1GB =/= 1000MB!!!

[Helepolis]

I tried using google today to do a few quick conversions of MB to GB. In the past its always been 1GB = 1024MB but now its showing 1GB = 1000MB. Even though storage capacities use the later measurement, i'm a little shocked to see it as the standard measurement that cannot be changed. Even top results use it as an answer, and 1024 comes after those. Have a I missed the memo and 1024 is no longer the standard outside of marketing?

 

/nerdrage

[RadiatingLight]

Yes. Google is right.

 

1 Gibibyte = 1024 Mebibytes.

1 Gigabyte = 1000 Megabytes.

[BlueChinchillaEatingDorito]

Difference between marketing and the actual backend.

[Nena Trinity]

The reason is that each memory module has slightly more so they always hit strange numbers like 1024mb or 512mb...

[RadiatingLight]

[RandomBK]

Over the past several years, people have mostly standardized on GB = 1000MB. Gibibyte is now used to denote 1024MB (https://en.wikipedia.org/wiki/Gibibyte).

[Guest]

1 minute ago, Nena360 said:

The reason is that each memory module has slightly more so they always hit strange numbers like 1024mb or 512mb...

2^10=1024

[crazysteve240]

1 minute ago, Helepolis said:

I tried using google today to do a few quick conversions of MB to GB. In the past its always been 1GB = 1024MB but now its showing 1GB = 1000MB. Even though storage capacities use the later measurement, i'm a little shocked to see it as the standard measurement that cannot be changed. Even top results use it as an answer, and 1024 comes after those. Have a I missed the memo and 1024 is no longer the standard outside of marketing?

 

/nerdrage

 

I think companies made it into 1000MB to make it more consumer friendly, since the extra 24MB or GB won't make a huge difference in terms of the entire disk. Regardless, computers still get information in 1024s.

[SpaceGhostC2C]

1 minute ago, RandomBK said:

Over the past several years, people have mostly standardized on GB = 1000MB. Gibibyte is now used to denote 1024MB (https://en.wikipedia.org/wiki/Gibibyte).

Only for HDDs/SSDs, though. No one would accept an 8GB stick of RAM with only 8,000MB in it.

[Wolther]

1 Gigabyte = 1000 Megabytes 

1 GiB = 1024 Mebibytes

 

Google is correct, there are two different terms that lead to confusion between the two since they're so similar 

[RandomBK]

Just now, SpaceGhostC2C said:

Only for HDDs/SSDs, though. No one would accept an 8GB stick of RAM with only 8,000MB in it.

True. We're in a weird transition period, where Gibibyte and friends are being used outside of the storage industry as well. Personally, I would like everyone to standardize on 1GB = 1000MB like the other SI units, and 1GiB = 1024MiB for binary

[zMeul]

11 minutes ago, Helepolis said:

I tried using google today to do a few quick conversions of MB to GB. In the past its always been 1GB = 1024MB but now its showing 1GB = 1000MB. Even though storage capacities use the later measurement, i'm a little shocked to see it as the standard measurement that cannot be changed. Even top results use it as an answer, and 1024 comes after those. Have a I missed the memo and 1024 is no longer the standard outside of marketing?

 

/nerdrage

I know the feeling ...

they went full retard and devised a new scale for mass storage that multiplies by 1000 - they should not call this shit "byte" 

1024 multiplication is still valid for RAM

[SpaceGhostC2C]

Just now, RandomBK said:

True. We're in a weird transition period, where Gibibyte and friends are being used outside of the storage industry as well. Personally, I would like everyone to standardize on 1GB = 1000MB like the other SI units, and 1GiB = 1024MiB for binary

I don't think that's a good idea: digital information is stored in base 2, whether we like or not. Hence, we can have round exact numbers of base 2 quantities (like 4096 bytes = 4KB), but if we start relabeling everything according to base 10 we'll end up with 13.22GB of RAM and so on (just like HDDs/SSDs not only are now reported in "fake" GBs, but they don't even have the stated number of "GBs", since it's a rounded-up approximation anyway).

