Jump right to Contenders for purchasing.

An SSD is a Solid State Disk Drive (also known as a Solid State Disk), a new kind of storage device that has entered the mainstream in 2014. The disk drives you’ve been used to all along are regular HDD’s. A regular HDD contains a round disk called a platter on which the bits of information are written. It contains a spindle that spins, or “drives”, the platter so that the “head” can read and write information from and to the platter as it spins underneath. It is a “Hard” Disk Drive (HDD) to differentiate it from “floppy” (or flexible) platters that were used in the past (ending in the 1990s).

A Solid State “Drive” doesnt “drive” anything, and doesnt contain a disk or platter. So it’s a misnomer. But since it is a replacement for a HDD it is known by the term Solid State “Drive”.

On the inside it is basically a large USB Drive, configured to work as an internal hard disk drive (HDD).

It’s a drop-in replacement for an internal hard drive. With some caveats which i’ll discuss.

It is a drop-in replacement for 2.5″ drives. SSDs tend to be sized the same as 2.5″ HDDs. Ie laptop hard drives. Also, they have a SATA port and interface, so they will plug into the current standard drive interface in nearly all PCs today. You might want v3 of SATA (SATA IIII aka SATA 6.0Gbits/s) to take advantage of the SSD’s additional bandwidth capabilities, but it will work with the previous STAT II installed in most PCs that are a few years old.

So, you can pull out your old drive, and plug in your SSD, and it will work.

Well, sorta.

3.5″ drives, or desktop drives

The first problem is if you have a desktop. SSDs are 2.5″ so they plug right into laptops. But most desktop drives are 3.5″. So you’ll need a physical adapter. The adapter is 3.5″ on the outside, and 2.5″ on the inside. You plug your 2.5″ drive (eg, your SSD) into the adapter (or screw the adapter onto your 2.5" drive), and then the adapter plugs into your desktop PC. Sometimes you need to get a special one from the PC maker.

Erase-before-Writing Requirement

The other 3 issues:

• wear leveling
• write amplification
• Partition Alignment

require some background info to be able to talk about them.

It’s all about how writes happen. Reads are easy and just work.

Here are some key facts to keep in mind, which are crucial to understanding what’s gonig on:

• Data is written 4KB at a time. These are typically called pages.
• An SSD can NOT overwrite data. It has to erase it first. So to write, it has to perform a sequence of:
  1. erase
  2. write
• Data is erased 256KB at a time.
• A brand new SSD is empty, ie fully erased.
• Most file systems (developed with HDDs in mind) when they delete files, they dont erase them. They just mark the spots as available for overwrite and leave the old data there. This is problematic for SSDs, cuz that old data DOES have to be erased. More below.

These facts mean you write 1 page at a time (4KB), but you erase 64 pages (256KB) at a time.

In the picture below, you write the blue block. But if you want to erase, you have to erase a BUNCH of adjacent 4KB blocks simultaneously.

https://en.wikipedia.org/wiki/File:NAND_Flash_Pages_and_Blocks.png

(Note that in the picture above, to save space, only 9 x 4KB pages are shown, but in reality there would be 64 x 4KB pages in each 256KB block.)

So, for a new SSD, everything is golden, and writes of only 4KB happen everywhere into free (erased) pages and we’re fine.

The problem comes when we run out of space and have to re-use some old pages.

The OS filesystem will do this. It will have some disk pages that contain data that it knows is from a file that has been deleted. It will instruct the disk (SSD in this case) to overwrite this space.

But the SSD can not simply overwrite, it has to erase first, then overwrite.

And to erase, it has to erase 256KB (64 pages). If ALL of those 64 pages are unused (aka stale), and the SSD knows it (the OS filesystem knows more than the SSD does), then it can simply erase all 64 pages. But more than likely, some of the 64 pages contain real data that must be preserved.

So…the SSD has to copy the good pages to a safe place, erase the whole 256KB block, copy back the good pages, then write the new page that was requested.

In reality, what the SSD really does is pick another entire 256KB block somewhere that’s entirely empty, and copy the good pages to it, re-map where those good pages data is, then write the new page that was requested, keeping in mind its new mapping.

