iPad iPad Pro 10.5/12.9 SSD Storage Speed test 64/256/512GB

Discussion in 'iPad' started by IcedT1, Feb 11, 2018.

  1. IcedT1, Feb 11, 2018
    Last edited: Feb 16, 2018

    IcedT1 macrumors newbie

    IcedT1

    Joined:
    Dec 6, 2016
    #1
    I am a student who owns the iPad Pro 10.5 64GB gold and I have been very happy with it so far. One day I downloaded an app to test my iPad disk speed and noticed the write speed is half the one on my iPhone 8 plus 64GB. I saw a similar thread for iPhone 8/plus/X somewhere on this website and decided do it for the iPad. This is the app I used:
    PerformanceTest Mobile
    https://itunes.apple.com/us/app/performancetest-mobile/id494438360
    Please Kindly reply with your read/write speed as I will update the average speed on this thread regularly.

    iPad Pro 10.5 write/read (MB/s)
    64GB:
    72.3/1214

    256GB:
    320/1221
    512GB:

    iPad Pro 12.0 Gen 2 write/read (MB/s)
    64GB:
    203/1149
    256GB:

    512GB:
    390/1025
     
  2. sparksd macrumors 68020

    Joined:
    Jun 7, 2015
    Location:
    Seattle WA
    #2
    10.5 LTE 256GB --

    Write 320 MB/s
    Read 1221 MB/s

    Jumped around some over multiple tests, above are eyeballed averages. Are you numbers for read/write for 64GB transposed? They must be write/read, not read/write.
     
  3. IcedT1 thread starter macrumors newbie

    IcedT1

    Joined:
    Dec 6, 2016
    #3
    Very interesting yet not surprising result. I noticed that comparing your result to iPhone 8/plus/X your write speed is about 100-200MB/s lower and my iPad is 100MB/s lower than the same capacity iPhone. it is pretty disappointing knowing Apple didn't use the same SSD on the iPads which was released the same year.
    Yeah thanks it was transposed. I have reversed them now.
     
  4. Eric5273 macrumors 6502a

    Joined:
    Apr 12, 2009
    Location:
    New Jersey
    #4
    iPhone 8 has the A11 processor while the iPad Pro 10.5 has the A10X, so it is no surprise that the iPhone 8 would be faster at reading/writing to the storage.
     
  5. rui no onna macrumors 603

    rui no onna

    Joined:
    Oct 25, 2013
    #5
    That's just the nature of SSDs. The NAND flash is kinda arranged like internal RAID-0 so until you hit another bottleneck (processor, bus, etc), higher storage typically means higher speed, too.

    Storage performance is also affected by how full the SSD is. A brand new iPad with barely anything in it is going to have faster storage performance than an iPad with less than 5GB free space remaining.

    And really, the more important metric for OS performance is random small block read/write.
     
  6. IcedT1 thread starter macrumors newbie

    IcedT1

    Joined:
    Dec 6, 2016
    #6
    I disagree. Processor has nothing to do with storage in terms of read/writing. Processors are no longer a bottle neck when it comes to loading application, storage speeds play a big role.
    --- Post Merged, Feb 12, 2018 ---
    But my point is that the iPad Pro has slower writing speed than the latest iPhone which I think should be using the same SSDs but apparently they don’t. If you spend so much money on an iPad wouldn’t you want it to have higher writing speed? My iPad only has 70MB/s compared to 190MB/s on my iPhone 8 Plus.
     
  7. rui no onna macrumors 603

    rui no onna

    Joined:
    Oct 25, 2013
    #7
    Iirc, the SSD controller on iOS devices is now built-in to the chipset so I expect processor speed definitely plays a part. Also, there's improvements to NAND flash year to year and support for newer NAND is dependent on the chipset/controller.

    What tasks would actually see benefit from much faster sequential write speed? Thus far, only 4K video exports come to mind. Downloading and installing apps? Bottlenecked by internet connection. Restoring from iTunes on computer? Bottlenecked by USB 2.0. Downloading videos (even from LAN)? Bottlenecked by LTE/Wi-Fi. App loading? More read speed than write speed. Real world usage, the faster sequential write speed doesn't make much of a difference.

