Tests Show 32GB iPhone 7 Models Have 8X Slower Data Write Speeds Than Other Capacities

[MrXiro]

8 times slower is NOT normal.

Real world speed of transfer wasn't 8X slower... it was more like 1.5X
[doublepost=1476883155][/doublepost]

I don't understand why some folks give this guy such a hard time. I think it's awesome that someone decided to create a channel that mixes tech and satire.

And, really, he's laughing all the way to the bank with 6.4M subscribers. So.

I think it's cause he comes off like a "gotcha" kind of jerk. Like his "bend test" on the 6 Plus and then the subsequent bend videos he did after it.

 

[Carlanga]

I think he means apple being cheap

IMO I don't think so, but he/she never replied back

 

[TechZeke]

So this is the catch for putting 32GB as the base. I knew the storage tiers this year made too much sense to be true. Apple had to keep their BS $10 margin somewhere.

Apple looks like the hero for finally making 32GB the base, but Timmy always gets his last laugh.

 

[agsystems]

Did anyone notice that although the write speed using that app causing the 32gb phone to be 8x slower, the 4.2gb movie write test only was about a minute longer (closer to 1.5x). If it was 8x longer, the 2:34 time from the 256gb version would be stretched out to approx 20 mins.

It appears the write speed of a large file improves dramatically compared the tiny file that app uses. I hypothesize that just initialization of the writing processes is what's really lagging 8x behind. Then it quickly improves to only 1.5x behind once it gets going.

Make sense?

You are making too much sense now and going against the forum

 

[MysteryMii215]

Now the iMac sales decline, Apple must put the pile of 5400 RPM HDD stock somewhere.

Mac Minis!

 

[tbrinkma]

8 times slower is NOT normal.

Unless the 32GB writes *negative* amounts of data over the tested time span, "8 times slower" isn't *possible*, either.

1/8th as fast is not the same as 8 times slower.
Base speed: 8 units per second
8 times slower: base speed - (8 * base speed) = 8 units per second - 64 units per second = -7 units per second
1/8 as fast: base speed / 8 = 8 units per second / 8 units per second = 1 unit per second

And, yes, massive speed differences *are* common, because larger flash storage includes multiple (often stacked) dies, allowing parallel writes, even within a single physical component.

And, like always, raw read/write tests don't show the real story in practice, just theoretical maximums.

The 256GB model wrote the 4.2GB movie in 2 minutes and 34 seconds while the 32GB model wrote the same file in 3 minutes and 40 seconds.

4.2GB in 154 seconds vs. 220 seconds means a real-world difference where the 256GB write speed is less than 1.5 times faster than the 32GB.

 

[rui no onna]

Umm, NO it's not. Not 8 times slower. Maybe 10-15% slower for the smallest capacity of SSD's.

That depends. Keep in mind that SSDs pretty much have internal RAID-0. As long as it's not bottlenecked by controller or bus speed, doubling the storage could potentially double the performance. Example is the Intel X25-V 40GB and Intel X25-M G2 80GB. Sequential write is double on the 80GB model. Also check out HDTach and HDTune tests for the Samsung 840 EVO 120GB and 250GB. The 840 EVO uses a pseudo-SLC cache so first few GB of writes is fast. However, once the cache is used up, sustained sequential writes on 120GB drop down to around half that of the 250GB.

Granted, the reason SSDs are good for operating systems is the 10-100x faster than HDD 4K random read/write. Sequential performance, while nice on paper, doesn't really matter all that much for OS. One wouldn't really notice much of a difference in performance between 32GB and higher models for most workloads.

So, it took 154 seconds to download the movie file to the iPhone 265GB model. And it took 400 seconds to do the same file to the 32GB model. That's roughly 2.6 times slower, not 8 times slower. If you do that math, the 256GB iPhone only transferred the movie at 27.9 MBytes/sec (not the app stated 341 MBytes/sec or 12.2 times slower) and the 32GB iPhone did it at 10.8 MBytes/sec (not the app stated 42.4 MBytes/sec or 3.9 times slower). So where is the bottle neck in both phones? And/or is the app even accurate when it comes to calculating the WRITE speed of the storage? My problem here is, that between the app he used and the actual speeds, the iPhone 256GB model apparently had more of a performance hit vs the 32GB, based on the app he used for the benchmark.

Umm, no. It took 154s (2:34) to do the transfer on 256GB or ~28MB/s, and 220s (3:40) to do the transfer on 32GB or ~19MB/s.

Keep in mind, USB 2.0 is a bottleneck and tops out at real world speeds of ~40MB/s. As I mentioned, PassMark only uses 90MB for their disk tests which is too small and the results may be further skewed by disk caching.

