Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
iPad Widely differing NAND flash speeds for iPad Air 2
- Thread starter userjohn
- Start date
- Sort by reaction score
Thanks, the 75 value is surprising. Do you remember about how many trials you did, and were any other results even close to that?
I also had one very low write result (85.7) for the 128GB, but of 20 tests each for 64GB and 128GB, no other result was <155 for 64GB or <170 for 128GB. Except for that outlier, the variability for the 64GB and 128GB sizes was substantially lower than for the 16GB.
Could you enlighten us as to your typical, daily use case for your iPad? I'm curious as to what real world usage - not running benchmarks, but actual productivity - that is both accurately representing how you use it, and is substianially more demanding than Newton's Apple's usage. Also, how much time (again, actual productivity, not benchmarks) have you actually measured has been wasted by slow NAND performance?
Pot, meet kettle.
Here's a thought: perhaps Apple is not aiming to please the benchmarking crowd. There are of course, other tablet models; I hear the Surface Pro 3 is pretty good. And as I've said many times, the one vote Apple truly cares about is the the one you make with your wallet.
I did 10 trials or so, had another result around 85 if i remember well. I have done another 10 just now, results dont vary much this time, one value around 75 again and the rest around 165.
Since you seem fixated on quantifying "productivity", perhaps you should use Microsoft ("At our core, Microsoft is the productivity and platform company...") For Apple's different core mission, see the wonderful "Intention" video from WWDC 2013: "The first thing we ask is: What do we want people to feel? Delight. Surprise. Love. Connection. Then we begin to craft around our intention." The delight I've experienced from all 6 iPad generations transcends productivity. I'd care about better speed even if all productivity/financial benefits (which I also appreciate) were disregarded.
From the Jobs biography: "If it could save a persons life, would you find a way to shave ten seconds off the boot time?" [Jobs] asked. Jobs went to a whiteboard and showed that if there were five million people using the Mac, and it took ten seconds extra to turn it on every day, that added up to 300 million or so hours per year that people would save, which was the equivalent of at least 100 lifetimes saved per year." [Engineer, Larry Kenyon] was suitably impressed, and a few weeks later he came back and it booted up twenty-eight seconds faster," [Bill] Atkinson recalled.
(I only calculate ~10 saved lifetimes/yr with 5 million users and 10 secs/day. However, Apple now has > 500 million iOS and Mac users.)
Although I appreciate a faster boot time, the subject of this thread is more important. Even a single file transfer can be delayed by more than 10 seconds (even minutes), and such delays can occur many times each day rather than only once. Although my usage is exceptionally high, I'd care even at vastly lower usage, and I'm sure some who have cared to run these benchmarks have moderate usage.
Even split-second delays matter to me and many others. For years I noticed slower response times (vs. iOS) on all my Android devices (e.g., 5 generations of Nexus phones and other flagship Android phones and tablets). I never measured such differences, but these articles provide some examples (e.g., Galaxy S4: 114ms, iPhone 4: 85ms, iPhone 5: 55ms):
http://bgr.com/2013/09/20/iphone-android-touch-screen-responsiveness/
http://bgr.com/2013/10/09/tablet-touch-screen-responsiveness/
Split-second delays in Android substantially degraded my user experience. That is one reason I've chosen to spend vastly more time using iOS than Android. Some people probably don't consciously notice such delays but are still affected, and some may not even be able to perceive them.
FYI, Google's testing showed a 200ms delay in web search response (which I would certainly notice and care about) caused a measurable decline in usage.
False equivalence. I never made an unprovoked attack (like claiming people who don't care about this subject are silly). I also didn't initiate the irrelevant claim of being "busy". Obviously people can greatly appreciate faster speeds without being busy at all (which is why the initial sarcastic comment that I must be "very busy" was ridiculous). There are countless threads about topics that don't matter to me. If I'm ever so close-minded and rude as to label them silly, please call me out.
