Apple A11 SoC

Discussion in 'Mobile Devices and SoCs' started by iMacmatician, Aug 18, 2017.

  1. roninja

    Regular

    Joined:
    Feb 9, 2002
    Messages:
    268
    Likes Received:
    0
    I suspect they’ve taken the basic Rogue 7XT schema and further customised it. It’s a bit like those customised BMWs which come with another motor Co. badge on them.
     
  2. mfaisalkemal

    Newcomer

    Joined:
    May 26, 2017
    Messages:
    48
    Likes Received:
    24
    with improvement of ALU 2 test(GFLOPS) and texturing test(Fill Rate) iPhone 7 Plus vs iPhone 8 Plus i think not far from this Furian(6 Cluster) Improvement claim by Imgtec.
    [​IMG]
     
    #102 mfaisalkemal, Sep 26, 2017
    Last edited: Sep 26, 2017
  3. Entropy

    Veteran

    Joined:
    Feb 8, 2002
    Messages:
    3,050
    Likes Received:
    1,011
    True, but ”fill rate is in the same neighbourhood as we might have expected from a Furian marketing slide. Kinda.” is an awfully broad data point.
    We would need a few specific instances of Furian vs. Rogue design changes (and a means to test them) to actually build a case.
     
  4. mfaisalkemal

    Newcomer

    Joined:
    May 26, 2017
    Messages:
    48
    Likes Received:
    24
    Tech Insight A11 Die Shot
    [​IMG]
     
  5. Pressure

    Veteran Regular

    Joined:
    Mar 30, 2004
    Messages:
    1,329
    Likes Received:
    261
    Why would it have HDMI? Wouldn't it be eDP if anything?
     
  6. loekf

    Regular

    Joined:
    Jun 29, 2003
    Messages:
    613
    Likes Received:
    61
    Location:
    Nijmegen, The Netherlands
    Yeah, should be eDP or DP. See the A10X teardown at ifixit. In an AppleTV they just hook up the SoC to an external DP to HMDI TX IC.

    Bit surprised there is an USB PHY on it as well, with its 5V I/O.

    And from the Ifixit teardown, might be an PCIe lane somewhere as well...
     
  7. Grall

    Grall Invisible Member
    Legend

    Joined:
    Apr 14, 2002
    Messages:
    10,801
    Likes Received:
    2,171
    Location:
    La-la land
    Lightning is just repackaged USB...

    PCIe has been used for the single-chip SSD Apple's been using from iPhone 6 onwards.
     
  8. Ailuros

    Ailuros Epsilon plus three
    Legend Subscriber

    Joined:
    Feb 7, 2002
    Messages:
    9,416
    Likes Received:
    178
    Location:
    Chania
    Aren't those IMG claimed persentages cluster for cluster at the same frequency between 7XT and 8XT generations?

    Either way again with the subnote that these are preliminary results for the iPhone 8 Plus:

    iPhone8 Plus:
    Battery Lifetime Manhattan3.1: 181 min
    Long term performance Manhattan 3.1: 43.9 fps

    iPhone7 Plus:
    Battery Lifetime Manhattan3.1: 301 min
    Long term performance Manhattan 3.1: 40.3 fps

    Latter results are run onscreen and I double checked both devices have a 1080p resolution and therefore comparable. Similar average performance after 30 continious runs of Manhattan3.1 for both devices whereby the 7 Plus seems to last a lot longer. I know they are merely indications, but I consider those indications far more interesting than raw peak low level test results.
     
    Squilliam, roninja and Grall like this.
  9. mfaisalkemal

    Newcomer

    Joined:
    May 26, 2017
    Messages:
    48
    Likes Received:
    24
    Those claimed compared with Series7XT Plus in the same process technology and at same clock frequency for cores with similiar silicon area budget. because the same cluster differ 50% in raw GFLOPs, Series8XT has larger area 10/9 (1.11111x) than Series7XT.

    your data is wrong because of different times when taking. below the correct data.
    #Iphone 7 Plus
    Long term performance Manhattan 3.1: 40.3 fps
    [​IMG]
    Battery Lifetime Manhattan 3.1: 74 min
    [​IMG]
    #Iphone 8 Plus
    Long term performance Manhattan 3.1: 43.9 fps
    [​IMG]
    Battery Lifetime Manhattan 3.1: 165 min
    [​IMG]

    For best Long term performance and Battery Lifetime Manhattan 3.1 we have to wait data from anandtech because data is taken with all displays are set to 200 nits, with location and other background services disabled to reach a useful relative comparison condition and battery from 100% to 0%.
    examples:
    [​IMG]
    [​IMG]
    [​IMG]
     
    Squilliam likes this.
  10. Ailuros

    Ailuros Epsilon plus three
    Legend Subscriber

    Joined:
    Feb 7, 2002
    Messages:
    9,416
    Likes Received:
    178
    Location:
    Chania
    Kishonti's suite cuts averages when you seek for it but that's how it exactly handles fillrates and all other results also. Obviously the iPhone7 Plus isn't lasting "just" 74 minutes battery life (estimated) while running Manhattan3.1 where Anand's graph itself shows 3.16 hours.

