Patent Applications Reveal Apple's Research Into 3D Chip Packaging

Discussion in 'MacRumors.com News Discussion' started by MacRumors, Jun 12, 2018.

  1. MacRumors macrumors bot

    MacRumors

    Joined:
    Apr 12, 2001
    #1
    [​IMG]


    Apple's persistent quest for better performance, longer battery life, and slimmer form factors appears to be driving its research into advanced chip packaging technologies. So-called "2.5D" and "3D" packaging methods stand to offer significant gains in all of these areas by increasing memory bandwidth, reducing power consumption, and freeing up space for higher-capacity batteries.

    Apple has been an aggressive adopter of new device packaging methods, mostly thanks to integrated fan-out (InFO) innovations provided by foundry partner TSMC. TSMC's success has spurred it into further developing and diversifying its packaging offerings, and TSMC has emerged as an industry leader in packaging techniques.

    [​IMG]

    While versions of TSMC's InFO packaging have brought performance improvements to Apple devices, such as better thermal management and improved package height, it has largely not been a direct enabler of improved electrical performance. This is set to change with future packaging techniques and is already seen in some products that utilize interposers for higher density interconnects to on-package memory, such as High Bandwidth Memory (HBM).

    The primary memory candidate for inclusion in such a package would be conforming to the Wide I/O set of standards described by JEDEC, and mentioned by name in several of the patents. This memory improves on LPDDR4 by increasing the number of channels and reducing the transfer speed per channel, thus increasing the overall bandwidth but lowering the energy required per bit.

    Interposers do, however, pose several issues for mobile devices. Significantly, they introduce another vertical element to the package, increasing total height. Interposers must also be fabricated on silicon wafers just like active ICs, with their dimensions driven by the footprint of all devices that need to be included in the package. These solutions are typically termed as "2.5D" due to some components being placed laterally with respect to one another rather a true stacking of chips.

    Rather than adopt interposers for its products as a next step in advanced packaging, the direction of Apple's focus, according to several patent applications [1][2][3][4], appears to be on true "3D" techniques, with logic die such as memory being placed directly on top of an active SoC. Additionally, a patent application from TSMC seems to suggest a level of coordination between Apple and TSMC in these efforts.

    [​IMG]
    3D stacking process flow

    The process has similarities to the existing InFO techniques in that they both involve a redistribution layer (RDL) where contacts on a logic die are routed inside a molding compound with the help of vias directly in the molding compound. Where the 3D process diverts from this is that there is now RDL content on both sides of the die, necessitating the use of through-silicon vias (TSV) directly in the logic die so that interconnections can be made with the top of the die. A key feature of these RDL layers is that interconnect pitches finer than available substrate or interposer types is possible.

    Subsequent dies could then be attached to the molding compound, mating with the vias and RDL placed in the previous step. This step could be done multiple times, provided each stacked component has TSVs for the next level of integration, and this is already seen in HBM, which allows for the stacking of up to eight DRAM dies.

    [​IMG]
    Side view of memory die (110) attached to SoC (150) in 3D package

    Still, this approach has many technical challenges that have prevented its commercialization. TSVs are expensive to implement and are a serious yield detractor to ICs. Electrical isolation from nearby components' radiated energy can also be a concern, particularly when integrating RF and analog components in a package with other components that would have been separated by space and EMI shielding before. Apple describes techniques to incorporate shielding directly in the package to mitigate this.

    [​IMG]
    Package with integrated EMI shield

    This approach also presents thermal challenges since active dies become so closely coupled in mediums that have poor thermal conductivity and shared thermal paths. These concerns extend not only to normal device usage, but also the package integration and any solder reflow steps. Thermal stresses can induce warpage of the packaging components due to differing coefficients of thermal expansion (CTE) amongst the materials utilized in the package. This warpage can lead to broken or separated contacts, resulting in device failure.

