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

[SimonTheSoundMa]

1) What are the 4k write speeds of both?

2) He transfers a file via USB 2, he'll only get a maximum of 20MB/s. So the question is, why the huge disparity?

 

[ColdShadow]

This is very very normal for how solid storage works.

no it's not.
normal is minimal speed differences, not 10x slower.

 

[rui no onna]

1) What are the 4k write speeds of both?

Abysmal (~2MB/s).

 

[oliversl]

This is really bad, the iPhones are not the same as they use to be.

 

[inscrewtable]

Surely the real question is not how much of a difference it makes, which is not much at all, but how is it possible to be slower in storage write speed than a 32gb iPod, which is much much slower in every other metric! Surely this would indicate that some funny business is going on.

 

[Hasukazu]

For people claiming this is normal behaviour, please read a review of how performance is effected by disk size. 8x is NOT normal.

http://www.anandtech.com/show/8216/samsung-ssd-850-pro-128gb-256gb-1tb-review-enter-the-3d-era/11

Exactly.

It's just hilarious how many butthurt 32GB owners are in here, trying fervently to bury their head in the sand and come up with all sorts of ludicrous excuses for the 32GB model being significantly slower.

 

[rui no onna]

For people claiming this is normal behaviour, please read a review of how performance is effected by disk size. 8x is NOT normal.

http://www.anandtech.com/show/8216/samsung-ssd-850-pro-128gb-256gb-1tb-review-enter-the-3d-era/11

Lowest capacity in that review is already 128GB class. Drop down to 32-64GB and the performance difference becomes more noticeable as the NAND becomes a significant bottleneck. Also, 850 Pro uses higher quality/performance MLC NAND. As far as I'm aware, all the latest iPhone models use cheaper, lower performing TLC NAND.

AnandTech SSD 2015 Bench: 850 EVO 250GB vs 850 EVO 120GB

From AnandTech's 840 EVO review, notice the effect of TurboWrite (pseudo-SLC cache). Once you write enough data to use up the TurboWrite cache, you'll see performance drops into a lower steady state value with 120GB sequential write being roughly half that of the 250GB model. From reviews, it seems higher capacity iPhones have a similar pseudo-SLC cache. Meanwhile, there's just not enough space to put a similar cache in 32GB models.

AnandTech - Samsung 840 EVO Review - TurboWrite

AnandTech - Samsung 850 EVO Review - HDTach Bench for 120GB

Notice how sequential write drops to ~150MB/s once TurboWrite cache is used up.

[doublepost=1476924643][/doublepost]

It's just hilarious how many butthurt 32GB owners are in here, trying fervently to bury their head in the sand and come up with all sorts of ludicrous excuses for the 32GB model being significantly slower.

I've got 2x iPhone 7 256GB. I'm not trying to find ludicrous excuses because I own the 32GB model. Merely a logical explanation for the disparity in performance. My first SSD was an Intel X25-V 40GB and over the years, I've done quite a bit of reading on SSDs. You can't just point to one SSD with a completely different controller and configuration and make a blanket statement that maximum performance difference between capacities should just be 10-15%. The why and how are important, too. It's also worth nothing that one should be mindful of what a benchmark actually measures. Writing a mere 100MB-1GB is quite a different proposition compared to a sustained write of, say, 15GB.

As I've mentioned a number of times in this thread, though, sequential write, once you go past a certain speed, doesn't matter much for a system drive. Random small block (4k) performance does and I'd like to see Apple improve from 2MB/s (basically around the same as HDD). Even my 7-year old X25-V with 40MB/s sequential write is capable of 15MB/s random 4K writes. That's around 7.5x as fast as HDD and is the reason why SSDs feel faster than HDDs.

 

[dotnet]

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!

You had to ruin a perfectly joyful debate by letting rationality seep into it

 

[inscrewtable]

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!

This is very true, but isn't there something odd about a 32gb iPhone 7 plus having a much slower disk storage write than a 32gb iPod?

 

[Hasukazu]

850 Pro uses higher quality/performance MLC NAND. As far as I'm aware, all the latest iPhone models use cheaper, lower performing TLC NAND.

That really hits the nail on the head, but only in the case of the 32GB models.

"We're starting to get too much heat for having only 16GB in the entry models, so let's increase that to 32GB. To make up for that, let's use low quality cheap ass memory controllers which are significantly slower. People who want a good memory performance will have to buy 128GB or 256GB."

