We fix some problems with old SSDs. Problems with SSDs and their solutions Ssd disk failure

Nowadays, when buying a computer, many people have a question: a PC with which drive is better to buy, HDD or SSD. To answer this question, you first need to understand what the main difference between an SSD and an HDD is. HDD hard drives appeared back in the seventies and are still used today in millions of computers. Basic operating principle of HDD hard drive is in writing and reading information on special magnetic plates. Reading is recorded using a head movement lever, while the magnetic disks themselves rotate at very high speeds. Due to the mechanical component of the HDD hard drive and the write and read speed, it is inferior to SSD solid state drives.

How does an SSD drive work? built on recording and reading information from special high-speed memory chips included in its composition. The very speed of writing and reading information from an SSD is several times higher than that of an HDD. In addition, thanks to the microcircuit design, the SSD is less susceptible to damage from impacts and falls, and also has miniature form factors that allow it to be installed in tablets and ultrabooks. Main disadvantages solid state drives are price and life cycle. But progress does not stand still, so we can already see how the price of SSDs is gradually falling, and their rewriting cycle is increasing. In this article we will look at all aspects of working with a solid-state drive and describe their characteristics, so if you decide to switch from an HDD to an SSD, then this article will be very useful for you. In addition, we will look at problems when the BIOS does not see the SSD and many others.

What types of SSD drives exist and which one is better?

When choosing a solid state drive first of all you should pay attention to its form factor and different types of interfaces, through which they connect to the PC. The most common form factor, as with HDD hard drives, is the 2.5-inch case form factor. This solid state drive can be found in many laptops and personal computers. Below is a list that lists all the types of form factors available in SSDs today:

• Form factor type 2.5 inches;
• mSATA form factor type;
• Form factor type M.2.

Below is a picture of 2.5-inch solid-state drives, which is the most common and familiar to many users.

The drives listed above are quite popular models and are labeled as follows: GOODRAM CX200 240 GB, Kingston HyperX FURY SHFS37A/120G and Samsung 850 EVO MZ-75E250B. Such drives are connected using a standard SATA interface, which is used on most computers.

The second type of mSATA device, presented below, has been used mainly in laptop computers since 2009.

It is extremely rare to see mSATA on desktop motherboards, but it is not uncommon in ultrabooks and tablets.

The third form factor M.2 represents a new development that should replace mSATA devices. Below is a picture showing an M.2 disk from Samsung.

We've sorted out the formats of solid-state drives, now let's try to figure out the type of memory used in them. Now on sale you can find devices with SLC, MLC and TLC types of NAND memory. The table below shows the memory characteristics relative to NAND chips.

NAND chip specifications	SLC	MLC	TLC
Number of bits per cell	1	2	3
Number of rewrite cycles	90000 - 100000	10000	3000 - 5000
Chip read time	25 us	50 us	~ 75 us
Programming time	200 – 300 us	600 – 900 us	~ 900 – 1350 us
Erase time	1.5 - 2 ms	3ms	4.5ms

From the characteristics of the table it can be seen that disks built on SLC chips have 90,000 - 100,000 rewrite cycles. It follows from this that such discs will last longer. But buying an SLC drive nowadays is a very expensive pleasure, so most users prefer MLC and TLC drives. To give our readers an idea of ​​the lifespan of an SSD, we have prepared a table that describes it.

Resource of SSD drive on TLC memory
Number of rewrite cycles	3000	5000
Memory	120GB	120GB
Average recording volume per day	12GB	12GB
	10x	10x
	One cycle = 10 * 12	One cycle = 10 * 12
SSD resource formula	SSD resource = 3000/120	SSD resource = 5000/120
Estimating the life of an SSD drive	8 years	13.5 years

It is noticeable from the table that we took as a basis the cheapest drive with TLC memory chips. The formula shows that our SSD goes through one rewrite cycle per day, and this is not so little. For example, a PC user can rewrite much less information, 120 GB per day. But even under such unforgiving conditions, this disk is capable of working for 8 or 13.5 years.

Below is a table for a drive with SLC, MLC memory chips.

