New iPad Mini Has Downclocked A15 Chip Compared to iPhone 13

[theorist9]

Why has Apple only binned AS based on the no. of GPU cores and the CPU clock, but not the no. of CPU cores? Here I speculate on three possibilities:

1) Their use of both efficiency and performance cores requires them to maintain, for each chip design, a fixed ratio of the two. A chip with one missing core of either would be unbalanced and perform poorly.

2) Even if their chips worked properly with a missing CPU core (as they do with a missing GPU core), the full CPU core count is still needed to meet certain minimum performance characterstics.[E.g., perhaps Apple decided the performance hit from reducing the GPU core count in the Air by one is OK, while that from reducing the CPU core count by one is not.]

3) Because of the nature of how AS chips are produced, bad CPU cores are much less common than bad GPU cores, such that the former does not occur with sufficient frequency to justify creating a separate SKU for a lower-CPU-core-count device.

 

[Solomani]

Question: if Apple underclocks a device like the iPad Mini 6. Could Apple also "undo" that in a future firmware update?

 

[Shirasaki]

Question: if Apple underclocks a device like the iPad Mini 6. Could Apple also "undo" that in a future firmware update?

They might be able to, but I’d argue they don’t want to do that. What if said update bricks a lot of iPad mini 6 smh or cause random crashes? “Overclocking” a subpar chip should not be something that is performed by manufacturers unless they want to market as such. Silicon lottery exist for a reason.

 

[Marty80]

Compared to the previous iPad mini even if the A15 is underclocked it still is a huge leap in terms of:

- Physical redesign
- Screen realestate ( smaller bazels, Liquid Retina display)
- Smaller in height when compared to the outgoing model, much easier to hold.
- Better processor & gpu performance when compared to the outgoing.
- More Ram compared to the outgoing model
- Better Camera resolution
- Flash for the first time.
- USB C
- Apple Pencil 2

These factors make this the best update we had for the IPad Mini in years, and outweigh the downclocked A15 chip.

My only grip is the cost, but it was expected with the chip shortage.

 

[Codpeace]

Yes, you are completely off. The way I explained it is exactly how it works. You have critical paths. The critical paths vary in length based on slight variations in geometries or doping of semiconductor structures or layers. Your maximum clock frequency is 1 divided by the worst critical path.

So a critical path, in this case, is literally a (set of) designated physical path(s) a signal takes through the chip for testing purposes, right? Then your maximum clock freq is 1 divided by the resulting completion speed(s) for each chip undergoing QC testing post-manufacture.

I'd like to hear more about the variabilities leading to differing results among sibling chips off one wafer (or whatever the terminology is; I'm ignorant). You state it is slight variations in geometries, which sounds like "organic" variabilities arising inevitably in the manufacturing process -- mostly uncontrollable by better engineering at the current technological state-of-the-art. Things like variation in crystal structure, etc. (e.g. "geometries"), leading to a bell curve of critical path results that are inherently individually unpredictable but expected in an aggregated set of chips.

But you also mentioned doping of structures or layers. That IS within engineering control, no? Doping is intentional with predictable, desired results, isn't it? How could doping not be predictably pre-factored into the results, i.e. an intentionally engineered outcome, and therefore be a factor in variability of output that necessitates binning? In other words, variability that requires quality-binning wouldn't be a result of doping, would it?

I'm not trying to hold your feet to the fire, I'm just trying to understand more. It could be you were just typing up a response on a board and not trying to be super-precise with language, which I get.

I would think that, from what I've learned today, Apple sends the A15 to TSMC expecting a pretty standard bell curve yield of, say, 1m chips. All good ones work for the iPad Mini, most for the iPhone 13, and enough for the iPhone 13 Pro. Then they've got a set of rejects that can be used for whatever else or will be trashed. Roughly accurate?

 

[Eagle1001]

I have Apple iPad Pro 10.5 (2017) buy iPad mini 6 on first sight. Ideal for navigation/bed/travel. Bit hesitant to 256GB but trade it for 5G. Hope navigation will work via hotspot from iPhone.

 

[cmaier]

So a critical path, in this case, is literally a (set of) designated physical path(s) a signal takes through the chip for testing purposes, right? Then your maximum clock freq is 1 divided by the resulting completion speed(s) for each chip undergoing QC testing post-manufacture.