Either way this process is SLOW and would essentially reverse all of the speed benefits of SSDs in the first place.

In the pic below, you see 4 pages (A-D) written.

Then 4 more pages are written (E-H).

Then the original 4 are “overwritten”, meaning the OS has written new versions (A’-D’), and told the SSD that A-D are invalid. So the SSD knows A-D are now unused.

Until an erase happens, the supposedly free pages A-D can not be used by the SSD. So to re-use A-D, the SSD has to copy the good pages (E-H and A’-D’) somewhere else first so it can erase the block:
(click to enlarge) https://en.wikipedia.org/wiki/File:Garbage_Collection.png

The other problem with this whole erase & write issue is that it causes extra writes (re-writing perfectly good data) to happen. These extra writes wear out the SSD faster. The extra writes are called “write amplification” (see below).

In fact without remediation an SSD would die in a month or 2. So all SSDs are remediated.

Write Amplification

Because of the erase-before-writing requirement (see above), there’s all kinds of re-writing of good data going on. This means, that if you measured the writes going into the SSD at its interface, you would measure a certain amount of writes, but if you looked at what was going on inside the SSD, you’d see a lot more writing going on.

More writes going on inside than the OS asked for is called write amplification.

Garbage Collection

It would be impractical for the SSD to wait for a write command from the OS and then do its erase-then-write. That would be too slow, and no one would buy SSDs. So, instead, SSDs go about doing proactive garbage collection. This is a process by which they move data around creating entire blocks which are all unused data. Then they can erase those blocks when the SSD is idle, and be ready for a speedy write when asked for by the OS.

They also re-map pages. That means the OS asked me, the SSD, to write data at A but i am going to put it at B, and remember the mapping from A to B, so that when the OS asks to read the data at A, i know where to get it.

That’s another way to keep a full set of 64 pages of good data in a block, and keep other blocks completely empty.

All good for speed, but all this proactive activity creates more re-writes of good data, and thus increases “write amplification”.

Reference: https://en.wikipedia.org/wiki/Write_amplification

Wear Leveling

So, another limitation of SSDs (they are not quite a silver bullet, not quite the best thing since sliced bread), is that a given bit of storage can only be re-written 1000 times, for a low end device. Up to 100k for a very hi-end (and expensive!) device. The amount of times you can write a cell is called “write endurance”.

You could easily exhaust that resource in a month of windows running on a drive. There’s lots of things that windows overwrites constantly.

So…

SSDs try to ensure that each bit is written the same number of times, or as close as they can get.

But, some data, like OS files, or maybe your pictures and videos, are written to the disk and just stay there, for YEARS. Unchanged. If SSD did nothing, those bits would have 1 write cycle, while other bits have 100s and 1000s and the drive would die.

So, SSDs move static data around, so it can write some of that rapidly changing data to the fresh bits.

But, these extra writes increase write amplification. In the end, however, it’s a net increase in life for the drive.

Reference: https://en.wikipedia.org/wiki/Wear_leveling

Partition Alignment

The OS tends to read and write X-size blocks. You really want those to be aligned to the natural alignment of the SSD.

If not, then the OS block will overlap TWO SSD blocks. So when the OS writes that block, you will have to write (and thus erase) TWO SSD blocks. that’s wasteful, so align your OS.

That means your partition has to be aligned. I’m not sure how to do this yet.

OverProvisioning

That’s when you say the drive has a certain size, say 500GB, but it really has more, say, 600GB, which would be an extra 20%. The SSD uses this space internally to perform writes with erase, and to do garbage collection, and to move around data proactively to evenly distribute wear (“wear leveling”).

SLC, MLC, TLC

SLC – Single Level Cell

Means 1 bit per cell.
The most expensive.
The most number of writes, 100,000 range, lifetime

MLC – Multi Level Cell

Means “multi” bits per cell, but in practice it is almost always 2.
Mid-priced.
Middle number of writes, 10,000 – 30,000, lifetime

Altho, http://www.tomshardware.com/reviews/600-pro-ssd-review,3498-4.html said the MLC’s will do only ~5,000 P/E cycles, and something called “eMLC” (“e” for enhanced, i would guess) could do ~25,000-30,000 P/E cycles.

