480/512GB SSD question

[darkfalz]

Do these differing capacities in SSD tiers only relate to companies using GB vs GiB or is it because some vendors choose to reserve more NAND for bad blocks?

 

[hawtdawg]

I think it has to do with factory over-provisioning. For example; I have a Sandisk Extreme Pro 480GB and a Samsung 850 Pro 512 GB. The Sandisk has OP built in and doesn't even give OP options in their software, whereas the Samsung software recommends you set about 10% of the space for OP. The end result is that they end up with a comparable amount of usable space (actually a tad more with the Sandisk) though theoretically you could partition all of the Samsung drive, while the option isn't there for the Sandisk.

edit: here's a picture to maybe better explain. (Disk 2 is the Sandisk and disk 3 is the Samsung)

 

[darkfalz]

Okay, so one does it by default, the other gives you the option. Cheers.

 

[darkfalz]

A minor correction to anyone who may read: my 256/512 do indeed turn out to be 240/480 respectively. These things do genuinely have 256/512 GB of flash memory, but 7% lost to the OP. Appears the OP your Samsung tool offers is additional to the built in OP.

 

[Cerb]

Do these differing capacities in SSD tiers only relate to companies using GB vs GiB or is it because some vendors choose to reserve more NAND for bad blocks?

The GBs are the same size, and they have the same size to work from (512GiB total NAND, in this case).

They get used for added over-provisioning, added redundancy, and caches, varying by drive (the Sandisk Ultra II and Extreme Pro do all three of those, for example, with the reserved the flash).

 

[darkfalz]

It's strange to see so often a repeated myth that the GB to GiB was not to be confused with the 7% lost to OP. Kingson's own page on OP debunks this.

The GB/GiB let traditional HD manufacturers claim more GB than they actually were offering. This is true. But SSD capacity is in powers of 2 and has nothing to do with the base 10 "GB" long established in use by HDD manufacturers. 512 GB is the actual size. 480 GB is the available size after OP. It's merely coincidence really that they are fairly close in difference to the GB/GiB.

 

[Cerb]

A 512GB SSD has 512GiB of NAND, and 512GB of user-visible space ("available space"). A 480GB drive has the same 512GiB flash, but more of it than the ~7% has been taken, for whatever reason, and only offers 480GB of user-visible space. In SF drives, and the Crucial M500, it's entirely used by RAID-like redundancy. In others, it has been plain added OP. In others, it has been added OP and fast caches (M600, 840 Evo, 850 Evo, Ultra II, Extreme Pro...). Some, like Sandisk's Ultra Plus, and most new Crucials, use a bit less over-provisioning, and give the "full" size, even with redundancy and caching added in that spare space.

The advertised sizes are equal to those found in HDDs, and has everything to do with the base-10 values. A 512GB drive will have approximately 512,000,000,000B. A 480GB drive will have approximately 480,000,000,000B.

512 GB is the actual size. 480 GB is the available size after OP. It's merely coincidence really that they are fairly close in difference to the GB/GiB.

Not at all. 512GB is the actual size, which is equal to 476GiB. A 480GB drive has 447GiB. It's the same as a 500GB HDD giving you 465GiB.

 

[darkfalz]

Not at all. 512GB is the actual size, which is equal to 476GiB.

When it comes to SSDs 512 GB is the same as 512 GiB. It's powers of 2 (and therefore, powers of 1024).

 

[Cerb]

When it comes to SSDs 512 GB is the same as 512 GiB. It's powers of 2 (and therefore, powers of 1024).

No, they aren't. GBs are the same between HDDs and SSDs.

That's an 840 Pro 512GB. As you can see, about 512 billion bytes (IE, 512GB), and 476GiB (Windows' GB is GiB). A 480GB drive will have 447GiB. A 500GB drive will have 465GiB. Whether HDD or SSD, it doesn't make a difference. Your space in GiB is (10^9/2^30) the marketing space in GB. For TB, it will be (10^12/2^40). That NAND is typically done in powers of 2, and redundancy has traditionally been done in coarse stripes (M500's 1:16 RAIN, and SF's RAISE), makes it convenient to use sizes like 256GB, 512GB, 240GB, 480GB, and so on, but you don't get any different amount of space per GB.

 

[CiPHER]

Do these differing capacities in SSD tiers only relate to companies using GB vs GiB or is it because some vendors choose to reserve more NAND for bad blocks?

No.

The NAND is like DRAM chips: always power of 2. The chips themselves are described in gigabits, like 128Gbit, which is 16GiB. The NAND chips are usually grouped in a power of 2, usually 8 chips or 16 chips. Earlier Intel SSDs had 10 channels and this resulted in different capacities like 40, 80, 160, 320, 640, etc. Usually with 8 channels it is 32, 64, 128, 256, 512.

When i say 512 i mean 512GiB, not 512GB. The latter (GB) is 512.000.000.000 bytes while the former (GiB) is 512 * 1024 * 1024 * 1024 bytes.

Now, assume an SSD with 512GiB of NAND. Now assume the following capacities:

512GiB -> almost impossible because the SSD has no spare space for anything; extremely simple USB sticks could do it, but impracticle as replacing bad NAND pages is impossible.

512GB -> the most usual; the difference between GiB and GB (6,8%) is used for overprovisioning and other stuff (reserve NAND, internal storage for SMART/mapping tables, etc).

But you also see 480GB and 500GB sometimes. Or 50/60GB instead of 64GB, or 100/120GB instead of 128GB. What is the story behind this?

Well, aside from overprovisioning, many modern SSDs also have parity information to protect against 'bad sectors' aka unreadable NAND pages. One unreadable NAND page at the wrong location and your whole SSD could be bricked because the mapping tables are damaged. Protection against bitrot is pretty much mandatory as you continue producing the NAND at smaller process technology (50nm->34nm->22nm->16nm).

As you may know, modern SSDs are like a 16-way RAID0 device of independent NAND flash. But in fact, it is not RAID0 interleaving that is used, but more like RAID3 or RAID5.

The M500 had 1/16-level bitcorrection, meaning: a RAID5 of 16 NAND devices. This also means that only 15/16 of the capacity can be used for storage. The 1/16 is exactly the difference between 240GB and 256GB; 480GB and 512GB. This has nothing to do with overprovisioning!

Now the M550/MX100 and up has 1/128-level bitcorrection; and thus lower overhead. This overhead can be contained in the usual difference between GB and GiB. This does mean, however, that these SSDs have fewer overprovisioning than the M500. It also has better good protection against bitrot; probably the Micron 16nm NAND has matured enough so the raw Bit-Error-Rate (BER) is sufficient to use only 1/128-level bitcorrection instead of the former 1/16-level.

Hope this answers your question.

 

[hojnikb]

No.

The NAND is like DRAM chips: always power of 2. The chips themselves are described in gigabits, like 128Gbit, which is 16GiB. The NAND chips are usually grouped in a power of 2, usually 8 chips or 16 chips. Earlier Intel SSDs had 10 channels and this resulted in different capacities like 40, 80, 160, 320, 640, etc. Usually with 8 channels it is 32, 64, 128, 256, 512.

Thats not true at all. NAND can be any size you desire, its just a industry norm to use powers of 2.

A good case in point is the new 850pro, which uses 86Gbit dies, which are not a power of 2.

 

[CiPHER]

Fair enough.