So this is the catch for putting 32GB as the base. I knew the storage tiers this year made too much sense to be true. Apple had to keep their BS $10 margin somewhere.

This has always been the case. If you think there's no performance disparity between 16GB iPhones/iPads versus higher capacity models, then think again. That's just the nature of solid state storage. It just happens to be getting a lot of media attention at the moment.

 

[sleepydinosaur]

Because moving 4gb on an iPhone happens so regularly....

 

[TechZeke]

This has always been the case. If you think there's no performance disparity between 16GB iPhones/iPads versus higher capacity models, then think again. That's just the nature of solid state storage. It just happens to be getting a lot of media attention at the moment.

Yes, I know how SSDs work. A 128GB SSD has always been slower than the 256GB, 512GB, or 1TB in Macs. 8x slower seems a little bit much, even if accounting its lower capacity. This seems like mix of the fact that it's smaller storage with the fact that Apple is using cheaper flash in the base model.

 

[iosuser]

Well then I guess paying the extra $100 for my 128GB is justified /s

I'm well aware of how SSD operates. Generally, higher capacity = additional NAND packages = higher write speed. Think of it as adding more cores to the SSD.

Just more FUD and click baits. He should stick to bending iPhones

 

[Tycho24]

Haha, I'm late to the party but glad to see I'm not the only one that's maths can't make 3.5 minutes equal 8x 2.5 minutes!

 

[wigby]

8 times slower is NOT normal.

8x is not normal but it's also not noticeable by typical users. The customers that try to save $100 or $200 by buying a smaller capacity phone are not clocking speeds on their devices. They don't care. Now if the larger iPhone was speed crippled compared to the smaller one, we might have an issue.

 

[Samuelsan2001]

Lmao at people immediately dismissing this guy for attacking Apple. And no, 8 times slower is NOT NORMAL. "How many people transfer 4GB movies regularly using iTunes?"

Excuses, pathetic from Apple.

It is normal that is how flash storage works the more you have the higher the write speeds this was found on all ssd's 128gb half the write speed of 256gb. There is a limit that the effect but at lower storage levels it's always been the case.

 

[rui no onna]

Yes, I know how SSDs work. A 128GB SSD has always been slower than the 256GB, 512GB, or 1TB in Macs. 8x slower seems a little bit much, even if accounting its lower capacity. This seems like mix of the fact that it's smaller storage with the fact that Apple is using cheaper flash in the base model.

Nah, they're all likely using cheap TLC NAND. Difference is with 128-256GB models, there's enough space for a pseudo-SLC cache. That said, short of benchmarks, there are few usage scenarios that will tax sequential write performance. Transferring to iPhone from computer is bottlenecked by USB 2.0 speed. Transfers over wifi/LTE? Bottlenecked by wifi/LTE speed (40MBytes/s translates to 320Mbit/s). Launching/running games and apps? Bottlenecked by CPU and GPU.

One of the few use cases is GSMArena's 4K video trim test wherein the 128GB iPhone took 17s and the 32GB iPhone took 52s. Not sure how many folks who plan on shooting 4K videos on their phone just buy a 32GB model, though.

That said, I do wonder if the pseudo-SLC cache helps with random 4k writes. Unfortunately, it seems most mobile benchmarks just focus on sequential performance when it's random small block performance that really helps boost system speed.

 

[Tycho24]

Because moving 4gb on an iPhone happens so regularly....

Haha, right?
ESPECIALLY on your iPhone that you had 28gb available right out of the box.

 

[AndrewR23]

Nah, they're all likely using cheap TLC NAND. Difference is with 128-256GB models, there's enough space for a pseudo-SLC cache. That said, short of benchmarks, there are few usage scenarios that will tax sequential write performance. Transferring to iPhone from computer is bottlenecked by USB 2.0 speed. Transfers over wifi/LTE? Bottlenecked by wifi/LTE speed (40MBytes/s translates to 320Mbit/s). Launching/running games and apps? Bottlenecked by CPU and GPU.

One of the few use cases is GSMArena's 4K video trim test wherein the 128GB iPhone took 17s and the 32GB iPhone took 52s. Not sure how many folks who plan on shooting 4K videos on their phone just buy a 32GB model, though.

That said, I do wonder if the pseudo-SLC cache helps with random 4k writes. Unfortunately, it seems most mobile benchmarks just focus on sequential performance when it's random small block performance that really helps boost system speed.

So what's the difference in shooting 4K video between a 32gb 7 and a 128gb 7?

I don't shoot much videos but I'm going to Disneyland for Christmas and rely on my iPhone 7 for video. In what way will this affect me ? Ty

 

[Deasnutz]

In the upside, people with that little storage won't routinely be transferring such huge files onto and off their phones, so I guess it kinda cancels out.

Right.....

