iPhone 13 A15 benchmarks reveal 21% CPU speed gain over iPhone 12

2»

Comments

  • Reply 21 of 33
    It's odd Apple did not call out a 20% speed increase during the launch presentation. If Intel or AMD managed a 20% CPU performance increase that's all you would hear about for weeks. Instead they talk up their low single digit performance increases.
    MisterKit
  • Reply 22 of 33

    However worth noting is that there has been significant press stating that the A15 would only be a mild bump, and that Apple has "hit a performance wall" - This is interesting because it's not just wrong, but appears to be a wide effort at astroturfing.
    Is it astroturf, or just crabgrass? 
    tmaywatto_cobra
  • Reply 23 of 33
    It's odd Apple did not call out a 20% speed increase during the launch presentation. If Intel or AMD managed a 20% CPU performance increase that's all you would hear about for weeks. Instead they talk up their low single digit performance increases.
    We are yet to see AnandTech's detailed performance analysis of A15. So any judgements on performance of A15 need to wait for at least few weeks. Also, performance of A series processors has gone way ahead of the software advancements in the last 5 years. There is no need to focus much on raw CPU performance unless new use cases come up in software which need more than the raw power of the A series SoCs (mainstream ones, not the niche ones like high-end gaming). If Apple is able to improve them further, great. If not, so be it. Apple's customers are not going to lose anything, with other features being rolled out every year.
    edited September 16 tmay
  • Reply 24 of 33
    ppietra said:
    frantisek said:
    KITA said:
    "Single Core Score: 1728 (+7.6% over A14)
    Multi Core Score: 4695 (+9.3% over A14)"

    A14 @ 2.99 GHz
    A15 @ 3.22 GHz
    = +7.33%

    M1 @ 3.2 GHz Single Core: 1721 (iPad Pro https://browser.geekbench.com/v5/cpu/9837998)
    +0.4%

    M1 @ 3.2 GHz
    A15 @ 3.22 GHz
    = 0.6%

    Conclusion, the difference mainly boils down to an increase in frequency and presumably the bigger cache in multi core score.


    Insider info leaked long time ago said that A15 design will be focused on power efficiency. It looks like that . They made it less hungry, probably cooler, better manufacturing process then allowed them to hick frequency and so the performance.  When we see proclaimed battery life improvements we can say that A15 give more juice for the same power. We will see what Anand tech will find up.
    It is very likely made on TSMC's newer and improved N5P node, if so the better thermal efficiency is likely free lunch. Meaning Apple didn't put much/any work into it, but is profitting from TSMC's node progress.
    TSMC reported that N5P node would only make a 5% increase in performance at isopower, so clearly Apple found anoother place to get some extra performance.

    What extra performance are you talking about. As I stated, there is no extra single core performance over A14, except due to higher frequency. M1 is running A14 cores closer to A15 speed and is scoring virtually the same in Single core performance as A15. Likely A15 is running at 3.23GHz within the same thermal envolope as A14 is at 2.99GHz, due to improvements from N5P node.


    williamlondonradarthekat
  • Reply 25 of 33
    Will anyone really notice the speed increase in day to day usage?
    watto_cobra
  • Reply 26 of 33
    tedz98 said:
    How much performance do people want out of a handheld computer? How much performance is actually needed? Unless Apple significantly increases the number of megapixels in their camera I’d argue you don’t need any more performance than what the 13 Pro offers. The challenge Apple has is they need to continuously offer “compelling” reasons to justify the premium price of the iPhone. Android has a superior market share. But Apple has an incredible share of the market measured in dollars and average price per device sold. The performance numbers are more important with the M series of laptop and desktop CPUs.
    As if megapixels were the only worthwhile metric in measuring a camera. You somehow miss the Cinematic Mode presentation or introduction of ProRes shooting?
    spock1234radarthekatwatto_cobra
  • Reply 27 of 33
    Fred257 said:
    And yet look at the speed tests done with Android phones side by side.  All that power is not being utilized in the least bit.  Apple I’m very disappointed in you.  I’m thinking about selling my iPhone 12 mini and going back to my first generation iPhone SE.  Software runs at about the same speed and that’s five generations older.  

    Apple, please optimize your software as my newest phone feels and acts just as slow as my older phone.  Shame on you 😂
    Unadulterated horseshit. 
    williamlondonthttmayStrangeDayswatto_cobra
  • Reply 28 of 33
    Will anyone really notice the speed increase in day to day usage?
    Well depends - doing things with less energy improves how long you can do those things for.
    I think that could be why Apple moved to battery life improvements as that a measure because that is real world.
    But actual CPU performance improvements at this point are probably not seen outside academic tests or on larger devices. 