Perhaps you don't know Apple cares greatly about benchmarks and justifiably brags about them in keynotes. Apple internally runs numerous benchmarks (including NAND benchmarks), despite only highlighting a few for marketing purposes. I'm glad Apple knows tiny delays affect user experience and they aggressively advance the performance state of the art (e.g., see A7 and A8). I was delighted and surprised by the much faster NAND flash and other performance improvements in iPad Air 2 (best iPad upgrade ever).
I have extensive experience with PC's and Windows and Android devices, and I prefer OS X and iOS and Apple hardware, so I have little interest in the Surface Pro 3.
I don't know why you mention Apple caring about wallet votes, since that's generally true for any company but is not what makes Apple very special. I've certainly voted with my wallet, as I've been both a large customer and a large shareholder for many years.
----------
Thanks for the update. There seems to be a pattern where something occasionally happens upon the write causing a significant delay that is relatively consistent in magnitude.
Thanks for your message. Perhaps you (or someone) can answer these further questions: How does this greatly improved flash in the iPad Air 2 compare to the best flash in other comparable mobile devices (not much higher power devices)? What has been the impact of technology from Anobit on Apple's flash performance and reliability? How feasible would it be for some future version of iOS to support a backing store on the iPad Air 2 (e.g., can this flash handle that level of activity?)
Just looked at the teardown and it looks like Apple's back to using NAND flash chips instead of eMMC packages. I'm just going to assume they put that Anobit acquisition to good use and now have the storage controller built into the A8(X). I reckon the Anobit acquisition was done more for vertical integration, patents and cost reduction while at least maintaining performance and reliability (e.g. MLC-level performance and longevity with cheaper TLC NAND).
Here's a good review for storage performance comparison:
http://www.anandtech.com/show/8666/the-apple-ipad-air-2-review/4
While the sequential performance is truly impressive, it's still slightly behind Samsung in terms of random performance which is somewhat more important for the OS. To be honest, the random write performance isn't really much better than what you'd get with a 3.5" 7200 RPM HDD. As I don't work for Apple and thus not privy to details of internal testing, I can't tell you if they made the right call. That said, seeing as I know how awful having a pagefile on the same HDD as your OS can be and taking longevity issues into consideration (maybe not so much for the 128GB but definitely for the 16GB), I would guess Apple made the right choice.
Seriously, if I were you, I'd stop worrying so much about sequential performance. For an OS drive, random read/write is more important and I reckon those don't differ much between different capacity iPads. As for the drops in performance, it could very well be just garbage collection at work.
Here's a good review for storage performance comparison:
http://www.anandtech.com/show/8666/the-apple-ipad-air-2-review/4
While the sequential performance is truly impressive, it's still slightly behind Samsung in terms of random performance which is somewhat more important for the OS. To be honest, the random write performance isn't really much better than what you'd get with a 3.5" 7200 RPM HDD. As I don't work for Apple and thus not privy to details of internal testing, I can't tell you if they made the right call. That said, seeing as I know how awful having a pagefile on the same HDD as your OS can be and taking longevity issues into consideration (maybe not so much for the 128GB but definitely for the 16GB), I would guess Apple made the right choice.
Seriously, if I were you, I'd stop worrying so much about sequential performance. For an OS drive, random read/write is more important and I reckon those don't differ much between different capacity iPads. As for the drops in performance, it could very well be just garbage collection at work.
Thanks, I read that great Anandtech review when it came out, but I forgot about Apple's relative random weakness, especially for writes, compared to Samsung. Although I also care about sequential performance (especially for large files), I've always realized random performance is most important for the system drive, and is the reason for the incredible difference between using a HDD or SSD based Mac (as I first experienced to my amazement in 2008 with the X25-M in my MacBook Pro). I'd be interested if anyone can suggest an iOS app that includes random read/write benchmarks.
This is nothing new. I've seen this same speed difference with usb memory sticks and SSD drives. Normally the larger drive on the same controller from the same manufacturer will give better performance. Why should it be any difference here? And if you bought the 16GB in the first place your going to run out of space before the speed becomes an issue.