[Nena Trinity]

13 minutes ago, deXxterlab97 said:

2^10=1024

https://qph.ec.quoracdn.net/main-qimg-39bab7143aae5e81f10f1188f68e697e-c

Each of those black squares...

[RandomBK]

7 minutes ago, SpaceGhostC2C said:

I don't think that's a good idea: digital information is stored in base 2, whether we like or not. Hence, we can have round exact numbers of base 2 quantities (like 4096 bytes = 4KB), but if we start relabeling everything according to base 10 we'll end up with 13.22GB of RAM and so on (just like HDDs/SSDs not only are now reported in "fake" GBs, but they don't even have the stated number of "GBs", since it's a rounded-up approximation anyway).

I think the change should be towards using the IEC/binary units, i.e. 4096 bytes = 4KiB.

 

From a broader perspective, it has always been very confusing that SI prefixes like GB were base 2 when talking in bytes.  After all, 1 KHz = 1000 Hz, 1 kilowatt = 1000 watts, 1 Km = 1000 m, etc. Why should bytes be the exception?  Better move to a different unit altogether.

[WMGroomAK]

27 minutes ago, Helepolis said:

I tried using google today to do a few quick conversions of MB to GB. In the past its always been 1GB = 1024MB but now its showing 1GB = 1000MB. Even though storage capacities use the later measurement, i'm a little shocked to see it as the standard measurement that cannot be changed. Even top results use it as an answer, and 1024 comes after those. Have a I missed the memo and 1024 is no longer the standard outside of marketing?

 

/nerdrage

Gamers Nexus did a half way decent little video about the conversion and confusion between Gibibytes, Gigabytes, Mebibytes and Megabytes a couple of weeks ago with the math conversion.  Think it was titled 'Explaining "Stolen" Gigabytes in Windows'.

 

 

 

[CBojorges]

1 hour ago, Nena360 said:

The reason is that each memory module has slightly more so they always hit strange numbers like 1024mb or 512mb...

Those aren't strange numbers. They are powers of 2.

 

2^9=512

[straight_stewie]

1 hour ago, Helepolis said:

I tried using google today to do a few quick conversions of MB to GB. In the past its always been 1GB = 1024MB but now its showing 1GB = 1000MB. Even though storage capacities use the later measurement, i'm a little shocked to see it as the standard measurement that cannot be changed. Even top results use it as an answer, and 1024 comes after those. Have a I missed the memo and 1024 is no longer the standard outside of marketing?

It's about moving away from the internals of computers. One of the selling points of them is that users don't really need to be concerned with the inner workings or the how's and why's in order to use them.

The numbers follow a very specific pattern; they are powers of two. 

But why powers of two? Well, it's because of how a computer works. A single bit can be either a 0 or a 1 (aka off or on). This means that a single bit can represent two values. This is known as base 2 numbers. The number of permutations (or in other words, number of different unique values that can be expressed) with n digits in base 2 is given by the function 2ⁿ.

So what does this mean for memory. Let's start at the bottom:

1. A single bit means that we can address two memory locations.
2. Two bits means that we can address four memory locations.
3. Three bits means that we can address 8 memory locations
4. Four bits means that we can address 16 memory locations.
5. Eight bits gives us 256 addresses.
6. Nine bits gives us 512 addresses.
7. And finally, the important lesson. Ten bits gives us 1024 addresses. The technically correct name for this is a "kibibyte" although it is also known as "1k".

You will find that computers are largely based on powers of two, mostly because 99% of what computers do is to generate, manipulate, and make decisions based on permutations of 2 to the n^th power, where n is normally the "width" of a computer, such as 8 bit, 16 bit, 32 bit, or 64 bit.

[KuJoe]

GiB can DIAF.

[Nena Trinity]

39 minutes ago, CBojorges said:

Those aren't strange numbers. They are powers of 2.

 

2^9=512

That is strange! o3o

[MoonlightSylv]

2 hours ago, RadiatingLight said:

Yes. Google is right.

