Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Why are storages powers of 2?

S

samiznaetekto

macrumors 65816
Original poster
Dec 26, 2009
1,016
24
Why is it that it's always 16GB, 32, 64, 128 for phone storage and such? Why not, say, 20, 30, 50, 75?

The amount of data you need to store does not grow exponentially with time, more like linearly. And in real life, we don't double each step and don't buy things in powers of 2. Is 10-pound bag of potatoes too small for you? Buy 15.

So, why?
 
Data is

Kept in he firm of 1/0. It's either a yes or no form. 2 choices means it's a base of 2.
 
Everything is a base of 2

Data has to be in some form of 2 to a nth power. That's why its 16/32/64/128.... Even a 1 tb of data is not going to be a perfect 1 tv it will be 1 tb - the remainder of 2 to a nth power that's closest to 1 tb.
 
Superjams said:
Data has to be in some form of 2 to a nth power. That's why its 16/32/64/128.... Even a 1 tb of data is not going to be a perfect 1 tv it will be 1 tb - the remainder of 2 to a nth power that's closest to 1 tb.
Click to expand...

Nope. Hard drives, SSDs can be of any size, 120, 160, 500, etc. Even those 16, 32, etc. are not actual powers of 2, they are 16 000 000 000, etc.

----------
Superjams said:
Kept in he firm of 1/0. It's either a yes or no form. 2 choices means it's a base of 2.
Click to expand...

You're confusing binary representation with storage amount. You can have a file with size 123 bytes, it doesn't have to be 128.

----------
Jessem1133 said:
Why not 15, then 27, then 31, then 50?
Click to expand...

Indeed, why not?

----------
Jessem1133 said:
Why not 15, then 27, then 31, then 50?
Click to expand...

Indeed, why not?
 
samiznaetekto said:
Nope. Hard drives, SSDs can be of any size, 120, 160, 500, etc. Even those 16, 32, etc. are not actual powers of 2, they are 16 000 000 000, etc.
Click to expand...
SSD capacities are actually powers of 2. They just reserve a part of it for wear-leveling and other housekeeping tasks, so not all of it is accessible to the end user. E.g. a 120GB SSD typically has 128GB of physical flash memory.
Indeed, why not?
Click to expand...
Solid-state memory comes typically in powers of 2 because they use binary addressing. It's theoretically possible to make memory chips that have non-power-of-2 capacities (i.e. 2^n-x), but that would mean that x bytes of the possible address space would not actually have memory cells and you could thus set the address bus to point to invalid memory. Also, chip designers hate waste so they would not want to underutilize the address pins on their chips. ;)

Hard drives can have other sizes because they are not addressed in binary (but rather in tracks and sectors).
 
Rigby said:
SSD capacities are actually powers of 2. They just reserve a part of it for wear-leveling and other housekeeping tasks, so not all of it is accessible to the end user. E.g. a 120GB SSD typically has 128GB of physical flash memory.
Solid-state memory comes typically in powers of 2 because they use binary addressing. It's theoretically possible to make memory chips that have non-power-of-2 capacities (i.e. 2^n-x), but that would mean that x bytes of the possible address space would not actually have memory cells and you could thus set the address bus to point to invalid memory. Also, chip designers hate waste so they would not want to underutilize the address pins on their chips. ;)

Hard drives can have other sizes because they are not addressed in binary (but rather in tracks and sectors).
Click to expand...

Thanks for the explanation. But even with RAM, you can mix and match, e.g. have 8GB in one slot and 4 in the other, total 12. My Note 4 has 3GB of RAM. The SSDs are also combos of several chips, not one, so in principle, mix and match should be possible.

And if 120 is actually 128 as you say, what's the physical size of 160GB SSD? 256? That's a lot of waste for housekeeping.
 
samiznaetekto said:
Why is it that it's always 16GB, 32, 64, 128 for phone storage and such? Why not, say, 20, 30, 50, 75?

The amount of data you need to store does not grow exponentially with time, more like linearly. And in real life, we don't double each step and don't buy things in powers of 2. Is 10-pound bag of potatoes too small for you? Buy 15.

So, why?
Click to expand...

Because Apple builds them that way.
 
samiznaetekto said:
Thanks for the explanation. But even with RAM, you can mix and match, e.g. have 8GB in one slot and 4 in the other, total 12. My Note 4 has 3GB of RAM. The SSDs are also combos of several chips, not one, so in principle, mix and match should be possible.
Click to expand...
Yes, that is true. But the individual memory chips and modules all have capacities in powers of 2 for the above mentioned reasons. In theory Apple could offer e.g. a 96 GB (composed of 64+32 GB memory chips) iPhone as well, but they probably want to use only a single memory chip for cost and space reasons.
And if 120 is actually 128 as you say, what's the physical size of 160GB SSD? 256? That's a lot of waste for housekeeping.
Click to expand...
True. What I wrote above is an over-simplification. The total flash memory will be composed of chips with a power-of-2 capacity, maybe 3*64=192GB. It gets more complicated because drive manufacturers typically give the capacity in GB (i.e. 10^9 bytes) rather than binary GiB (i.e. 2^30).
 
Jessem1133 said:
Why not 15, then 27, then 31, then 50?
Click to expand...

