Media Notes / Flash Memory

2 Following

Flash Memory is a type of erasable ROM intended for use as Mass Storage. Ever since The '80s, virtually all ROM chips produced could be somehow erased and rewritten, though most of them could be only erased as a whole: either by exposure to UV light, or by high voltage^note relatively, the voltages involved are ~10-15 V, but it's still rather large for the semiconductors pulse. After erasing the chip, it could be written to, quite similar to normal RAM, but it would hold the information after the power was switched off. This technique of erasing and writing is commonly called "flashing" in technical lingo, a rudiment of the "ancient" times when ROM chips were written by burning the specific connections inside them using these pulses. Flash memory takes its name from this process, but differs from these early chips by the simple fact that it can be "flashed" partially, leaving other parts of the chip untouched.

Each flash memory cell is built using a MOSFET type transistor, with the only difference being the gate (the part that acts like the switch) is electrically insulated from the transistor called a "floating gate." This floating gate is similar to a capacitor in that it stores charge. To fill or empty the charge in the floating gate, a high voltage is passed through another gate called the Control Gate. Compared to the older magnetic or optical drives, it has no moving parts, and so these devices are much more durable and they also require little physical space, electrical power, or cooling. Also because there's no moving parts, finding and grabbing the data is much faster. Because of these things, flash memory is ideal for portable devices (i.e. tablets, smartphones, etc.).

Flash comes in two flavors: NAND and NOR flash due to the way the flash cells resembles their respective logic gates. NAND flash offers higher density and faster erase speeds at the cost of not being byte-addressable, which isn't a problem for mass storage since almost all of them can only address "sectors" of data (typically 512 or 4096 bytes). NOR flash is the opposite, trading high density and erase speed for byte addressability, which is useful in embedded systems as they can run apps directly from flash memory (i.e., not copy the application into RAM), they usually don't need high capacities, and are rarely re-programmed so the downsides aren't a problem.

Despite all these great features over other storage drives, flash memory has several problems:

• Because flash memory is built up of semiconductor chips and requires complex and expensive tools to manufacture, flash memory costs more per byte than mass storage that came before it. By mid 2025, it went down to $0.80 per gigabyte ^note This was before the spike in demand from data centers. Compare this to a hard drive which reached costs as low as $0.021 per gigabyte.
• When there's no data on the chip, data can be written to any part of it at random and hardly causing any fatigue. However, it's erasing the data that's tricky. While flash chips can be erased partially, it must be done in blocks, by using high-voltage pulses to reinitialize the floating gates in whole sections of the chip, commonly called "pages". It's this sectional nature that means that Flash cannot be used as a normal non-volatile RAM similar to core; they need to have additional controller chips or built-in circuits to handle the erasing and addressing, which makes them more similar to disk drives in actual usage.
• To add insult to injury, the high voltage needed to erase the flash memory degrades the structure of the semiconductor itself, so in cases of intensive write and delete cycles, it wears out somewhat faster than most other types of Mass Storage. This makes flash (much like CD/DVD-RW discs) generally unsuitable for use as RAM.
• Even worse, unlike most other transistor based technology, flash memory is one of the few ones that gets worse as size shrinks. This is due to the floating gate needing to be a minimum size to hold a detectable amount of charge. Also, in order to increase storage density, flash cells were made to hold various amounts of charge for a number of voltage levels. These have become Single Level Charge (SLC, 1-bit), Multi-Level Charge (MLC, 2-bits), Triple-Level Charge (TLC, 3-bits), and Quad-Level Charge (QLC, 4-bits). While higher level charges store more data, it comes at the cost of write performance and the number of times they can be erased before it becomes unusable.