    The use of a carrier substrate in the process flow mitigates some of the thermal concerns. The direct integration of heatsinks into device packaging is also addressed at various levels of the package assembly, such that higher dissipating die, such as a SoC featuring CPU and GPU cores, could be placed on the bottom of the stack or at a higher level of integration, providing stackup flexibility not seen in previous PoP arrangements.

    [​IMG]
    Package with integrated heatsink component (310)
    The benefits of the methods described are many. The use of higher bandwidth memory will yield performance improvements. The flexibility of component placement shortens the distance between connected active and passive devices, either lowering the energy required to communicate between them, or reducing parasitic effects that can cause unwanted power loss or dynamic performance degradation. The most notable tasks that stand to benefit are gaming and image processing tasks, which often require large amounts of bandwidth over short time intervals.

    Apple Watch Implications

    These enhancements would be applicable to all of Apple's mobile devices, but multiple patent applications specifically mention methods of multiple components married together in a System in Package (SiP), as seen in the current Apple Watch. The methods described below are an enhancement on the existing SiP solutions found in Apple Watch in that they introduce true 3D stacking elements enabled by both TSV and Through Oxide Vias (TOV).

    [​IMG]
    Array of TOVs for connecting stacked die to package pins
    Additionally, the patent describes TSV and TOV pitch in explicit detail, suggesting that keeping package heights down allows them to create very small width vias, with the TOV forming interconnect rows at sizes smaller than even the TSVs. The effect of TSVs stressing active parts of the die, including hurting transistor performance, is also discussed, and the reduced pitches help to mitigate this.

    [​IMG]
    Active die keepout zones around TSVs

    Inclusion of RF transceivers and active devices on substrate types not currently used in Apple mobile devices are covered, indicating all types of active and passive components found in Apple Watch products could be housed in the SiP proposed.

    [​IMG]
    Bottom level view of an SiP with stacked heterogenous die interconnected with TSV and TOV

    Timeline

    Packages featuring 2.5D and 3D connected components have been in consumer devices for several years, but most of the methods described above have yet to debut in mobile devices. The steps described are set to increase manufacturing complexity, and cost and throughput will likely suffer as a result.

    Due to cost and yield concerns, a primary candidate for first inclusion of these methods would be a high-margin, low-quantity device. While the iPhone is the highest margin of Apple's mobile products, it is also the largest volume category, with a huge initial demand for each generation. The iPad Pro is a good candidate because of its low volume nature and its classification as a high-performance device. The inclusion of 120Hz refresh rate is something that will benefit from increased memory bandwidth, specifically.

    The focus of many of these patents seems to be specifically on SiP methods seen in Apple Watch internals. The Apple Watch is a lower-volume device, and it stands to benefit because its internals are extremely sensitive to package size given the importance of its form factor and battery size. It seems reasonable to expect some of the methods described to be incorporated as soon as the next revision of the Apple Watch, and more progressively in future revisions.

    Article Link: Patent Applications Reveal Apple's Research Into 3D Chip Packaging
     
  2. fhall1 macrumors 68040

    fhall1

    Joined:
    Dec 18, 2007
    Location:
    (Central) NY State of mind
    #2
    I'm really surprised these are patentable. I worked in microelectronics for the Gov't since the early 80's and National Semi, Fairchild, TI, and others all had ideas for stacked die, flipped chips, multiple die per package ICs, etc so not sure what's new and novel without digging into it I guess. Manufacturing wasn't up to the task back then and must be they never patented the ideas, or they expired.
     
  3. Ronlap macrumors regular

    Ronlap

    Joined:
    Sep 7, 2007
    Location:
    San Francisco Bay Area
    #3
    Funny how these patents look exactly like the the SanDisk 3D patents from around 2006. Didn't the patent officer notice that?
     
  4. mozumder macrumors 6502a

    Joined:
    Mar 9, 2009
    #4
    OK now add a translucent photovoltaic solar layer, and a micro-LED display layer.
     
  5. slimtastic Suspended

    slimtastic

    Joined:
    May 17, 2018
    Location:
    Your Mother's Bedroom
    #5
    That's cool. I have no idea what I just read. But that's cool.
     