...because it's binary

2^0=1
2^1=2
2^2=4
2^3=8
2^4=16
2^5=32

so on...

----------
Rigby said:
SSD capacities are actually powers of 2. They just reserve a part of it for wear-leveling and other housekeeping tasks, so not all of it is accessible to the end user. E.g. a 120GB SSD typically has 128GB of physical flash memory.
Solid-state memory comes typically in powers of 2 because they use binary addressing. It's theoretically possible to make memory chips that have non-power-of-2 capacities (i.e. 2^n-x), but that would mean that x bytes of the possible address space would not actually have memory cells and you could thus set the address bus to point to invalid memory. Also, chip designers hate waste so they would not want to underutilize the address pins on their chips. ;)

Hard drives can have other sizes because they are not addressed in binary (but rather in tracks and sectors).
Click to expand...
Yup

Then keep in mind OP, storage is technically gibibits, not gigabits as manufactures advertise as
 
Lawd, OP I hope you don't get into networking. Subnets and supernets will kill you Lolol
 
I think it's because of memory addressing and - depending on OS and drivers, represent sections of your disk conveniently.

SSDs coming in crazy sizes is a different reason! So, when you buy a 480GB SSD, you might actually be getting a device with 512GB of flash memory on it (or maybe more). The extra flash is saved for buffer, and flash block failures depending on the model.
 
JoeG4 said:
I think it's because of memory addressing and - depending on OS and drivers, represent sections of your disk conveniently.

SSDs coming in crazy sizes is a different reason! So, when you buy a 480GB SSD, you might actually be getting a device with 512GB of flash memory on it (or maybe more). The extra flash is saved for buffer, and flash block failures depending on the model.
Click to expand...

You're right about overprovisioning. But it's not necessarily in powers of 2. Here's an example:

http://www.edn.com/design/communications-networking/4390180/Dissecting-the-Intel-710-Enterprise-SSD-

200GB drive consisting of 20 16GB chips. Neither 200, nor 320 are powers of 2.

Again, most Android flagships have 3GB of RAM (either 3x1 or 2+1), not 2 or 4.

From consumer perspective, it would make sense for storages to have linear increments, not exponential.
 
samiznaetekto said:
Why is it that it's always 16GB, 32, 64, 128 for phone storage and such? Why not, say, 20, 30, 50, 75?

The amount of data you need to store does not grow exponentially with time, more like linearly. And in real life, we don't double each step and don't buy things in powers of 2. Is 10-pound bag of potatoes too small for you? Buy 15.

So, why?
Click to expand...

As many have stated, it's because computers are working with a base 2 system. You could very well have a 15GB device, but it would have to be 8GB + 4GB + 1GB + 1GB + 1GB. That would be five separate chips that would have to be manufactured, as opposed to one chip for 16GB of storage (not to mention that multiple chips take up more physical space).

Also, the cost of data storage (or the cost of computing) decreases exponentially over time (Moore's Law), so the cost of producing a 256GB SSD will eventually be equal to or less than the cost of producing a 128GB SSD.
 
Here's an explanation of chip stacking:

http://www.samsung.com/global/busin...ort/package-info/package-datasheet/flash#none

So, while each die in the stack necessarily has a power-of-2 capacity, why their number must be a power of 2? Why not like the SSD example I gave above (20 chips)?

----------
bizack said:
Also, the cost of data storage (or the cost of computing) decreases exponentially over time (Moore's Law), so the cost of producing a 256GB SSD will eventually be equal to or less than the cost of producing a 128GB SSD.
Click to expand...

I think Moore's law stopped working a few years ago. Every exponential growth is bound to hit the ceiling and stop working.
 
samiznaetekto said:
Here's an explanation of chip stacking:

http://www.samsung.com/global/busin...ort/package-info/package-datasheet/flash#none

So, while each die in the stack necessarily has a power-of-2 capacity, why their number must be a power of 2? Why not like the SSD example I gave above (20 chips)?

----------



I think Moore's law stopped working a few years ago. Every exponential growth is bound to hit the ceiling and stop working.
Click to expand...

You'll notice in Samsung's chip stacking spec, that they're stacking in powers of 2. That's because memory addresses are based on powers of 2 (16-bit, 32-bit, 64-bit). The first example sites 16 stacked dies. If you only stacked 15 dies, you're essentially 'wasting' your high order bit (which would address the 16th stacked die).

If Moore's law stopped years ago, the iPhone 6/6 Plus would cost Apple *a lot* more than ~$200-$300 per phone to produce. Computers haven't been around that long... we're not even close to approaching the asymptote of that growth curve.
 
bizack said:
You'll notice in Samsung's chip stacking spec, that they're stacking in powers of 2. That's because memory addresses are based on powers of 2 (16-bit, 32-bit, 64-bit). The first example sites 16 stacked dies. If you only stacked 15 dies, you're essentially 'wasting' your high order bit (which would address the 16th stacked die).
Click to expand...

So, why SSDs have no problem with wasted address bits (as in 20-chip example above)?
 
samiznaetekto said:
So, why SSDs have no problem with wasted address bits (as in 20-chip example above)?
Click to expand...

SSDs are configured differently than the storage in an iPhone. The 6 and 6+ have a single flash memory chip installed in them, so whatever capacities that those chips are manufacturered is what Apple can use.
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back