When theiPhone 16 linelaunches this September , we expect it to usher in a new A - series mainframe . As has been the case in the last few year , the top - end “ Pro ” model will get a raw processor while the regular iPhone 16 non - Pro model will likely get a version of the A17 Pro flake found in this year ’s iPhone 15 Pro .
As we do every class , we ’ll look at the history and performance of the A - series processors from the last several year , along with the recent rumors and overall industry trends , to look for what we might expect from Apple ’s unexampled iPhone silicon this fall . As always , this is all just guesswork ; Apple is a magnificently closelipped company , and the item about its product are often not fully reveal until they are relinquish .
To “Pro” or not to “Pro”?
Last year , Apple introduce what might be an all - young wrinkle into the expected upgrade radiation diagram . exchange the name from “ A17 Bionic ” to “ A17 Pro ” for the first time , the logical implication is that there would be a Modern non - Pro chip each generation too . That has n’t yet occur to pass , but we might see something different this year .
There are three probable possibilities for what Apple may do in plus to wee a new A18 Pro for the iPhone 16 Pro and Pro Max :
It ’s hard to say which is more likely , only that the high-pitched - remnant chip will probably once again carry the “ Pro ” name instead of “ Bionic ” and launch only in the iPhone 16 Pro models . We ’ve heard rumors that thelower - death chip in the iPhone 16 will be called “ A18 ” this yr , but even if that ’s dependable we do n’t know if it will be an truly newfangled splintering or just a rebranded iPhone 17 Pro . And we discover similar rumors last yr that did n’t pan out .
Our analysis here is targeted at the A18 Pro , what we take over Apple will call the higher - end adaptation of the chip powering this year ’s iPhones .
Still made with a 3nm process
The A17 Pro was the first mass - market consumer chip to apply a 3 nm fabrication process . And whileTSMC is on runway with the 2 micromillimeter chipsthat arrive next , we do n’t expect to see an Apple product using that outgrowth until next year a the earliest . So the A18 Pro will be made with TSMC ’s 3 nm cognitive process again .
The 3 nm procedure used for the A17 Pro was known as N3 , and as the first looping of 3 millimicron applied science from TSMC it was pretty expensive . The company has a newer process known as N3E , which is far easier to manufacture at scale . It should be slightly more superpower efficient than the original N3 process but also slightly less “ dense ” , which means a chip build using N3E would be slightly larger than an identical chip made with N3 .
There are rumour that the A18 Pro will be a physically larger chip than the A17 Pro — the switch to N3E may be one ground , but we expect Apple will also increase transistor reckoning to more then 20 billion ( A17 Pro is a 19 billion electronic transistor chip ) . To put that in linear perspective , a modern Intel “ Raptor Lake ” generation laptop CPU is estimated to have a junction transistor count around 26 billion transistors , which the top - tier large Nvidia GeForce 4090 RTX background GPU is a 74 billion junction transistor chip .
CPU performance will probably improve
All the rumors have focused on AI performance , and why not when that ’s the hot topic of the Clarence Shepard Day Jr. ? But Apple is always concerned about CPU performance , and shares core designs between its A - series iPhone / iPad chips and it ’s M - series Mac ( and iPad Pro ) chips .
Just look at this chart of individual - center Geekbench 6 performance for iPhone scrap all the way back to the A11 .
Foundry
What strike me as most remarkable is just howsteadilysingle - kernel public presentation has increase . We ’ve judge the A18 performance by continuing the vogue from the last several chips , and it gives us a single - core Geekbench 6 score of almost 3,200 . That ’s just a touch faster than an Intel Core i9 - 13900KS , a very high - end desktop CPU .
Of course , herculean desktop and laptop computer CPUs have many more cores than a telephone set central processor . We think the A18 Pro will have the same core configuration as most late A - serial chip : two carrying out center and four efficiency burden .
Faster core , cache , and memory board will agitate up the multi - core mark even if Apple uses the same number of cores . We ’ve seen a pretty steady carrying into action trend for multi - core public presentation as well , and do n’t expect that to dead change . A multi - core score of over 8,200 would put the A18 Pro in the same territory as a mid - tier up Intel or AMD laptop CPU from about three yr ago .