I'd like to hear more about the variabilities leading to differing results among sibling chips off one wafer (or whatever the terminology is; I'm ignorant). You state it is slight variations in geometries, which sounds like "organic" variabilities arising inevitably in the manufacturing process -- mostly uncontrollable by better engineering at the current technological state-of-the-art. Things like variation in crystal structure, etc. (e.g. "geometries"), leading to a bell curve of critical path results that are inherently individually unpredictable but expected in an aggregated set of chips.

But you also mentioned doping of structures or layers. That IS within engineering control, no? Doping is intentional with predictable, desired results, isn't it? How could doping not be predictably pre-factored into the results, i.e. an intentionally engineered outcome, and therefore be a factor in variability of output that necessitates binning? In other words, variability that requires quality-binning wouldn't be a result of doping, would it?

I'm not trying to hold your feet to the fire, I'm just trying to understand more. It could be you were just typing up a response on a board and not trying to be super-precise with language, which I get.

I would think that, from what I've learned today, Apple sends the A15 to TSMC expecting a pretty standard bell curve yield of, say, 1m chips. All good ones work for the iPad Mini, most for the iPhone 13, and enough for the iPhone 13 Pro. Then they've got a set of rejects that can be used for whatever else or will be trashed. Roughly accurate?

Just like anything else, doping has manufacturing tolerances. Layer thicknesses too. The individual polygonal shapes that make up each wire, transistor, via between metal layers, etc. There are hundreds of billions of shapes. Each layer has to align with the one below, and there can be slight variances. Everything has a tolerance. The wafer is vertically curved by a tiny degree. You are using optics to make the shapes using photolithography, so the center of the water is slightly closer to the light source than the edge of the wafer. This means the shapes are slightly different on the edges. Same with thicknesses of layers / deposited material can have different thicknesses based on where it is on the wafer. It can also vary wafer to wafer depending on random effects, tolerances (the vapor deposition lasts X seconds, plus or minus a ms, because the computer that turns it off has that tolerance, e.g.)


We are talking about manufacturing tolerances on order of wavelengths of light. And doping where the carrier concentration is on the order of 10 to the negative 18. Very very touchy stuff.

So these tiny tiny differences have tiny effects - a picosecond here or there. And, as you note, one over the path length gets you your max frequency.

Everything you wrote is accurate.

 

[Pezimak]

You mean iPhone 13 not the iPhone 13 Pro. Pro has 2GB more RAM, better cameras, pro motion, faster speed A15. I think they did this so the upcoming iPad Air will have the same speed as the iPhone 13.

No I meant the iPhone Pro, which I is why I said it. In my country the new iPad Mini is half the cost of the new iPhone 13 Pro. The only thing you'll notice on the Pro are the better cameras, you aren't going to notice 300mhz faster processor. Maybe 2GB ram will help.

 

[ChromeCloud]

Cmaier is right. It’s not underclocking. It’s due to chip manufacturing variation. Apple probably had an internal tolerance specification where a chip passes QA if it exceeds a certain speed.

It’s a similar with cars and specified power output from motors. The HP/kW value they specify is extremely conservative, since it’s Ok to sell you something with more power, but not OK to sell you a product with less. 🤷🏻‍♂️

Even if Apple were to include clockspeed, it would be pointless because all their products would exceed it. They’d advertise whatever conservative value they want, even if some exceed the spec by 10-15%.


BTW, I’m not saying I like it. 🤷🏻‍♂️ However, it’s not underclocking.

I understand your angle about the definition of “undercloking”.

But your car analogy is incorrect.

In this case Apple systematically puts a less powerful version of the same SOC in a specific product, without disclosing it.

It would be like a car manufacturer producing two models of a car advertising a certain power spec and then putting a less powerful version of the same engine in one specific model, without disclosing it.

Also the argument about Apple having to advertise the “conservative” clock speed does not stand.

Clock frequencies are decided when you decide how to bin the processors.

Every single A15 they put in the iPhone 13 Pro runs at 3.2 ghz.

Every single A15 they put in the iPad Mini runs at 2.9 ghz.

It’s not a lottery, but a systematic difference in performance. Therefore it should be disclosed in the specs.

 

[kaycrystal626]

Doubt anyone notices any difference for daily use, unless you have to render some long videos.