The demand for improved efficiency in handling memory and 'cleaning up' memory cells resulted in a number of handling tricks the industry calls "Garbage Collection" being introduced, to help ensure that the lifespan of memory cells was used as efficiently as possible to maximise the overall operational lifespan. A further technique called 'wear leveling' ensures that wear and tear on flash memory cells is evenly distributed, rather than concentrated on a potential point of weakness. However, before anyone gets worried about flash having a finite number of write-cycles, most of the highly regarded SSD manufacturers guarantee a rating of at least 150 TBW (terabytes written) per 250GB of capacity. Assuming the average person does about 20 GB of writes per day, this comes out to around nearly 20 years of usage per 150 TBW out of the SSD before it wears out. Given the life expectancy of most electronics that are daily driven, it's safe to assume something else on the SSD will croak before you've ran the flash cells down.

Some flash memory based mass storage also employs a few things to help with write performance. The highest performing drives have on-board DRAM and user-transparent SLC flash memory to act as write cache. This is usually good for dozens of gigabytes before speed dramatically drops off. However, most writes typically aren't this large so most users won't notice. Cheaper models drop either or both, but the metric that suffers is write performance.

And as a compromise between performance and storage capacity, assuming a single 1 or 2TB drive can't suit most of your needs, it's been recommended to get an SSD for your programs as these benefit the most from their read performance. For data like pictures, music, videos, and documents, a hard drive is still plenty performant.

Common form factors:

• Solid State Drives are HDD-sized devices that hold quite a few Flash chips and are roughly similar in capacity to modern HDDs. They are usually marketed as direct replacements for hard drives for people that prefer their speed and indifference to rough handling, and are not afraid of their price, small sizes, and the whole "fatigue" thing. As such, they are typically found in standard laptop HDD sizes, and use common HDD IDE or SATA interfaces. Slightly easing the "fatigue anxiety", though, modern SSDs have roughly the same mean time between failures as modern hard drives ^note In a highly simplified manner and assuming equal fatigue, MLC flash memory, commonly used in SSDs, have a life span of tens of thousands of program/erase cycles. For a 128GB drive, this amounts to well over 1 petabyte of write-lifetime before being worn out. Further extending the life is TRIM support, which nullifies data marked as "deleted", making it ready for fast, efficient writing. On the other hand, they're still much more expensive. Nintendo's Wii has a rather small 512MB SSD instead of its competitors' multigigabyte hard drives. (The Wii U upped it to 8/32 GB, depending on the configuration purchased.) It's common for homebuilt PCs owners to offset the cost problem of an SSD by having a relatively small SSD to hold key programs and the operating system(s), with one or more larger HDDs for general purpose storage. Some more specialized devices, like the Apple Macbook Air and certain extra-thin models of Ultrabook laptops, utilize an extremely minimized form factor for their solid state drives that connects to the motherboard using a very slim PCI-E connector. This style of SSD connector (called M.2) has also propagated to desktop computers as a faster alternative to SATA. One of the largest SSDs currently in production is a 30 Tb server drive from Samsung.
• High Capacity Solid State Drives are not necessarily refereed to by this name, but these are intended as a compromise between the high performance of a regular SSD and the high-capacity of a mechanical drive. Their structure can make writing slower due to how data is "shingled" like roofing tiles and how said data may share a certain area; any modification means the entire area must be rewritten, but this is still a major improvement over conventional hard drives and their capacities can rival said drives cost efficiently. To alleviate their performance bottlenecks, a smaller portion of regular flash memory can be added as a "cache" where frequently-modified files are stored and almost make the slowdown nonexistent.
• Solid State Hybrid Drives strive to combine both magnetic turn-tables and solid state memory into a single unit. They can simply be a traditional hard drive with a secondary solid state partition on board, sharing the same SATA port (rarely); or they can be a seam-less unit that appears like a traditional hard drive, with an invisible cache of solid state memory that the firmware uses for small, frequently-used files.
  