    Now if sequential write speed is so important, then just buy a higher capacity model.

    As points of comparison:

    iPad Air 2 128GB (A8X)
    Read: 378 MB/s
    Write: 122 MB/s

    iPad 9.7 128GB (A9)
    Read: 627 MB/s
    Write: 274 MB/s

    iPad Pro 9.7 256GB (A9X)

    Read: 726 MB/s
    Write: 329 MB/s

    iPad Pro 12.9 512GB (A10X)
    Read: 1025 MB/s
    Write: 390 MB/s

    iPhone 7 32GB (A10)
    Read: 615 MB/s
    Write: 39.9 MB/s

    iPhone 7 256GB (A10)
    Read: 877 MB/s
    Write: 323 MB/s


    Frankly, despite the 8x difference in sequential write, there's no discernable difference in performance between the 32GB and 256GB iPhone 7.
     
  8. IcedT1 thread starter macrumors newbie

    IcedT1

    Joined:
    Dec 6, 2016
    #8
    Perhaps you are right that newer chipset affects the storage speed substantially. Though I am aware I don't notice any performance difference between my 8 plus and iPad (which they really don't) but as a "perfectionist" it kinda bugs me. I shoulda gone for the 256GB iPad but I guess its way too impractical. But 5 years from now 64GB is going to be outdated anyway I guess I would just go for higher capacity.
     
  9. IowaLynn macrumors 65816

    IowaLynn

    Joined:
    Feb 22, 2015
    #9
    UFS 2.x is the special sauce. UFS 3 will kick it up to 2000MB/sec. Samsung Galaxy 8+ is slightly slower than Note 8 but manage ~700MB/sec

    Faster reads help instantly load/reload apps, and their data. Not just for 4K video. Higher capacity, higher number of concurrent channels (RAID0) for I/O.
     
  10. rui no onna macrumors 603

    rui no onna

    Joined:
    Oct 25, 2013
    #10
    Reads seem to be nearly the same across capacities of the same gen. It's the sequential writes that vary and the use cases where sequential write is the bottleneck is quite limited.

    More important is random small block performance. As of Apple A10(X), I believe that's still just ~25MB/s for random 4K read and ~2MB/s for random 4K write.
     
  11. petsk macrumors 6502

    petsk

    Joined:
    Oct 13, 2009
    #11
    iPad Pro 12.9 gen 2, write/read (MB/s)
    64GB:
    203/1149
     
  12. rui no onna macrumors 603

    rui no onna

    Joined:
    Oct 25, 2013
    #12
    Interesting. I wonder how big of a part RAM caching affects these results.
     
  13. IcedT1 thread starter macrumors newbie

    IcedT1

    Joined:
    Dec 6, 2016
    #13
    Thats a strange result. I own the same capacity ipad pro 10.5 but only getting one third of your write speed. Does it affect your usage when it comes to installing applications? Mine seems to be a little bit slow in that department.
     
  14. rui no onna macrumors 603

    rui no onna

    Joined:
    Oct 25, 2013
    #14
    More a function of CPU, internet and random read/write rather than sequential write speeds. If sequential write speed was the primary bottleneck, then installing a 1GB app would take, for example with the iPhone 7 results above, 25s with the 32GB model and 3s with the 256GB model. Yes, the 256GB iPhone installs slightly faster but certainly not 8x as fast and definitely not within just 3s.

    iPhone 8/8+/X have a faster CPU than iPad Pro it's normal that they're faster when it comes to app installs.
     
  15. IcedT1 thread starter macrumors newbie

    IcedT1

    Joined:
    Dec 6, 2016
    #15
    Makes sense products with A11 SOC have higher writing speeds. But as you can see it was a ipad pro 12.9 gen 2 with faster writing speed than my 10.5 and they both have the A10X SOC.
    Btw, should I worry about my iPad 10.5 (80MB/s)? It is relatively slow, I am just afraid it can be a bottleneck few years from now.
     