 

[Applebot1]

So what's the difference in shooting 4K video between a 32gb 7 and a 128gb 7?

I don't shoot much videos but I'm going to Disneyland for Christmas and rely on my iPhone 7 for video. In what way will this affect me ? Ty

8 times slower I assume...reality you won't see the difference.

 

[rui no onna]

So what's the difference in shooting 4K video between a 32gb 7 and a 128gb 7?

I don't shoot much videos but I'm going to Disneyland for Christmas and rely on my iPhone 7 for video. In what way will this affect me ? Ty

Shooting video and then copying to Mac/PC or iCloud, there's no difference. The test that GSMArena did was to basically make a duplicate copy of the 4K video in the phone itself (well, not exactly duplicate since they trimmed the video). Basically similar to copying a large file in your HDD/SSD to the same HDD/SSD.

That said, I had a 64GB iPhone when we went to Disneyland a couple of years ago and I easily filled it up in one day with just 1080p30 video and some 720p240 slo-mo. It's a good thing I was carrying two iPhones with me (6 64GB and 6+ 64GB) so I was able to switch to the other phone when storage filled. I wouldn't bother with 4K video on just 32GB unless you don't plan on taking a lot. I believe the camera defaults to 1080p and you have to go to settings to change to 4K.

 

[extrachrispy]

There are a couple of factors in play here:

• Parallel I/O channels: yes, it's quite possible that the higher-end device has more parallel I/O capability than the lower end one. You pays your money and you takes your choice.
• More significantly: a 32GB device is very quickly going to fill its storage and reach an "aged" state compared to a 256GB device. That means that on the 32GB device, you will quickly reach a point where writing any new data is going to make the controller find a couple of partially-written blocks, read them, coalesce them, mark the old block for garbage collection, and write the coalesced blocks to a new location. This is how NAND flash works: once it has been written, it cannot be rewritten without an intervening erase cycle. In the extreme case, it will end up having to do this for every write, just to make space for the new data. It is quite possible to see performance degrade to 50% of its fresh-out-of-the-box value from this phenomenon alone, and this happens to ALL NAND flash storage, without exception. Combine a 2x slowdown from steady-state garbage collection with perhaps a 4x slowdown from fewer I/O channels, and there's your 8x difference. You pays your money and you takes your choice.
• TLC NAND is slower than SLC and MLC NAND. As manufacturers cram more bits into the same space, they have to switch first from SLC (one bit per cell) to MLC (two bits per cell) and then to TLC (three bits per cell). The physics of reading and writing MLC and TLC compared to SLC mean that you have to make multiple attempts with varying voltages to sense the charge in the cell, and that takes more time. Note as well that there is an order of magnitude reduction in write endurance at each step as well, from SLC to MLC to TLC. For phone users who upgrade every year or two, this is not going to be an issue.

Conclusion: the unbox dude is a sleazy sensationalist clickbaiter.

 

[bashman]

Ok this may be a stupid question. Will the 32GB only be slow when transferring stuff from computer to phone? What about downloading apps and installing?

 

[BornAgainMac]

Apple shouldn't even bother with the 32 GB model. 128 GB should be Grandma's iPhone.

 

[MrXiro]

Ok this may be a stupid question. Will the 32GB only be slow when transferring stuff from computer to phone? What about downloading apps and installing?

Downloading is only as fast as your Wifi or Network Data allows. But typically most data plans and wifi aren't anywhere near as fast as solid state storage. This whole "-gate" is nit picky whining that will effect at most 1% of people in real world usage. People who are buying 32gb iPhones aren't transferring 5gb of data at a time onto the phone and even if they ARE, the extra minute/minute and half it would take is negligible.

 

[bashman]

Downloading is only as fast as your Wifi or Network Data allows. But typically most data plans and wifi aren't anywhere near as fast as solid state storage. This whole "-gate" is nit picky whining that will effect at most 1% of people in real world usage. People who are buying 32gb iPhones aren't transferring 5gb of data at a time onto the phone and even if they ARE, the extra minute/minute and half it would take is negligible.

Thanks

 

[MrXiro]

So what's the difference in shooting 4K video between a 32gb 7 and a 128gb 7?

I don't shoot much videos but I'm going to Disneyland for Christmas and rely on my iPhone 7 for video. In what way will this affect me ? Ty

In no way at all, short of running out of storage space much more quickly.

You should just shoot in 1080p60 though IMO. There is really no point in shooting your vacation trip in 4K unless you want to 5X crop in on things during post production... which in all fairness you can crop at 1080p as well but probably no more than 2x before you start seeing blurring and artifacts from the compression.

Shooting in 1080p60 allows you to slow down the footage in post by 1/2 and if you use the 240fps slow-mo you can slow your footage by 8X, which IMO is more useful than crop.