  These units have added complexity, but are typically much less expensive than an SSD of the same capacity. They can speed up boot-times tremendously, and launch your favorite applications almost like a pure SSD. Huge files, like High-Definition video are stored on the turn-table like normal, as the disk-platters are excellent for continuous access to huge data chunks. Hybrid Drives often utilize optimization techniques to allocate the flash memory space provided for the most-accessed files to speed up disk operations.
• Thumb Drives ^explanationThe brand name for portable USB flash drives created by Singaporean technology company Trek, other common names include pen-drives, USB drives, USB sticks, jump drives and flash-disks (after floppy disks) are rather similar, differing only in that they have smaller size (only slightly larger than the plug — typically male USB — they are integrated with) and capacity, and generally aren't designed to be the main system volume, but intended to to replace removable media, similar to the early floppy disks. They usually use not the internal HDD interface, like SSDs, but one of the common outside buses like USB. Most (but not all) thumb drives contain flash memory. This is usually permanent, but upgradeable ones instead contain flash readers that take a standard flash card.
  • And then there's the USB bracelet.
    • And USB Flash watches, James Bond style! Some Swiss Army knives come with a USB drive attachment as well.
  • At the extreme, some manufacturers like Sandisk have made their Cruzer Fit USB flash drives so tiny they barely stick out of the port when plugged in and can be easily mistaken for something like one of Logitech's nano-receivers for wireless mice.
  • As with floppy disks, USB flash drives can be configured to serve as boot disks and installers. There are utilities to facilitate booting into Windows, Linux, and even Mac OS.
• Flash Cards are simply boards with flash chips and the minimum number of support chips needed to make the memory work, all squeezed into a standard-sized Cartridge. Numerous formats are available, with varying degrees of physical and logical compatibility, in approximately chronological order:
  • Numerous game consoles use proprietary memory card formats to save game progress between play sessions separately from the game itself. Since consoles don't use rewritable optical drives, this became pretty much mandatory once games started coming on optical disc instead of magnetic disk or cartridge.
  • Several flash card standards have a rather bizarre origins from the laptop extension buses that make them somewhat similar to modern SSDs.
    • PCMCIA, first of them, was originally a laptop memory card format, based on a pin-reduced version of the good, ol' ISA (technically, on its hard-drive subset, the IDE interface) bus. Later it became "PC Card" and explicitly started allowing IDE drives and other peripherals. Has 68 pins compared to the 16-bit ISA's 100.
    • Its successor, CardBus, is commonly called "PCI for laptops", as it could switch back and forth between ISA/IDE and PCI modes on-the-fly. Also 68 pins.
    • CompactFlash is a tiny version of PCMCIA and is electrically compatible with IDE, requiring only a pin converter. Has 50 pins.
    • A variant of CompactFlash, CFast, is based on the Serial ATA (SATA) interface instead of the Parallel ATA (PATA) interface that the original CF used. CFast cards are not physically or electrically compatible with CompactFlash, but because SATA can emulate the PATA protocol, existing CompactFlash drivers could be used. 7 data pins, 17 power pins.
    • A more recent CompactFlash variant is XQD, which is based on PCI Express (PCIe). Incompatible with any previous CF variants. Pinless.
    • XQD has since been superseded by CFExpress (aka CFx), also based on an updated version of PCIe, giving it even faster read/write speeds. CFx is physically and electrically compatible with XQD; devices with XQD slots need only a firmware upgrade to read and write CFx cards.
    • PCMCIA's successor, ExpressCard, is based on PCI Express and USB, and is being used for some truly enormous flash cards that are basically a whole bunch of readers for other formats squeezed into one card. 26 pins.
  • The SIM cards in many cellphones are used as their primary (sometimes ONLY) mass storage device
  • MMC (MultiMediaCard,) RS-MMC and MMCmicro.
    • SD (Secure Digital), miniSD and microSD are an updated version of MMC. ^note Due to this, SD card readers tend to be backward compatible with MMC.
  • xD Picture Cards, a variant developed by Olympus and Fujifilm specifically for digital cameras. (Now phased out in favor of the better-supported SD format.)
  • Memory Stick and Memory Stick PRO/PRO Duo/Micro are Sony's proprietary format. Now in the process of being phased out in favor of SD.
• Nearly all modern MP3 players and all smartphones (and whatever you call their non-cellular-capable brethren) and tablets use internal flash memory, similar to an SSD. In a similar vein, Apple laptops with solid-state storage as their only storage option, like the MacBook Air, have it on several chips that are soldered directly onto the motherboard wherever the designers could make them fit.