  16. rui no onna macrumors 603

    rui no onna

    Joined:
    Oct 25, 2013
    #16
    Wouldn't be worried at all. The read speeds are still very fast and that has more impact for day to day use. Afaik, random speeds are practically the same across all capacities. Before sequential write speeds are even an issue, I expect you would have long updated for faster/more CPU/GPU/RAM.

    As for the 12.9 gen 2 being faster than 10.5, as I mused earlier, it could be a function of caching or even free space. We need more samples. That said, the storage bench in PerformanceTest is so short that it's quite likely not super accurate.
     
  17. graceyj20 macrumors newbie

    Joined:
    Apr 21, 2019
    #17
    iPad Air (Gen 1, 2015) 32 GB
    write/read (MB/s): 48/105

    iPad Air (Gen 3, 2019) 64 GB
    write/read (MB/s): 218/1553
     
  18. EugW, Apr 21, 2019
    Last edited: Apr 21, 2019

    EugW macrumors 603

    EugW

    Joined:
    Jun 18, 2017
    #18
    I guess that explains my results with my USB 3 CompactFlash reader, through my Lightning to USB 3 camera adapter. I got about 70 MB/s to my 64 GB iPad Pro 10.5". It was an old CompactFlash card so I figured it could be a limitation of the card, but it looks like it is a limitation of the SSD write speed in the iPad.

    Interestingly, when I tried the same setup on my iPad Air 2, it either would take forever, or else it would simply fail. I wasn't sure what was going on, until I tried an old USB 2 CompactFlash reader. Then it worked fine, at very decent USB 2 speeds.

    It seems there is a compatibility issue on the USB 2 iPad Air 2 with my USB 3 CompactFlash reader, at least with that specific UDMA-enabled card.

    Note that the iPad Air 2, iPad 9.7, and iPad Pro 9.7" are all USB 2, so you will never be able to achieve those write speeds from external storage via the Lightning port.
     
  19. tps3443, Apr 21, 2019
    Last edited: Apr 21, 2019

    tps3443 macrumors 6502a

    tps3443

    Joined:
    Jan 24, 2019
    Location:
    NC,USA
    #19
    Oh I wanna join in on this. What app are you guys running these test on?

    My write speeds are in the 1,400 range which doesn’t sound right. Seems backwards to me. I’ve tried the test numerous times, I cannot get it to complete the disk benchmark and show both read and write speeds. Only write speeds show up, which are oddly high and look more like read speeds to me.

    [​IMG]


    It says I’ve scored 238,920 overall and looking on there website, this is there disk benchmark chart showing recently uploaded benchmark scores of 1TB iPads at the top with the highest at 195,000 range. But, there’s no way to see a break down of what there read/write MB speeds are. So, I have no idea what my write or read speeds actually are..

    So I’m really confused by this. I wish I had all of the information here.


    [​IMG]
     
  20. EugW macrumors 603

    EugW

    Joined:
    Jun 18, 2017
    #20
    It's in the original post.

    https://itunes.apple.com/us/app/performancetest-mobile/id494438360

    However, for my CompactFlash test, I just copied the files over and timed it.
    --- Post Merged, Apr 21, 2019 ---
    For this model I got 76.0/790 MB/s, with a pretty full iPad (6.3 GB free according to PerformanceTest Mobile, and 17.2 GB free according to the General Settings info).
     
  21. rui no onna macrumors 603

    rui no onna

    Joined:
    Oct 25, 2013
    #21
    Impressive score. That said, as I mentioned previously, the disk benchmark in PassMark is probably too short. I wonder if/how Apple's storage controller employs RAM and/or SLC caching. There's way too much variance running the test in succession on the same device (6th gen 128GB 1st run: w288/r936; 2nd run: w186/r708).

    Too bad AnandTech hasn't done one of their in-depth reviews for iPads in a while. They have an excellent review of the iPhone XS (Max). Unfortunately, it didn't cover storage performance.
     
