Last month Apple introduced its latest generation of iPad Pro tablets, and much to the surprise of many, the new iDevices didn’t come with a high-end variant of Apple’s newest A13 SoC. Instead, the iPads used an SoC that Apple was calling the A12Z, clearly indicating that it was based on the same Vortex/Tempest architecture as the earlier A12X, which was used in the 2018 iPad Pros. The unusual move from Apple left us suspecting that the A12Z may not have even been a new chip, but rather a new bin of the A12X, and today we finally have confirmation of that theory thanks to TechInsights.

In a brief tweet, the technical analysis and reverse engineering firm published a note announcing their findings, along with side-by-side die shots comparing A12Z and A12X. In short, the two chips are seemingly identical, with every last functional block in exactly the same place and the same size on A12Z as it was A12X.

While TechInsights’ die shot analysis doesn’t suss out some finer details such as chip steppings – whether A12Z is even on a newer stepping, or if it’s the same stepping as the A12X that Apple was shipping in 2018 for the IPad Pro launch – it’s clear that in terms of silicon, A12Z doesn’t bring anything new to the table.

Instead, the notable changes between the two chips is in their binning/configuration: whereas the A12X only ever shipped with 7 of its GPU clusters enabled, A12Z ships with all 8 enabled. And while a small change in the grand picture of things, it makes sense for Apple to finally enable the 8th cluster for a bit more performance. A12X is produced on TSMC’s 7nm line, and when it was released in 2018, it was one of the biggest 7nm chips being churned out. So Apple should be enjoying much better yields 18 months later, reducing the need to bin to a lower spec to salvage chips.

Apple SoC Comparison
  A12Z A12X A13 A12
CPU 4x Apple Vortex
4x Apple Tempest
4x Apple Vortex
4x Apple Tempest
2x Apple Lightning
4x Apple Thunder
2x Apple Vortex
4x Apple Tempest
GPU 8-cluster, A12-gen 7-cluster, A12-gen
(+ 1 disabled)
4-cluster, A13-gen 4-cluster, A12-gen
Memory Bus 128-bit LPDDR4X 128-bit LPDDR4X 64-bit LPDDR4X 64-bit LPDDR4X
Manufacturing Process TSMC 7nm (N7) TSMC 7nm (N7) TSMC 7nm (N7P) TSMC 7nm (N7)

As for why Apple would opt to re-use A12X for their 2020 tablets instead of commissioning an A13X, while we can only speculate, it almost certainly comes down to economics, as the tablet market is quite different from the smartphone market. Apple is virtually unchallenged as far as high-performance Arm tablets go, and even then, the number of iPads they sell has always been a drop in the bucket compared to the number of iPhones they sell. So there are fewer devices to amortize the costs of chip development against, and all the while chip development costs are continuing to rise with each new generation of photolithography technology. In short, at some point it has to stop making sense to create new chip designs on a yearly basis for mid-volume products, and Apple may very well have finally hit that mark with their tablet SoCs.

Source: TechInsights

POST A COMMENT

57 Comments

View All Comments

  • kefkiroth - Monday, April 13, 2020 - link

    There must be a huge bump in refurb 2018 iPad Pro orders these days. Most people won't use the LIDAR capabilities, and the A12X is for most purposes almost identical to the A12Z. The time the 2018 models gets slow should be around the time the 2020 models feel slow. I recently picked one up after holding out with an old Nexus 9, and in some respects, it feels snappier than my 11 Pro. Reply
  • André - Tuesday, April 14, 2020 - link

    The only difference is the amount of RAM available. The 2018 model only the 1TB model has 6GB of RAM. Now every 2020 model has 6GB. Reply
  • s.yu - Tuesday, April 14, 2020 - link

    They need to sell the 2018 iPP refurbs cheaply enough. I don't know about the current market price but the 2018 model was priced higher than the 2020 models at launch so they need a more significant cut to make it make sense. Some may also appreciate the upgraded camera. Reply
  • brucethemoose - Monday, April 13, 2020 - link

    Maybe Apple had some of those tablet SoC engineers working on a ARM Macbook chip?

    Given the slip ups in the x86 market over the past few years, it would've been the perfect time to pull the trigger. I'd be *very* interested to see how a couple of thier ARM cores stack up against Skylake and Zen in the same TDP.
    Reply
  • watzupken - Monday, April 13, 2020 - link

    Its a shoddy iPad Pro refresh to be honest. If they want to release a new iPad, then make it a proper upgrade instead of just slapping some fancy rear facing camera array. Apple is increasingly recycling design, components, etc, over the years. For a premium price that they are charging, there is little reason for these to stay stale, especially when this is the top end tablet. Reply
  • zepi - Tuesday, April 14, 2020 - link

    New pro is still better than the old one - better camera + more memory/SSD and prices stayed the same.

    Personally this will not make want to buy one, but it is not like anyone is forced to upgrade or entitled to bigger perf upgrades.

    Compare with your average yearly update of HP, Lenovo or Dell laptops. Nobody expects to upgrade their business laptop on a yearly or bi-yearly cadence. I think most larger companies have 36-48 month upgrade cycles.

    My guess is that iPad Pro aims for similar. If they release a small feature upgrade every 18 months and a new SoC + other things every 36 months, businesses won't complain.
    Reply
  • s.yu - Tuesday, April 14, 2020 - link

    Cheaper, for mid-high tiers. The base models stay the same though. Reply
  • rpg1966 - Monday, April 13, 2020 - link

    It's (mildly) interesting, in that Apple has so much money they could just say "who cares if we spend a $xx million on an A13X", and yet they take what appears to be the more financially-sensible route.

    Yes, of course there's the "look after your pennies and the pounds will look after themselves" thing, but when you have *that* much money, it's still a (slightly) interesting insight into their thought process.
    Reply
  • brucethemoose - Monday, April 13, 2020 - link

    Its reportedly more like $xxx million, and growing: https://semiengineering.com/big-trouble-at-3nm/

    Thats enough to eat into *Apple's* margins. And given those eye watering projections, this won't be the last shortcut they take.
    Reply
  • flgt - Tuesday, April 14, 2020 - link

    Interesting reading that article from 2018. Now GF’s is out. Intel might be next if their next node can’t match 14 nm in performance. Maybe only TSMC and Samsung left. Although the government may need to step in and help Intel if they want to keep any leading edge fabs in the US. Reply

Log in

Don't have an account? Sign up now