 

1 Gibibyte = 1024 Mebibytes.

1 Gigabyte = 1000 Megabytes.

THANK YOU

[manikyath]

the issue lies within that the prefix standard "prescribes" powers of 10, and that hardware pretty much "exists within powers of 2"

 

it's mostly a matter of "no, manufacturers arent gonna write freaking gibibyte on their ram sticks", and it's down to the person picking the parts to understand what is what.

[straight_stewie]

Just now, Nicholatian said:

This is a bit of an oversimplification, actually. The bit width you are referring to has two main contexts that can apply to it in the world of computer hardware: first, the width of each CPU instruction, and second, the width of memory addresses. Computers are not, by nature, “32-bit” or “64-bit”; it’s a misnomer that occurred because Intel and AMD have kept the width of x86 instructions and the width of memory addresses in x86 the same at most every point in the architecture’s history.

Obviously I wasn't going to write a text book here. But there are even some simplifications in your response. Instruction width largely doesn't matter outside of cost (which is a very complicated subject in and of itself). The important factors, atleast as I was/am taught, are address bus width and data bus width. It just so happens that in modern Von Neumann machines the data bus and address bus are usually the same size.
 

More directly, the width of instructions is a non sequiter. It can theoretically be any width, regardless of data or address width. It can even be dynamically sized, as most modern Intel chips support (with instruction set extensions, especially VLIW). 

And computers can be whatever data bit length and address bit length their designers decide upon. It just so happens that most designers, even those that do not implement the x86 instruction set, settle on multiples of 8 bits for data, and multiples of 8 bits greater than or equal to 16 bits for addresses (even in 8 bit data machines). Throughout history there have been many machines, and even some very important ones, that had weird bit lengths all around. 

 

8 minutes ago, Nicholatian said:

If you must term a given CPU as X-bit, then the usual convention is to go by the width of a single instruction. Many people will contest that though as several RISC CPUs have simplified instruction set modes where the instruction width is, say, 16-bit instead of 32-bit… but those 16-bit instructions can still access 32 bits of memory. So it’s best to be explicit with what you’re talking about in any case.

This is completely false. The normal convention is most commonly data bus width, and secondly; address bus width. As I alluded to previously, instruction width can even be dynamic (infact it is in modern x86 chips).

 

 

But ultimately, you are correct. The 32/64 bit thing is largely a result of the popularity of x86. This is something that I have mixed feelings about. However, I still believe that most designers would still settle on data bus widths that are multiples of 8 bits largely because of heirarchical design and the fact that a byte is 8 bits in length.

[manikyath]

2 minutes ago, straight_stewie said:

owever, I still believe that most designers would still settle on data bus widths that are multiples of 8 bits largely because of heirarchical design and the fact that a byte is 8 bits in length.

i guess that it's mostly a matter of "more sensible backwards compatibilty"

 

if you'd, for example, make a chip that can do both 16-bit, and something newer with "more bits", it'd make more sense to go 32 than for example 24, because you can (as is a thing with x86) split the regs in "low" and "high" 16-bit parts, instead of having an 8-bit piece kinda just sit there.

[straight_stewie]

9 minutes ago, manikyath said:

i guess that it's mostly a matter of "more sensible backwards compatibilty"

 

if you'd, for example, make a chip that can do both 16-bit, and something newer with "more bits", it'd make more sense to go 32 than for example 24, because you can (as is a thing with x86) split the regs in "low" and "high" 16-bit parts, instead of having an 8-bit piece kinda just sit there.

Exactly. And there are more reasons than that as well, such as hierarchical design patterns. For example, if you give me a register that can hold 1 bit, I can string two together to hold 2 bits, 2 of those together to hold 4 bits, two of those together to hold 8 bits, two of those together for 16 bits, and so on. In this fashion, I can largely ignore how the smaller parts work together, and worry about making the whole, rather than the pieces. It's not necessarily multiples of 8, it's more about powers of 2, yet again. It just so happens that all powers of two greater than 8 are multiples of 8. and there just so happens to be 8 bits in a byte, and nearly all modern computers are byte addressable.

Actually, mentioning that last part reminds me of something else. Every power of two is a multiple of all powers of two less than that value.