I wouldn't be surprised if Apple bought a billion 16GB chips for pennies on the dollar years ago. These could be the same spec chips used in iPads from the very beginning.
Only solution is to pony up for the higher end models.
I wouldn't be surprised if Apple bought a billion 16GB chips for pennies on the dollar years ago. These could be the same spec chips used in iPads from the very beginning.
Only solution is to pony up for the higher end models.
Oh, I'm sorry. See, I presumed - falsely, it would seem - that people used computing devices to do things. As such, one would be able to to at least enumerate what it is that they do, and what it is about a product that is preventing them from performing those tasks.
Apparently this is not the case; the primary use for iPads is to run benchmarks all day. My bad!
Fair enough. Let me rephrase the question then: If you weren't running benchmarks, and obsessively asking others to run benchmarks to collect data, could you honestly say that you would in fact not be feeling love, connection, delight, surprise through the use of your iPad Air 2?
Follow on question: What about obsessively running benchmarks, and asking others to obsessively run benchmarks to gather data, allows you to feel love, connection, delight and surprise?
Can you show us data that proves that the boot time of the iPad Air 2 is singificantly slower than any previous generation of iPad? Can you show us data which conclusively correlates the use of certain types of NAND flash to slower boot times compared to other types of NAND flash?
Lastly: How often do you power down your iPad? Mine shows an uptime of 46 days, 9 hours, 29 minutes. I attribute this to the fact that I don't have a need to power down and cold boot my iPad, because it sleeps and wakes instantaneously. Sounds like a lot of lives have been saved to me, if we accept Jobs' premise, because iPads don't typically need to boot from a power-off state very often at all.
So, that argument is pretty much irrelevant. On to the next logical fallacy:
So again, we're back to your claim that slow NAND - and not network speed limitations, USB interfaces, or other factors - is slowing down the super-feelsy-things you do. This sounds like a lot like you're trying to be productive, even though you seem to abhor that word. So again... do you have actual measurements showing how much time has been wasted, by NAND latency, and not by other factors?
If you noticed them, then you can measure them.
Speaks for itself. But you're asking other people to do your work for you.
These articles measure touch screen responsiveness, not NAND flash performance. They also don't measure the iPad Air 2, and the data is over a year old. How does that relate to your premise about iPad Air 2? Where's the data for that? And what does this have to do with your claim about TLC NAND Flash?
Explain to me how people are affected in their ability to feel love, delight, connection, surprise, because of something that you even admit they may not be able to percieve?
Is Google using TLC NAND, and is that what's causing the problem? If not, what does this have to do with your initial claim?
Indeed, your arguments are full of them.
And false accusations, it would seem.
Right, we get it: you're a dreamer. You love to feel love and delight and surprise and connection and all that jazz. So again I ask: How does obsessing over imperceptible (by your own admission) differences in NAND Flash performance inhibit your ability to feel all that goodness?
Are you sure it's the actual components, and not your hangups over what someone has told you ABOUT the components, that are causing this angst and not-cool feels that you're feeling?
Can they, really, if they can't actually perceive them? You can run benchmarks over and over, sure, but you do know that most NAND Flash, including both MLC and TLC, have limited write cycles, right? And that you're shaving off just that little bit more of useful life off the storage media every time you run a benchmark?
Maybe I'm wired differently emotionally, but incessantly flogging my iPad's storage capacity doesn't give me good feels and joy. Taking photos, listening to and making music, and yes, even doing busy, productive things, makes me feel a whole lot better.
That's nice. This topic interests me though. It fascinates me, even. It makes me feel surprise, connection... not love though. That's pushing it a bit.
So, I'm going to keep reading and responding as long as the thread is open.
Well, I personally haven't called you "silly," and looking back at the thread, the word "silly" was used to describe the repeated benchmark tests run, not the people running them. So, you shouldn't have any need to call anyone out here.
You, on the other hand, seem to be trivializing my and others' questions to you... and questions are a normal part of discourse and conversation. So, I think I should take the opportunity now to call you out on that.