 

[Homme]

10 answers down finally an informative answer.

So they did it before.

Chances are, iPad A15 is way too fast for normal users anyway, nobody needs that much power for the apps available on the iPad. So trade-off for battery life / heat generation, I am guessing. Which would be a very good trade-off.

For example doesn't get hot after hours of gaming use - things like that, practical things Apple cares about more than feeds and speeds.

Plus, your game won't run faster with a faster chip, an A15 will likely make easy work of any available games - since iPad games have to support back devices from years and years ago.

yes I agree trade off for power for battery life. Plus Mini 6 has obviously a bigger battery than the 13 Max. Not sure about the capacity but I saw somewhere around the 5100 mah area don’t know the exact size

 

[spinedoc77]

As long as it can background apps so they don't reset information I'm happy, oh wait...

 

[RAS2MR]

The price to go from 64gb (a chip they keep in this scenario) and have 256gb instead is highway robbery.

Flat out eye gouging theft

Given that perspective e you definitely shouldn't buy the 256GB version. Others will look at the capability it provides to them relative to their needs and will decide that the capability is worth that price difference. If very few people were to decide that they want to pay the price difference for the 256GB model, then Apple would either lower the price or would pull the product off the market. Thats how free markets work.

 

[enc0re]

Doubt anyone notices any difference for daily use, unless you have to render some long videos.

Might even be the opposite. I bet the mini chassis can dissipate heat better than an iPhone, perhaps giving better sustained performance.

 

[GuruZac]

That is lame... They stick an overkill M1 in the iPad Pros and hamstring these.

Hardly hamstringing. There's probably a reason for it, and no it can't be cost savings for Apple. If anything, having two different chip specs may cost them slightly more money. There is simply not going to be any tangible differences in performance. And though the M1 in the iPad Pro is overkill, the RAM increase is not. I am noticing on my 2018, 4GB isn't enough. It can't hold many apps open and will frequently refresh Notability on me, so a base of 8GB of RAM is a huge increase. I wasn't going to upgrade but the fact that the 2021 iPad Pro works with the Pencil 2 and Magic Keyboard, accessories I already have, I'll go ahead and upgrade.

I plan to pick up the iPad Mini and a portable ultrasound when I begin residency.

 

[jicon]

If tempted to buy, I'd suggest a bit of caution with this, and maybe wait out real world use to garner how reliable this is.
Going back to the original Mac mini with a G4 chip, it too was underclocked... presumably so the ATI 9200 card it carried could continue to work without a fan.

Except- that had compromises. This was one of the first Macs I think that enabled use of any monitor with a VGA or DVI-D connector. Run a monitor at 800x600 resolution over VGA, and nothing seemed wrong. But, if instead DVI-D was used, and resolution bumped to 1280x1024, well... then the GPU compromises started to show. Fairly bad clipping of the video signal. ATI blamed the outdated firmware and underclocked/undercooled chip... Apple ended up switching to integrated graphics on the next Intel based model, and I got a nice iPod Photo from Apple out of the whole debacle.

That was a long time ago, and maybe Apple's got some better handling of things now, but still, talk of underclock is a bit of a trigger subject for me when Apple is involved.

 

[cmaier]

Hardly hamstringing. There's probably a reason for it, and no it can't be cost savings for Apple. If anything, having two different chip specs may cost them slightly more money. There is simply not going to be any tangible differences in performance. And though the M1 in the iPad Pro is overkill, the RAM increase is not. I am noticing on my 2018, 4GB isn't enough. It can't hold many apps open and will frequently refresh Notability on me, so a base of 8GB of RAM is a huge increase. I wasn't going to upgrade but the fact that the 2021 iPad Pro works with the Pencil 2 and Magic Keyboard, accessories I already have, I'll go ahead and upgrade.

I plan to pick up the iPad Mini and a portable ultrasound when I begin residency.

The whining going on here is an example of people knowing just enough to get it wrong.

So they use lower bin parts. So what? Peak performance goes down by a handful of percent. But the iPad mini can maintain peak clock speed for far longer than an iPhone, due to cooling. So for anything other than a benchmark that runs for a very short time, the iPad mini completes it faster.


But everyone wants to whine because the clock didn’t go up and down as many times a second.