  6. Baymowe335 macrumors 68020

    Joined:
    Oct 6, 2017
    #6
    My guess is you don’t fully understand what you’re looking at.

    Patents can be extremely subtle.
     
  7. FelixDerKater Contributor

    FelixDerKater

    Joined:
    Apr 12, 2002
    Location:
    Nirgendwo in Amerika
    #7
    Keep the R&D focus! That's where the future is.
     
  8. x-evil-x macrumors 68040

    x-evil-x

    Joined:
    Jul 13, 2008
    #8
    way to confused 99% of the readers macrumors
     
  9. Baymowe335 macrumors 68020

    Joined:
    Oct 6, 2017
    #9
    Yet we have people saying this is a copy of another patent and the patent office will make a mistake granting this.
     
  10. chrmjenkins macrumors 603

    chrmjenkins

    Joined:
    Oct 29, 2007
    Location:
    CA
    #10
    Keep in mind these are applications, not granted patents. It’s possible that none are approved. This is more an introspective to the technologies they’re investigating.
     
  11. x-evil-x macrumors 68040

    x-evil-x

    Joined:
    Jul 13, 2008
    #11
    the 1%.
     
  12. Mikey44 macrumors member

    Mikey44

    Joined:
    Mar 6, 2012
    #12
    Just want to say thank you to MacRumors for posting this in-depth article. I appreciate articles that take the time to actually explain the science behind the ideas that have been patented and helps raise the level of discourse here.

    TL;DR I'm glad it's not all just Apple VS. Samsung!
     
  13. CJM macrumors 65816

    CJM

    Joined:
    May 7, 2005
    Location:
    U.K.
    #13
    Excellent write-up, MacRumors. Much appreciated, thanks!

    Anandtech doesn’t really cover Apple any more so it’s nice you’re taking up that particular torch.
     
  14. Relentless Power macrumors Penryn

    Relentless Power

    Joined:
    Jul 12, 2016
    #14
    An extremely diverse and dynamic article, but a very interesting read deep into tech. Good article.
     
  15. Cosmosent macrumors 6502

    Cosmosent

    Joined:
    Apr 20, 2016
    Location:
    La Jolla, CA
    #15
    One of it's primary purposes is to enable AAPL to better-support the the "concurrent capture" of Burst Photos AND Depth Data Maps.
     
  16. macduke macrumors G3

    macduke

    Joined:
    Jun 27, 2007
    Location:
    Central U.S.
    #16
    Thank you for these technical articles, much appreciated! I don't understand all of it, but I understand some and over time I hope reading more articles like this will help raise my level of knowledge and understanding. Before they even got to the Apple Watch section I was thinking "This sounds like it would be really useful in the Apple Watch," lol. I hope the author is right about possibly seeing this in the new Apple Watch this autumn as it could mean some serious speed improvements as well as space for more battery. Combine that with the non-physical buttons rumor and we could get some serious life out of this. The icing on the cake would be them switching to mLED for Apple Watch to get even better battery life but I bet that's still a few years out, right?
     
  17. chrmjenkins macrumors 603

    chrmjenkins

    Joined:
    Oct 29, 2007
    Location:
    CA
    #17
    Yes, micro LED exceeds LCD and OLED technologies in power efficiency, but is probably a few years out.
     
  18. Piggie macrumors G3

    Piggie

    Joined:
    Feb 23, 2010
    #18
    I had to giggle at the bit about the ability with this tech to fit higher capacity batteries.

    Like we all know just how hard Apple try every year to fit higher and higher capacity batteries to keep extending the lifespan.
    Apple would never compromise battery size, simply in an effort to create a thinner product would they ;)
     
  19. caligomez macrumors regular

    caligomez

    Joined:
    Feb 8, 2011
    Location:
    San Juan, PR
  20. name99 macrumors 6502a

    Joined:
    Jun 21, 2004
    #20
    What's being patented is not the vague idea of "wouldn't it be cool to stack chips on top of each other", it's a specific way of actually ACHIEVING that goal.