  22. tps3443, Apr 21, 2019
    Last edited: Apr 21, 2019

    tps3443 macrumors 6502a

    tps3443

    Joined:
    Jan 24, 2019
    Location:
    NC,USA
    #22
    I dunno 1,400mb write speed is strong. Very strong. It almost seems to good to be true, and now I’m really wondering what kind of SSD is in this 1TB iPad lol. If the write speeds are 1,400 then read MB speeds would be sub 2,000 easily.

    The test does fluctuate a little, but seems to stay between 1,330mb and 1,500mb write speeds. I’ve tried reinstalling, a hard reset, I cannot get the read test to work, I can click memory read and a blue bar loads really fast and nothing happens. Seeing how the score was at 240,000 sometimes I can only assume it’s high.

    I wish we had a real diskmark app for iOS.
     
  23. rui no onna macrumors 603

    rui no onna

    Joined:
    Oct 25, 2013
    #23
    If you're willing to shell out $0.99, maybe you can try this:
    Jazz Disk Bench by Guangyu Chang
    https://itunes.apple.com/us/app/jazz-disk-bench/id1298441490?mt=8

    I find 1400MB/s write to be unsustainable considering the iPad likely doesn't have the degree of parallelization as one might get from a larger SSD with multiple NAND flash chips. Burst thanks to either memory or SLC caching, sure, but not sustained. Mind, I'd be quite happy to be proven wrong on this regard.

    Note, I believe the storage controller is built into the A12X chipset and addresses raw NAND flash chips.
     
  24. tps3443, Apr 21, 2019
    Last edited: Apr 21, 2019

    tps3443 macrumors 6502a

    tps3443

    Joined:
    Jan 24, 2019
    Location:
    NC,USA
    #24
    They can pack some serious storage speed punch in to a tiny little area. If you look at Samsung 970 Pro m.2 SSD, with its (2) little nand chips, and baby sized storage controller, it is just a little larger than our built in setup on our iPads logic boards. I know our iPads can not compare to a 970 m.2, or even the majority of m.2 ssd’s on the market.

    There is so little technical information explaining iPads. No tear down videos, no hardcore in-depth internal chip information or break downs listed about what is even on our logic board or anything past pointing to where it’s speakers are and saying; these are the speakers, there are 4 of them lol..

    I’m looking at photos of the iPad Pro 2018’s Logic Board. And you can see the two micron memory chips directly on the CPU, you can also see where the internal storage is, (2) spaces for nand storage chips, one of which is missing a blank spot on the lower tier capacity models. right above them both on the logic board looks like the storage controller.

    I wish I could see the 1TB models logic board, I wish I could open mine up and see if it is the same type of Nand, or maybe a different brand, possibly even a different storage controller too. But probably not. Most likely just both NAND chip area soldered on, instead of just one.

    Also another thing, I think the missing 7% of ram that is not seen by iOS is a base set minimum used as GPU memory.

    Anyways, I do not doubt that the iPad Pro 2018 “any storage model” could easily put down a solid continuous 500mb+ write speeds, and read randomly at over 1000mb-1,500mb easily.

    I’m with you though, I just don’t see 1,400+mb write speeds continuously being realistic at all. But who knows maybe half of that is over a continuous period of time.
     
  25. rui no onna macrumors 603

    rui no onna

    Joined:
    Oct 25, 2013
    #25
    I'm familiar with NVMe performance. Thing is even those M.2 2280 usually have at least 2 NAND chips to help with parallelism. If minimum storage wasn't 64GB, I'd agree with you on 500MB/s writes. Alas, there's usually significant performance penalty due to lack of parallelism on very low capacities.

    Iirc, starting with the A9 the storage controller is built into the Apple chipset so I highly doubt the 1TB models would have a different storage controller. As you have surmised, what's more likely is the 1TB models having more NAND flash chip packages compared to lower capacities (likely 2 instead of 1).

    Mind, since majority of the magic happens in the Apple A12X SoC, rather than the logic board, what we need is more of this:
    https://www.anandtech.com/show/13392/the-iphone-xs-xs-max-review-unveiling-the-silicon-secrets/2
     

Share This Page

40 February 11, 2018