Would you rather have a car that goes 0-60 in 2 seconds but then can’t maintain a speed beyond 40mph, or a car that take 5 seconds to reach 60mph and can maintain that speed?

But people feel they are owed 0-60 in 2 plus that 60mph top speed otherwise they aren’t getting what they were promised, even though nobody promised them that.

 

[svish]

That’s surprising. Maybe due to thermal issues or for better battery life.

 

[FriendlyMackle]

I don't care about the price and the slower processor, I just wanted 6GB or 8GB of RAM.
4GB of RAM is not enough to make it future proof for the next 6 years.no

Totally agree! The low RAM (4GB) makes me think any web pages will constantly be refreshing, and apps will lose their place anytime you switch between them. I had this problem eventually with my original iPad Air. Not immediately, but after 2-3 years. Of course, that device had only 1-2GB RAM (not sure). But my current 2018 iPad Pro 11” is beginning to have the Safari refresh frequency problem—and I think it has 3GB RAM. But, that’s really the only disappointment I see with this device, aside from maybe an option with more storage.

 

[HobeSoundDarryl]

I suggest restarting your Pro and closing all apps not in immediate use to maximize your free RAM. I'm still rocking a Mini 2 (1GB ram) and it is NOT "constantly refreshing web pages" and apps are not "losing their place" UNLESS I have a number of tabs open and/or too many apps open. Empty your 3X-more-RAM than I have and I bet your experience will improve.

I have friends that were griping about "slow" iPads. In helping them, I always look at how many apps are open- usually every app they've ever opened because they don't know how to close apps- and then I show them how to restart their iPad to flush out any memory tied up for nothing. Magically, after freeing their available RAM, their iPads speed up.

I'm finally jumping on this 6 myself only because the 2 is no longer getting iOS upgrades and some apps I use seem to really "want" new iOS functionality. Since 1GB RAM still works pretty well on the 2, I'm thinking 4 times that is going to be amazing instead of "low."

Of course, we can always want more "specs" stuff. Had Apple put 6GB RAM in, someone would gripe why not 8? If there was 8, why not 16? There are plenty of posts about the A15 being slowed just a bit from the A15 in iPhone. Why not M1? Why not M1X? Why not M1X AND Intel so we can run Windows too? Why not a card slot for dedicated graphics? Why not 2 card slots?

I can "want" with the seemingly many craving what would be the iPad Mini Pro (though many posts read like they don't want to pay up for such a pro at what would likely be $1000+). Had Apple also rolled out such a pro, I would have paid the difference. To me, that would be the perfect Mini.

But this is what they offer. If we like it, we buy it. If we don't, we can buy something else or buy nothing. How dollars actually flow is the only tangible way to show Apple if they made good or bad product dev decisions. Coming from the "2" and this form factor being "ideal" IMO, it was a no-brainer purchase for me. For others wanting features not available here and/or wanting what is available here for substantially lower prices, this is not the product for such people.

I excitedly await this loaded 6 to arrive. Yes, I certainly wish it cost less, but I could have that same wish if it was priced at $600, $500, $400, $300, $200 or even $100. Much like the Mini 2's useful longevity, I expect this one to still be in daily use 4-6 years from now. Spread even the max config specs over- say- 5- years and we're talking < $200/yr for a daily use tool/device. Meanwhile, over in iPhone threads, people are clamoring to spend $1599 on (maxed) iPhone MAX that they'll be spinning as "long in tooth" as soon as next year or the year thereafter to rationalize paying $1799-$1999 or whatever Apple wants for iPhone 14 or 15. Many woke up in the wee hours desperately trying to spend much more than this cost ASAP... and then are gushing if they were "lucky" enough to get one.

 

[GuruZac]

Totally agree! The low RAM (4GB) makes me think any web pages will constantly be refreshing, and apps will lose their place anytime you switch between them. I had this problem eventually with my original iPad Air. Not immediately, but after 2-3 years. Of course, that device had only 1-2GB RAM (not sure). But my current 2018 iPad Pro 11” is beginning to have the Safari refresh frequency problem—and I think it has 3GB RAM. But, that’s really the only disappointment I see with this device, aside from maybe an option with more storage.

The 2018 iPad Pro were all 4GB of RAM except the 1TB storage models which came with 6GB of RAM. My 2018 12.9 is showing some of the RAM limitations. I’m planning to upgrade to the 2021 base model. With 8GB of RAM should future proof me for a while.