And they've shown - and it's been borne out in independent tests - that they are justified in bragging, compared to contemporary models. Which makes me wonder: have you tested the NAND Flash performance of competing tablets? How do they measure up? Maybe then you might have a case.
How do you know this?
That's a shame. Perhaps their Flash storage performs better. Perhaps you might feel more positive feels using one. You never know until you try. I have.
Very simple: If and when people shun Apple products to the point where it begins to hurt their profit margins, Apple will be better motivated to listen to the TLC vs MLC benchmarking crowd. You appear to making a choice to continue using Apple products despite the flaws you perceive.
It just seems to me that if something is so fudnamentally flawed, then a user should actively pursue alternatives. I did exactly that back in 2007. Otherwise, I woudn't be used anyting Apple-related today, and I'd still be on Windows.
Ah, then I suggest you use your many shares to make your opinions known to Tim Cook. I'm sure such a large shareholder would have his ear.
Last edited:
Among other mistakes, you presumed falsely that "doing things" only includes "actual productivity". It's already clear that I care about a single 10 second (or even less) delay per day from one file transfer, and I've already said my usage far exceeds that. I see no reason to enumerate further details about how I use my iPads to a troll.
Thanks for proving my point again by calling our interest in benchmarks "obsessive". Since I said I was already delighted even by the slower model, even before any benchmarks, your question is obviously nonsensical.
To use my prior analogy, loving the Tesla P85D and its amazing speed increase doesn't mean the prior P85 wasn't also delightful. A troll like you would claim that comparing the 0-60 times is obsessing about a meaningless benchmark, and you'd demand an enumeration of real world trips to determine whether the P85D's faster acceleration would improve "actual real-world productivity" by sufficiently shortening trips to the grocery store. Have fun with that!
I never claimed I boot my iPad Air 2 often, nor that it boots slower. In fact, I assume it would probably boot faster, since the NAND and other components are faster. The point of Jobs' quote (which was from many years ago about the Mac, not the iPad) was that even a 10 sec/day delay is important. The reason for the 10 sec/day delay (e.g., a Mac floppy long ago, or iPad flash today) doesn't matter. Of course, a smaller delay matters for the same reason, just proportionately less.
I've already stated my wireless speed (nominally 866Mbps) exceeds some apparent NAND limits, e.g., I've already done wireless transfers from the Internet to my iPad at > 400 Mbps (to RAM, not flash). Of course, you don't actually care about any such information.
Obviously I could have measured them if I obtained a high frame rate camera as described in the articles. I never claimed otherwise.
For iPad Air 2 NAND benchmarks, I volunteered my own work and others have volunteered theirs, while you've contributed nothing and continue to demand "answers" to your "questions".
People differ in their limits of perception. For example, some people can't notice flicker at lower refresh rates, or can't notice certain input lag, or can't see some DLP rainbow effects, or can't distinguish between a Retina and non-Retina display. You may lack some or all such abilities, but not everyone does.
100% of users would be able to perceive certain NAND delays (e.g., multi-second delays in a file transfer). For split-second delays, millions would perceive them. Those who can't perceive them would presumably be affected the least, but even they may be affected slightly because real world time is being lost for them (unlike with a slower refresh rate, for example).
Obviously the linked articles were about touch screens and not NAND and not iPad Air 2. The simple point is that tiny delays, for any reason (e.g., NAND, touch screen) for any device, can matter. The Google test was another such example. I hope you won't now ask me how flicker, input lag, rainbow effects, or non-Retina screens are caused by slow NAND (hint: they aren't).
I bet much of the decreased usage in the Google test resulted from an unconscious reaction to the 200ms delay, rather than a conscious perception and decision. Similarly Apple keeps improving iOS performance in hundreds of subtle ways that are not consciously noticed by some users but still increase their sense of ease and delight.