GPU performance always goes up
Using the modernistic 3DMark Wild Life bench mark in Unlimited mode , we get a good melodic theme of how much GPU performance has improved over time . Frame rates have more than quadruple since the iPhone X days . It ’s not quite as unfaltering as CPU public presentation – some years the uplift is 10 percent or less and in others long time it ’s closer to 30 percent .
We do n’t have any reason to believe that GPU performance wo n’t be well this clip around as well , though the rumors say Apple is n’t tally more GPU cores . Architectural efficiency can go a foresighted way . Our guess is a 10 - 15 percent improvement in substantial - humankind central processor performance .
The 3DMark Solar Bay test stress irradiation tracing carrying out , so you see much lower framing rate overall .
Performance took ahugejump up with the A17 Pro in this exam because Apple added computer hardware to help accelerate ray tracing . There may be some tweak this year , but we do n’t think a similar jumping can be expected in this test in the A18 Pro – again , we ’re probably get going to see a 10 - 15 pct encouragement .
Neural Engine may get a big upgrade
The Neural Engine is Apple ’s name for it ’s NPU or “ Neural Processor Unit”—specialized hardware made to optimally run neural networks for machine learning , deep learning , and AI software in the same way the GPU is made for graphic processing .
Apple is suppose to be all - in on AI this year , with heavy AI feature come to iOS 18 and the iPhone 16 line . I think that means we can expect a expectant improvement in Neural Engine public presentation .
But the Neural Engine has been produce quicker at a regular pace since it was first introduced in the A11 ( found in the iPhone 8 and iPhone X ) . That first Neural Engine could perform 600 billion operations per irregular , while the Neural Engine in the A17 Pro is touted as able to handle 35 trillion operations per second ( TOPS ) .
Here ’s how that performance increase over time has manifested in operation using the new GeekBench ML bench mark , place the NPU only .
Geekbench MLis a relatively fresh bench mark , only on version 0.6 , that runs a suite of machine learning task like image credit , object classification , persona super resolution , and linguistic communication processing to assess the operation of CPUs , GPUs , and NPUs . When running the test only the the Neural Engine , we see a more than 8 - fold improvement over the last six years , and an medium annual speedup of rough 20 % .
Notably , the score did not double from the A16 to A17 propagation even though Apple said Neural Engine performance jump from 17 acme to 35 TOPS . The claimed maximum performance specs often do n’t match real - world performance numbers .
Apple ’s Neural Engine already delivers much good performance than competitors like the NPU in the Snapdragon 8 + gen 1 ( around 2,800 point depending on which telephone it ’s in ) or Google ’s Tensor G3 ( under 2,400 points ) . just take the retiring public presentation gains and projecting it ahead would give us a score of almost 4,100 points and a commanding confidential information over most competitor .
I think we can safely push aside this chart for the A18 Pro , however . late rumor intimate that Apple will devote more Si to the Neural Engine and fight for a big performance encouragement in edict to power more advance AI features entirely on - twist . I would n’t be surprised if Apple touted a really big number for its Neural Engine this yr – over 50 TOPS is almost certain — and we could see a Geekbench ML score closer to 5,000 .
RAM, 5G, and more
Of course , there ’s more to the A18 Pro than the CPU , GPU , and Neural Engine . Then there are related chips that are n’t technically part of the A18 Pro but are still authoritative to iPhone users , like the cellular modem .
The newfangled LPDDR6 memory standard is still too newfangled to make it into this yr ’s iPhones , but Apple could move from LPDDR5 to LPDDR5x — providing for more or less more memory bandwidth and while using less power . The A17 Pro made the leap from 6 GB to 8 GB of RAM and it feels too presently for Apple to jump up again . On the other handwriting , AI manikin run to be very store - intensive and we could see Apple increase random-access memory just to enable bigger and more advanced AI model on the new phones .
On the wireless front , we should anticipate an rise to Qualcomm ’s Snapdragon X75 modem for fast and more reliable 5 G connections . Apple has been work on its own wireless microchip for year but is having a hard time getting them to a point where they supply a in force enough experience . There have also been rumors of anupgrade to Wi - Fi 7 , and of grade , we can expect Ultra Wideband , Bluetooth , and NFC lineament that are like to those in the iPhone 15 line .