    The point here is that data isn't selectively correct or wrong as displayed in Kishonti's database, it's rather highly indicative since there are several pitfalls to watch over and that goes for ALL results.

    In any case if the early A11 Gfxbench results aren't as misleading it COULD be that the iPhone8 Plus throttles more than former iPhones, which COULD mean that they mostly raised frequencies compared to the A10 GPU. Nothing has been stated or claimed in absolutes.
     
    BRiT likes this.
  11. Pressure

    Veteran Regular

    Joined:
    Mar 30, 2004
    Messages:
    1,329
    Likes Received:
    261
    Any benchmarks we are missing from the A11?
     
  12. iMacmatician

    Regular

    Joined:
    Jul 24, 2010
    Messages:
    766
    Likes Received:
    200
    SPEC 2000 and 2006?
     
    Lodix likes this.
  13. Pressure

    Veteran Regular

    Joined:
    Mar 30, 2004
    Messages:
    1,329
    Likes Received:
    261
    Anyone compiled a version for iOS?
     
  14. Voxilla

    Regular

    Joined:
    Jun 23, 2007
    Messages:
    708
    Likes Received:
    279
    What is the clock frequency of the A11 (big / little cores) ?
    Still can't find this information anywhere.
     
    Lodix likes this.
  15. Vitaly Vidmirov

    Newcomer

    Joined:
    Jul 9, 2007
    Messages:
    108
    Likes Received:
    10
    Location:
    Russia
    I hacked up a little app to get actual CPU frequency for all cores
    https://github.com/vvid/ios-cpu-tool

    I hope someone with iPhone8 could compile and run it.
    Just set 1..6 to "Threads" field and tap "Get CPU frequency" several times.
    In theory there should be 2 different frequencies of Mistral/Monsoon clusters.

    For example on my iPhone SE I got
    1846MHz - one thread
    1799MHz - two threads
     
  16. Pressure

    Veteran Regular

    Joined:
    Mar 30, 2004
    Messages:
    1,329
    Likes Received:
    261
    iPhone 8
    Core 1: 2186 MHz
    Core 2: 2303 MHz
    Core 3: 2234 MHz
    Core 4: 2303 MHz
    Core 5: 2303 MHz
    Core 6: 2302 MHz

    Second run
    Core 1: 2303 MHz
    Core 2: 2303 MHz
    Core 3: 2375 MHz
    Core 4: 2303 MHz
    Core 5: 2303 MHz
    Core 6: 2302 MHz

    Third run
    Core 1: 2304 MHz
    Core 2: 2303 MHz
    Core 3: 2372 MHz
    Core 4: 2303 MHz
    Core 5: 2304 MHz
    Core 6: 2304 MHz

    iPhone 7
    Core 1: 2242 MHz
    Core 2: 2239 MHz
    Core 3: 2240 MHz
    Core 4: 2242 MHz
     
    #116 Pressure, Oct 13, 2017
    Last edited: Oct 13, 2017
  17. Voxilla

    Regular

    Joined:
    Jun 23, 2007
    Messages:
    708
    Likes Received:
    279
    So all cores around 2.3 Ghz. (iPhone 7 is officially 2.34 Ghz)
    Would have expected the small cores to run slower.
    So given the claimed 25% faster CPU, that is again a huge increase in IPC. Must be well over Intel's best now.
    Can't imagine removing ARM 32-bit has enabled this.
     
    Grall likes this.
  18. Nebuchadnezzar

    Legend

    Joined:
    Feb 10, 2002
    Messages:
    972
    Likes Received:
    138
    Location:
    Luxembourg
    That app needs to fire up constantly in a power virus fashion on all CPUs otherwise you don't guarantee that the thread is actually running on a different CPU, so I don't think it works with that DispatchQueue.

    On the iPhone 7 you only have 2 logical CPUs so again with 4 threads we should be seeing about 1100MHz for the two cores (due to time-slicing), so somethings broken with the measurement.
     
    #118 Nebuchadnezzar, Oct 13, 2017
    Last edited: Oct 13, 2017
    mfaisalkemal and BRiT like this.
  19. Pressure

    Veteran Regular

    Joined:
    Mar 30, 2004
    Messages:
    1,329
    Likes Received:
    261
    Admittedly I didn’t look through the code.

    I simply compiled it and ran it.
     
  20. Nebuchadnezzar

    Legend

    Joined:
    Feb 10, 2002
    Messages:
    972
    Likes Received:
    138
    Location:
    Luxembourg
    It just needs to be modified and run those threads continuously while you have another thread sample the outputs occasionally.
     
Loading...

Share This Page

  • About Us

    Beyond3D has been around for over a decade and prides itself on being the best place on the web for in-depth, technically-driven discussion and analysis of 3D graphics hardware. If you love pixels and transistors, you've come to the right place!

    Beyond3D is proudly published by GPU Tools Ltd.
Loading...