I never claimed to have any "angst and not-cool feels". Unlike some (outside this thread) who were upset about the NAND, I said I was delighted by even the slower model. That doesn't prevent me from being even more delighted with a faster one and caring about their relative performance.
Benchmarks are a tiny fraction of my use, and the impact is immaterial to me, especially as I'll upgrade this year (probably quite soon to the so-called "iPad Pro").
Apparently you're wired emotionally to enjoy trolling. Unlike various others who've contributed benchmarks or other facts, you've not contributed a single piece of actual information in this thread, despite so many words.
Although my responses to you are partly a waste of time dealing with a troll, they have probably conveyed some real information to others (e.g., presumably not everyone already knew about the touchscreen response tests, or Google's search delay test).
You're splitting hairs. People running silly tests would be showing the behavior of a silly person. I never claimed you used that word, but you've been even more obnoxious by now.
The people who've offered benchmarks here (and many more elsewhere) have somehow not felt the need to offer or demand an enumeration of usage details. You're the only person who claims to have a desperate desire for such details.
My anti-iPad "case" only exists in your imagination. I never claimed to be unhappy with even the slowest model, but merely wanted to compare the speeds of the different models. You're the only person in this thread who (insincerely "suggested" another vendor.
I've no desire to provide those details, but feel free to confirm this yourself, if you're really under the delusion that Apple only runs the benchmarks that they publish.
My current Apple preference is not from lack of experience with alternatives. I've tried more competing phones and tablets (including Windows Phone and Surface) than anyone I've met, and still own multiple other platforms. Perhaps your myopic focus on "actual real-world productivity" prevents you from realizing why some people prefer Apple or certain other products, such as Tesla.
Apple is acutely aware of the differences between TLC vs. MLC and numerous other details about flash. You're the only person in this thread suggesting the iPad Air 2 could be "fundamentally flawed" because one model's NAND speed hasn't increased by as much as another. I assume others who've submitted benchmarks like their iPads, as I do.
FYI, you can also e-mail Tim Cook, without even being a shareholder. So perhaps you can tell him why you think the Surface Pro 3 is better.
You assume too much. The only thing I exclude from "actual productivity" is performing NAND Flash benchmark tests. If you can reasonably explain to me how performing such tests constitutes "actually productivity," I will gladly admit I'm wrong.
And you have yet to provide any measurable data that attributes this to the object of your obsession: allegedly too-slow NAND Flash storage.
I can say my iPad is blue and emits rainbow unicorns. That doesn't make it true. I think it would be reasonable for people to request proof of such a claim, and to be highly skeptical if I responded by calling them a troll for doing so. Just as I'm requesting some shred of proof that the NAND Flash is performing as badly as you say it is, and affecting you as badly as you say it is.
Right, so it's not true. Let's move on.
No, I described specific actions as obsessive, not one's interest in them. Though if I'm truly in the wrong on this, you could again, just give us some proof - any proof - that your productivity is being affected by slow NAND. Then the actions would clearly not be obsessive.
But oh, right... you don't HAVE proof, so here we are.
If you're delighted, why even benchmark? Why make the benchmark results an issue? Why insist there is a problem?
But were you running track tests to quantify the slowness of the P85 versus the P85D? Were you asking others to do the same and report the results? Did you equate the slowness of the P85 versus the P85D to the Jobsian notion that slowness waste lives as you did here with iPad's NAND Flash performance?
If you did not, then I fail to see how this is a proper analogy.
Actually, I'm not the one saying that the difference in speed "saves lives." You said that.
And I'm still not exactly sure what you're getting at with the point of these tests. Are you saying the P85 is defective vs. the P85D? Are you trying to say the iPad Air 2 is defective? I guess so, since it's such an alleged life-waster.
Then why mention boot times at all?
Actually, no, that's quite useful information. It's actually also the first time you've mentioned that speed in this thread, and it's also different from your earlier claim that your network speed was in fact 500Mbps (your post here). So the discrepancy really brings up more questions than it answers.
Sooo, in order to perceive a performance difference, one has to acquire a high frame rate camera and observe the difference through it? Why then, is the point of doing all this?