TLC – Triple Level Cell

Means triple (3) bits per cell.
The least expensive.
The least number of writes, 1,000 – 3,000, lifetime

Altho, http://www.tomshardware.com/reviews/600-pro-ssd-review,3498-4.html said the MLC’s will do only ~5,000 P/E cycles, and something called “eMLC” (“e” for enhanced, i would guess) could do ~25,000-30,000 P/E cycles.

How to optimize

SSD Speed Tweks at speedguide.net <http://www.speedguide.net/articles/ssd-speed-tweaks-3319 >

SSD Optimization Guide from TheSSDReview.com <http://www.thessdreview.com/ssd-guides/optimization-guides/the-ssd-optimization-guide-ultimate-windows-8-edition/>

Contenders

Samsung EVO 500GB	NO: Has TLC so many fewer writes (lower write endurance). $349.99 maybe free shipping at TigerDirect on 3/12/2014 The 840 EVO is missing power loss protection, cross-die redundancy – From: SSD Deathmatch: Crucial’s M500 Vs. Samsung’s 840 EVO – at tom’sHARDWARE.com <http://www.tomshardware.com/reviews/crucial-m500-1tb-ssd,3551-14.html> Reference: http://www.thessdreview.com/our-reviews/samsung-840-evo-msata-1tb-ssd-review-unheard-msata-capacity-along-turbowrite-great-security
Crucial M500 WINNER – March, 2014	480GB $239.99 at crucial.com on 3/12/2014. Is it 512GB factory overprovisioned? YES 1.2 million hours mean time to failure (MTTF) Write Endurance: 72TB total bytes written (TBW), equal to 40GB per day for 5 years
Seagate 600 480GB	They also have a Pro version. MLC. Is it 512GB factory overprovisioned? I believe NO since the Pro verrsion has a 400GB factory overprovisioned and a 480GB non-factory overprovisioned. I assume this one is also not factory overprovisioned as it’s the consumer market. $249.99 Free Shipping at TigerDirect 3/12/2014. Extensive Specs Nonrecoverable read errors, max : 1 LBA per 10^16 bits read 10^16 bits = 10^15 bytes, approx. 1GB = 10^9 bytes = 10^10 bits. so 10^16bits = 10^6GB which is 1million GB. This drive is 500GB, so you get 1 bit error every 2 million times you re-write the entire drive. Not bad at all. Annualized Failure Rate (AFR) : 0.58% Endurance : 40GBs (max capacity) host writes per day Limited Warranty With Media Usage: This warranty is based on the shorter of term and endurance usage of the drive: 36 months or 73 (max capacity) TBW (total TeraBytes written) or 73TB TBW (total bytes written) whichever comes first. That’s writing the entire drive 146 times over. Typical Data Retention with Power removed (at 40°C) : 12 months (Note: As NAND Flash devices age with use, the capability of the media to retain a programmed value begins to deteriorate. This deterioration is affected by the number of times a particular memory cell is programmed and subsequently erased. When a device is new, it has a powered off data retention capability of up to ten (10) years. With use the retention capability of the device is reduced. Temperature also has an effect on how long a Flash component can retain its programmed value with power removed. At high temperature the retention capabilities of the device are reduced. Data retention is not an issue with power applied to the SSD. The SSD drive contains firmware and hardware features that can monitor and refresh memory cells when power is applied.from: Seagate 600 SSD Product Manual.pdf)
Seagate 600 Pro	They also have a non-Pro version. Seagate 600 Pro-Series 200 GB SSD Review: For The Enterprise – at tom’sHARDWARE.com <http://www.tomshardware.com/reviews/600-pro-ssd-review,3498.html> The 600 Pro ships in six capacities: 100, 120, 200, 240, 400, and 480 GB. This is a fairly interesting set of configurations, since the 100, 200, and 400 GB models are simply factory-over-provisioned versions of the 120, 240, and 480 GB offerings
OCZ Vertex 460 480GB	$354.99 maybe free shipping at TigerDirect on 3/12/2014
Intel® 730 Series 480GB	$449.99 + shipping at TigerDirect on 3/12/2014 70GB writes per day for five years (compared to the industry typical 20GB)
SanDisk Extreme II 480GB

v2