 

[firewood]

But you also mentioned doping of structures or layers. That IS within engineering control, no? Doping is intentional with predictable, desired results, isn't it?

Doping is a diffusion process, and diffusion is a statistical process. Thus you are guaranteed to have non-zero local variance in density.

(unless and until there is some nano-engineering mechanism to individually position each atom at an exact location is a perfect crystal lattice. I think IBM labs has done that on a very small scale, using an electron microscope to place a small number of individual atoms to spell "IBM", but that's hardly a volume manufacturable process).

 

[Crow_Servo]

Many thoughtful posts in here. I do have faith now that Apple was trying to make sure the A15 could perform longer by underclocking it slightly. I still can't think of a personal situation where RAM has ever been an issue for me, so I'd say 4GB works for me at least. It's what my Air 4 has, and it runs like a dream.

Would it be nice for the Mini to get ProMotion? Of course, but I never expected it. I expected 4GB RAM, hoped for A15 and didn't really know what else to expect besides an 8.3" Liquid Retina display, USB-C and an improved rear camera. Oh, I also expected 5G (even though it doesn't include mmWave).

Did I expect Center Stage on the front camera? Absolutely not, but I have been saying ever since the Pros got Center Stage that it should become standard on all iPads. Looks like I got my wish sooner than expected. I also didn't expect a rear camera flash, since Apple has been careful not to include one on any iPad other than a Pro until now. So that was also a pleasant surprise.

But I do see how it can still be frustrating to know that we'll probably never get a Mini Pro. At least it's treated as a Mini Air instead of an entry-level iPad, though. For many, the Pro is overkill, and we never got an iPhone Mini Pro either, so Apple just doesn't like miniaturizing their Pros to the extreme, I suppose.

I definitely want this Mini. Hoping to secure one by the end of the year.

 

[code-m]

I decided to wait on the Mini.

My experiences with the M1 overheating on the iPad Pros almost instantly was a reminder. I also want to see if folks find it significantly lighter (7.5 grams doesn’t sound like much on paper.)

Hearing this news makes me feel slightly better about the A15 in the Mini insofar as heat, but I also understand why people are annoyed.

While it’s a personal choice to purchase or wait on any product consider this from what Apple has been doing in the past number of years. The iPad mini 6 may not be updated for the next 2 years.

2019 iPad 10.2” = A10
2020 iPad 10.2” = A12
2021 iPad 10.2” = A13
2022 iPad 10.2” = A14
2023 iPad 10.2” = A15
2024 iPad 10.2” = A16
2025 iPad 10.2” = A17

2019 iPad mini 5 = A12
2021 iPad mini 6 = A15
2023 iPad mini 7 = A17

2019 iPad Air 3 = A12
2020 iPad Air 4 = A14
2021/22 iPad Air 5 = A15/16

The iPad Air and mini have to stay a couple Apple Silicon chip generations ahead for a reason as per generation improvements are slowing down and will not be noticed for daily tasks for most users. Apple has decided to update the entry iPad annually for a reason possibly due to manufacturing costs or software related reasons.

iPad Air and mini may become every two year updates while the iPad an annual update iPad Pro maybe on a 1.5 year update cycle with AS schedule.

In 2023 the mini 7 will receive base 128GB storage and people will complain that it should have 256GB and the cycle continues. It will probably receive a 6GB RAM upgrade too. The wild card in all of this is if Apple decides to merge the iPad and Air line or merge the Air and Pro line. Due to feature overlap and at some point the entry iPad will gain most of what the iPad Air 3 had such as laminated screen, 10.5”, etc.

To keep manufacturing cost down and reduce consumer confusion the lineup should eventually resemble something like this:

iPad mini 8.3”
iPad 10.5” - 11”
iPad Pro 11”
iPad Pro 12.9”

in parallel to

iPhone mini 5.4”
iPhone 6.1”
iPhone Pro 6.1”
iPhone Pro Max 6.7”

 

[code-m]

A smaller chassis (like in the iPad Mini) likely has less ability to dissipate processor heat? Possible overheating could be one reason. But I'm sure I will be corrected by eggier eggheads here.

This logic is challenged as the iPhone 13 and Pro have an A15 with some differences in a smaller physical body.