    Reality is that these patents (I suspect) are not worth much, they're being filed just in case, as legal protection.
    The HARD part is in the trade secrets of how to actually get the zillion steps of the process working, and most of that is either tacit knowledge or (of you're really lucky) written up in some ISO9001 documentation (which is, of course, company internal, not public).
    --- Post Merged, Jun 12, 2018 ---
    Higher bandwidth yes, but also lower power. The closer you can bring DRAM to the SoC, the less power it takes to transfer each bit.
    A second possibility (it will come, but the technology of the patent may not yet be good enough) is the construction of the SoC as two distinct layers, one on top of the other. The primary benefit of this is to reduce the length of some long distance wires. This can be done (has been in the academic literature) in such a way as to avoid placing hotspots on top of each other, so avoiding thermal issues. An easy example (which may actually be the first real implementation of the idea) might be something like moving the L3 cache to a second physical layer, living between the logic at the bottom and the DRAM above. The tricky part is you now need metallization on both sides of the L3...

    There's a great discussion of this technology's past and future here:
    https://community.arm.com/arm-research/b/articles/posts/three-dimensions-in-3dic---part-iii
     
  21. thefourthpope macrumors 6502a

    thefourthpope

    Joined:
    Sep 8, 2007
    Location:
    DelMarVa
    #21
    Came here to say this exact thing.
     
  22. Zenithal macrumors 603

    Joined:
    Sep 10, 2009
    #22
    I understood integrated heatsinks into the chips themselves. That's all I understood.
     
  23. MentalFloss macrumors 65816

    MentalFloss

    Joined:
    Mar 14, 2012
    #23
    Well, the vast majority of consumers obviously strongly prefer thinner and lighter devices over heavier/thicker devices with higher battery capacity. So blame the market, not the manufacturer. Or blame me, as I am one of these consumers.
    --- Post Merged, Jun 13, 2018 ---
    Most. Definitely. Not.

    Whenever there was an article about patents here, all(*) the "this should not be patentable" comments that I have seen so far came from people who had no clue whatsoever what they were talking about. That is always proven already by the fact that these people never read the patents or patent applications in question before commenting, but they just comment on the (usually extremely simplified, frequently completely incorrect) summary on Macrumors. An expert in the field would never do that.

    (*) Yes, "all". Absolutely every single one. No exception so far.
     
  24. Piggie macrumors G3

    Piggie

    Joined:
    Feb 23, 2010
    #24
    How do you know as the consumers have NEVER been given the choice.

    Have Apple make 2 devices, one 7mm with a 1 day battery and the same model, 9mm with a 2 day battery.
    Show me no-one buys the 9mm model and then I'll agree with you.
     
  25. MentalFloss macrumors 65816

    MentalFloss

    Joined:
    Mar 14, 2012
    #25
    The consumers are given a choice. There are battery cases on the market, but I hardly ever see anyone who is willing to add bulk and weight to their phone to get a bigger battery capacity.

    Also, Apple's marketing department does not (despite what many people here seem to think) consist of idiots. They spend large amounts of money to do market research. If there was a real market for devices with high battery capacity among potential Apple customers, then Apple would offer such devices. If it's just 5% of the potential customers who would go for such a phone, then it is not worth investing in the design and production of a higher capacity model.

    Not that you really need to do a lot of market research to figure out that this is not a critical feature. The number of people who go more than a day without having access to a power outlet is extremely small. And those people do have the option to buy battery cases or battery packs for their trip through the Sahara. I actually own a battery pack and didn't even bring it on my last 12 hour long distance flight, because there were USB ports at every seat (which I didn't even need).

    Also, other companies offer phones with high battery capacity. I don't see anyone lining up to buy them. "Hey, why are you so excited about Notappletech's Unbelievaphone Z?" - "Oh well, the killer feature is of course the huge battery life!" ...said no one ever. If enough people were willing to switch to Android for longer standby/usage times, you'd have a constant barrage of Samsung ads, touting their new phone with four days of battery life. But apparently, it's not that critical a feature.
     

Share This Page