    If they can say for the next 5+ years  "We still have the most powerful SOC by miles and we increased battery life again" then that could use some gains to move the budgets towards things consumers will get a kick out of. 

    I think that is the real reason they have de-emphasised CPU metrics to concentrate on what it delivers Cameras and Battery.
    fastasleepwatto_cobra
  • Reply 29 of 33
    mattinoz said:
    Will anyone really notice the speed increase in day to day usage?
    Well depends - doing things with less energy improves how long you can do those things for.
    I think that could be why Apple moved to battery life improvements as that a measure because that is real world.
    But actual CPU performance improvements at this point are probably not seen outside academic tests or on larger devices. 

    If they can say for the next 5+ years  "We still have the most powerful SOC by miles and we increased battery life again" then that could use some gains to move the budgets towards things consumers will get a kick out of. 

    I think that is the real reason they have de-emphasised CPU metrics to concentrate on what it delivers Cameras and Battery.
    To add to your analysis, consider the context that Intel fouled up with the Pentium 4 towards trying to focus on performance, and truly whiffed it for power usage, they thought they could ramp up that design pattern to a 10 Ghz clock.  I suspect they could have eventually gotten there, but it'd require a cryogenic freezer around the CPU to make that work.  Intel's Israel team was working on a low-power Pentium processor that had a shorter pipeline, which also was a meaningful part of how it used less power.  Intel started from that as the basis, with the design goal of going for performance (not sure how perfectly they followed this) where it'd save power in the process.

    The net result is that Intel has still created CPUs that need a lot of cooling and use a bit of power, but they've attached rocket engines to a pig of a system and made that pig fly by a long string of incremental performance tweaks that takes power into account as a major design priority. Why?  If you have "fast" circuitry but it uses too much power, it'll overheat anyway, and you'll limit your upper range of what's feasible purely by physics.

    So, focusing on ways to reduce power usage for a given amount of computation translates long-term into headroom for improving performance while keeping the power/cooling costs reasonable.  By contrast, look how many Android phones can burst for a short period of time, but also need huge batteries (not that it would be a bad thing if Apple made larger batteries, but... that's not how Apple rolls historically) and will run hot.  It's not like an iPhone won't run rather warm/hot if you push it hard enough.

    As an owner of an M1 MacMini, in an apartment in the Seattle area without AC (it's rather odd to have an apartment or house in this area with central AC, this is a curious area between the climate and the culture in that regard) I welcome working on a system that doesn't add notable waste heat into my environment.
    mattinozradarthekatwatto_cobra
  • Reply 30 of 33
    Fred257 said:
    And yet look at the speed tests done with Android phones side by side.  All that power is not being utilized in the least bit.  Apple I’m very disappointed in you.  I’m thinking about selling my iPhone 12 mini and going back to my first generation iPhone SE.  Software runs at about the same speed and that’s five generations older.  

    Apple, please optimize your software as my newest phone feels and acts just as slow as my older phone.  Shame on you 😂
    Don’t do drugs, kids.
    tmaywilliamlondonwatto_cobrafastasleep
  • Reply 31 of 33
    This benchmarks conflicts with so called experts said iPhone 13 uses pretty much the same processor as iPhone 12. Apparently, iPhone 13 offers lot more improvements over iPhone 12.
    williamlondonwatto_cobra
  • Reply 32 of 33
    sdw2001sdw2001 Posts: 17,738member
    Cool.  Now, does the damn screen dim under load and/or heat so that amazing CPU performance is often useless?  Asking for a friend.  
    williamlondonwatto_cobra
  • Reply 33 of 33
    mcdavemcdave Posts: 1,826member
    Well that +3.2billion transistor budget (only 1bn shy of the M1) was spent somewhere and +1 GPU core doesn’t explain it. The +4 TOP neutral performance is a clue and it isn’t just the Neural Engine that’s improved. The Geekbench ML scores show what’s happened; the NE up is up 10% but the GPU score is the main bump at 40%
    Expect this on the M1X as I don’t think Apple will spend its transistor budget completely on the CPU or GPU cores. It’ll be audio, video codecs & layers that will receive the love, benchmarks be damned.
    edited September 20 williamlondonwatto_cobrafastasleep
Sign In or Register to comment.