 

[Tech_User]

With as much as Apple charges it's messed up they do this.
I never get the lowest storage but it's not about me and what I get.

 

[rui no onna]

That really hits the nail on the head, but only in the case of the 32GB models.

"We're starting to get too much heat for having only 16GB in the entry models, so let's increase that to 32GB. To make up for that, let's use low quality cheap ass memory controllers which are significantly slower. People who want a good memory performance will have to buy 128GB or 256GB."

Curious, I know Apple started using TLC NAND in 2014 with the iPhone 6/6+. Where they even using MLC NAND for the 16GB 6s/6s+?

In fairness to Apple, while the sequential write performance of the iPhone 7 32GB seem atrocious compared to higher capacities, it seems in-line with results from other manufacturers and is likely comparable to or better than the previous generation's 16GB model.

 

[derphelix]

This is very very normal for how solid storage works.

This makes totally sense.

It's 8 times slower? Doesn't that ring any bell?

Imagine you have a bathtub with one drain pipe which is able to hold 32L of water and has a diameter of 10cm. No you have a second bath tub with 8 x 10cm drain pipes holding 32L each, which sums up to 256L in total.
Now you pour in 4L of water in each bath tub. What do you expect to happen? That the first bath tub will be empty as fast as the 2nd? probably not...

 

[gnasher729]

Almost sounds like the 32 GB models are using some type of compressed memory algorithm.

That would be rather stupid when transferring videos. If you could compress them, they would be smaller. Unless Apple recognises that storing 4GB videos on a phone with a tiny screen isn't clever, and creates a video with lower resolution on the fly.
[doublepost=1476950729][/doublepost]Great news! iPhone 7 128 GB write speed is eight times higher than the 32 GB version!
[doublepost=1476950850][/doublepost]

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

4K video is _huuuuge_. Your 32 GB will be full in no time.

 

[Iconoclysm]

If I remember correctly, AnandTech wrote their own storage benchmark app so yes, I'd actually trust them more than GSMArena.

Which has nothing to do with the fact that they're not comparing the same things....

 

[rui no onna]

Which has nothing to do with the fact that they're not comparing the same things....

The question is how accurate are the numbers from PassMark? Because that's where the benchmark numbers posted on GSMArena come from.

AnandTech's sequential write shows ~150MB/s sequential write on the iPhone 7 256GB. Meanwhile, GSMArena reports 308MB/s for the iPhone 7 128GB. I get around the same PassMark scores on my iPhone 7 256GB but the disk tests finished in less than a second so I'm not sure how reliable those numbers are.

PassMark results are also inconsistent. On the same iPhone 7 256GB, I've had the following different results:

357MB/s write
907Mb/s read

283MB/s write
384MB/s read

315MB/s write
853MB/s read

 

[Hasukazu]

Curious, I know Apple started using TLC NAND in 2014 with the iPhone 6/6+. Where they even using MLC NAND for the 16GB 6s/6s+?

In fairness to Apple, while the sequential write performance of the iPhone 7 32GB seem atrocious compared to higher capacities, it seems in-line with results from other manufacturers and is likely comparable to or better than the previous generation's 16GB model.

No, it's not comparable at all. iPhone 7 32 GB has around 42 MB/s write speed: (Click for full size)


This makes the iPhone 6 (two generations back) two times faster, and the iPhone 6s (last generation) even a whopping four times faster!
So, even if compared to one and two generations back, the write performance of the iPhone 7 32 GB is still atrocious. Apple really dropped the ball there.

 

[rui no onna]

No, it's not comparable at all. iPhone 7 32 GB has around 42 MB/s write speed: (Click for full size)


This makes the iPhone 6 (two generations back) two times faster, and the iPhone 6s (last generation) even a whopping four times faster!
So, even if compared to one and two generations back, the write performance of the iPhone 7 32 GB is still atrocious. Apple really dropped the ball there.

I believe the iPhone 6 and 6s used in AnandTech's reviews are both 128GB models. I'm curious about performance of iPhone 6 and 6s 16GB. That's really the point of comparison for the iPhone 7 32GB.

 

[fbottone]

My 16gb iPhone 6s+:

 

[rui no onna]

My 16gb iPhone 6s+:

View attachment 667622

Thanks for this! So slower than or comparable to 32GB 7, basically (36MB/s vs 40MB/s).

 

[fbottone]

Yeah, seems like much todo about nothing with this whole story. My guess would be the 6 (non-s) are also similar.