) to N5.

    So if we do N7P (A13) to N5 (assumed A14), then we have:

    ~22% power consumption decrease with the same performance

    -OR-

    ~7.5% performance increase with the same power consumption
     Compared to the N7 A13 chip, an Apple A14 chip could either be 15% faster at the same amount of power or 30% more power efficient.” 

    am I missing something?
    thtwatto_cobra
  • Reply 16 of 28
    KITAKITA Posts: 375member
    MplsP said:
    KITA said:
    MplsP said:
    30% is a huge jump - when was the last time we saw anything near that with Intel processors? Even if the actual jump is smaller it is still quite significant. Along with everyone else, I’m very curious to see what the new Macs can do. 

    bluefire1 said:
    The best news about the upcoming iPhone 12 models is that they’ll all haveQualcomm inside.
    Why? I’ve yet to see what 5G will do for my iPhone beyond make it more expensive and power hungry. 
    30% decrease in power consumption while offering the same performance is going from N7 (A12) to N5.

    So if we do N7P (A13) to N5 (assumed A14), then we have:

    ~22% power consumption decrease with the same performance

    -OR-

    ~7.5% performance increase with the same power consumption
    “ Compared to the N7 A13 chip, an Apple A14 chip could either be 15% faster at the same amount of power or 30% more power efficient.” 

    am I missing something?
    The A13 uses N7P, the A12 uses N7. @AppleInsider still hasn't corrected this.
    muthuk_vanalingam
  • Reply 17 of 28
    KITAKITA Posts: 375member
    muthuk_vanalingam said:

    And it wasn't Android OEMs who showcased the benefits of higher refresh rates to the world. It was Apple themselves implementing it in iPad pros more than 2 years ago, with Android OEMs choosing to follow Apple's direction in this aspect. 
    The Razer Phone came out with a 120 Hz display and a variable refresh rate only a few months after the iPad Pro.

    This was not some last minute knee-jerk reaction by Razer to design their phone the moment the iPad was announced. It was very much a culmination of where the industry was headed and technology becoming available to the majority of OEMs. Even the Snapdragon 835 (used in the Razer Phone) was announced with features such as variable refresh rate (qsync) months before the iPad Pro was announced.

    The benefits of high refresh rates and variable refresh rates have been known for a very long time prior to the iPad Pro.
  • Reply 18 of 28
    tmaytmay Posts: 5,465member
    tmay said:
    KITA said:
    MplsP said:
    30% is a huge jump - when was the last time we saw anything near that with Intel processors? Even if the actual jump is smaller it is still quite significant. Along with everyone else, I’m very curious to see what the new Macs can do. 

    bluefire1 said:
    The best news about the upcoming iPhone 12 models is that they’ll all haveQualcomm inside.
    Why? I’ve yet to see what 5G will do for my iPhone beyond make it more expensive and power hungry. 
    30% decrease in power consumption while offering the same performance is going from N7 (A12) to N5.

    So if we do N7P (A13) to N5 (assumed A14), then we have:

    ~22% power consumption decrease with the same performance

    -OR-

    ~7.5% performance increase with the same power consumption
    I'm interested in the die size, transistor count, and the inevitable x-ray shot, which will give a good indication of the complexity and cost of the A14.

    Not that it really matter, since I'm going to be ordering/preordering an iPhone 12 Pro Max when available, with the only unknown for me being the screen refresh rate; 60 Hz or 120 Hz.

    60 Hz display in iPhone 12 generation would be a bummer, with even $200 Android phones coming up with 90Hz displays (when they shouldn't be, with SoCs lacking enough power to drive the display. Not an issue for Apple's SoCs for last 3 years). And it wasn't Android OEMs who showcased the benefits of higher refresh rates to the world. It was Apple themselves implementing it in iPad pros more than 2 years ago, with Android OEMs choosing to follow Apple's direction in this aspect. Apple themselves not implementing it in iPhones even after 2 years is unacceptable. If quantity is the issue, then it would lead to questions on Tim Cook's mastery over the supply chain management.

    You forget that Apple needs something like 150 million screens just for the new models that will be released in October, and will need 35 million or so screens for the Christmas quarter alone. All of those will have to be the highest spec screens available, and some of those may be 120Hz. Those screens are significantly better quality, and more difficult to source in volume than the 90Hz screens you mention.

    Oh, and the Razer phone is a gaming phone, and isn't really comparable to Flagships in features, so not a good comparison.
    edited August 2020 watto_cobra
  • Reply 19 of 28
    KITAKITA Posts: 375member
    tmay said:

    Oh, and the Razer phone is a gaming phone, and isn't really comparable to Flagships in features, so not a good comparison.
    The Razer Phone happened to be a flagship phone that just happened to come from Razer. That would be like dismissing their laptops for productivity because the company is associated with gaming.

    Going back to what the other user had said, suggesting that Apple led the industry to adopt high refresh rate displays is false.

    EDIT:

    Here's the 120Hz Sharp display from mid-2016


    edited August 2020 muthuk_vanalingamavon b7
  • Reply 20 of 28
    BeatsBeats Posts: 2,550member
    KITA said:
    muthuk_vanalingam said:

    And it wasn't Android OEMs who showcased the benefits of higher refresh rates to the world. It was Apple themselves implementing it in iPad pros more than 2 years ago, with Android OEMs choosing to follow Apple's direction in this aspect. 
    The Razer Phone came out with a 120 Hz display and a variable refresh rate only a few months after the iPad Pro.

    This was not some last minute knee-jerk reaction by Razer to design their phone the moment the iPad was announced. It was very much a culmination of where the industry was headed and technology becoming available to the majority of OEMs. Even the Snapdragon 835 (used in the Razer Phone) was announced with features such as variable refresh rate (qsync) months before the iPad Pro was announced.

    The benefits of high refresh rates and variable refresh rates have been known for a very long time prior to the iPad Pro.

    How come when the iKnockoffs come out with a feature months or years after Apple it was "natural" or "already in production" but when Apple INVENTS something and it releases a month after Samsung crapped a turd on the market to beat Apple then "Apple COPIED!!"

    Fu**ing stupid logic.
    thtwatto_cobra
Sign In or Register to comment.