Good research doesn't happen in a vacuum. People ask questions, and if the premise behind the work you're doing is sound, you can answer them without getting puffed up about it, or calling people trolls for merely asking those questions.
Seems to em the answer then, to satisfactorily prove there's a problem to 100% of the population, is to provide hard facts which prove that productivity is objectively and quantifiably affected by slow NAND.
Gee whiz, I wish the people making these claims could sure provide some of that hard data...
Oh right... it was something about "Delight. Surprise. Love. Connection."
But, apparently you're not experiencing a lack of this. And int he same post you say you care not about productivity... soo... what negative effects, exactly, are you experiencing?
So..... why collect this data, again?
Another false accusation. I'm just asking questions. Not getting very many answers. Lot of insults though.
Another false accusation. I'm asserting that iPad Air 2's are fine, regardless of NAND Flash performance, and my lengthy post history here would probably bear that out.
I'm also not the one collecting the data... for, what reason again?
Oh I'm sure I can, but I would imagine Mr. Cook might pay a little bit more attention to the guy that has shares, and also has collected hard data on NAND Flash performance, but yet is delighted and finds nothing wrong with it. Because... collecting data on perceived NAND Flash performance issues when you're really happy with a product is a thing people do these days? *shrug*
Last edited:
Honestly....who cares? The Air 2 is a beast and the fastest iOS device available. the 2GB of RAM makes this thing a dream to use. Swiping between apps now give almost zero delay before getting loaded back up. Before it would never register taps or swipes for a couple seconds. If you are that worried and geeky enough to care about RAM read/write speeds get a laptop. Good grief.
Last I checked, none of the iOS devices have USB 3.0 and wifi throughput tends to be a crapshoot unless you're sitting beside your router (and even then, best case you'll see around 50 MB/s real world throughput with 802.11ac). Given you're already bottlenecked by either wifi or USB, I don't really see how you'll even get near a 10 second delay when it comes to file transfers.
Mind, looking at the AnandTech review, it definitely looks as if the PassMark benchmark results posted here are showing effects of RAM caching.
Link speed is not the same as throughput. If your router can actually reach 866Mbps throughput, I'd love to know what router you have since even the best ones that SmallNetBuilder reviewed only reached around 500Mbps max throughput and around 250Mbps average (5GHz).
Last edited:
The upgrade to 2GB RAM was great, but you seem to have misunderstood this topic, which isn't about benchmarking RAM read/write speeds, but rather NAND Flash.
Thanks for your further input.
I'm especially intrigued by your comment about RAM caching. Can you please explain how much you think that affected the Passmark results, including how the impact varied by size? Any idea what size Anandtech tested (which I didn't see mentioned in the review)?
Wouldn't Wifi or USB limits be well above the 16GB write speed, potentially over 50% higher (even if ~31MB/s isn't inflated)? For example, couldn't a 2GB transfer be more than 20 secs faster at 50MB/s vs. 31MB/s?
Please note 10 secs/day was an arbitrary number I only cited because of the Jobs quote, and I added "(or even less)" since I'd also care about much smaller delays (e.g., split-second delays for much smaller transfers).
Of course, random reads and writes would be even farther below USB or Wifi limits, but you mentioned before that such speeds should be "similar" and not "differ much" by size. To clarify, should such speeds be entirely independent of size (with possible variations for other reasons)?
Am I understanding correctly that transfers from RAM to NAND (without use of USB or Wifi) can show the greatest speed disparity, such as ~87 MB/s vs. ~31MB/s, which is nearly triple the speed (perhaps more if the 31MB/s is inflated)?
I wrote "nominally" 866Mbps because I knew link speed wasn't real throughput (but thanks for pointing it out in case I didn't know). Even with the iPad Air 2 about 20 ft away from the newest Airport Extreme, in an adjacent room separated by a solid windowless wall, I could still obtain >400 Mbps Internet speeds. Usually the iPad would be closer to the router in the same room.