Calculation	Resource of SSD drive on SLC memory		Resource of SSD drive on MLC memory
Number of rewrite cycles	90000	100000	9000	10000
Memory	120GB	120 GB	120 GB	120 GB
Average recording volume per day	12GB	12GB	12GB	12GB
Increasing the volume of recorded information	10x	10x	10x	10x
Formula for rewrite cycles per day	One cycle = 10 * 12	One cycle = 10 * 12	One cycle = 10 * 12	One cycle = 10 * 12
SSD resource formula	SSD resource = 90000/120	SSD resource = 100000/120	SSD resource = 9000/120	SSD resource = 10000/120
Estimating the life of an SSD drive	750 years	833 years	75 years old	83 years old

Of course, the user can use more rewrite cycles per day, but then the table indicators will be different. For example, if you rewrite an SSD on MLC memory chips 10 times a day, then the life cycle of this disk will be 7.5 years. Judge for yourself, with a 10-fold rewrite on this disk, you need to rewrite 1200 GB of information per day, which is quite a considerable amount.

Based on the information described above, an SSD with TLC memory chips is quite enough for the average PC user.

We solve problems by upgrading old SSDs

All new drives have a built-in SSD a special subroutine that removes garbage as it becomes full. This garbage removal mechanism is needed to maintain SDD performance. Solid state drives have been on the market for quite some time. In older versions of SSDs, some models do not have a mechanism to protect against garbage cleaning, as a result write speed on such disks drops noticeably. You can solve this problem by completely erasing the information on the disk and subsequently reinstalling Windows. In order not to reinstall Windows or split new partitions on the disk, below we will describe a method that preserves the previous state of the system.

First of all, you need to download the image from http://clonezilla.org Clonezilla, which will help us save all partitions. You can also use other means of system cloning and recovery. The process of creating a system image using Clonezilla It is simple and can be handled by both an experienced user and a beginner. After creating a full backup, you can start cleaning the disk. For this we need an image Linux Parted Magic and utility UNetbootin. You can download this software from the following sites: https://partedmagic.com And http://unetbootin.github.io. Using the utility UNetbootin You can write our image to a USB flash drive, creating a bootable drive from it. After creating a bootable USB flash drive, you can boot from it.

Now on the desktop we will find the program “ Erase Disk" and let's launch it.

In the program window that opens, find the item “ Internal Secure Erase" and click on it. After this, a window should open asking you to select your SSD. Having selected the required disk, the overwriting process will begin. After cleaning, restore the system using Clonezilla. The restored Windows should function as if you had a new SSD.

With the help Linux Parted Magic the user can split and create new partitions on the SSD. You can partition and create a partition on a solid-state drive in the same way as on an HDD hard drive.

We solve problems with performance, BIOS and SSD firmware

The most common problem malfunction, or when the computer does not see the SDD, is old version of motherboard BIOS microcode. You can update the BIOS on any released motherboard. Most often, the problem with SSDs occurs with older versions of motherboards with a new UEFI BIOS. In most cases, updating the BIOS is done using a downloaded microcode file and a USB flash drive. The BIOS file is placed on a flash drive and is used to update. Each motherboard manufacturer has detailed instructions on their website for updating the BIOS.

Be careful when updating the BIOS, as an incorrect update can damage the motherboard.

You can find out which BIOS version is installed on a Windows PC using the CPU-Z utility.

Many PC users buy SSDs to significantly speed up Windows. But with such an upgrade, you should take into account that most older PCs only support the SATA-2 connector. When connecting a solid-state drive to SATA-2, the user will receive a data transfer speed limit of 300 MB/s. It follows that before purchasing, you need to find out whether your motherboard supports the SATA-3 connector, which provides a throughput of 600 MB/s.

To make the SSD more stable, you can get rid of most errors using firmware. The firmware for an SSD is a microcode similar to the BIOS, thanks to which the drive functions. The firmware, as well as the BIOS, can be found on the official website of the SSD manufacturer. Instructions for updating can also be found on the manufacturer's website. Such firmware can solve the problem on some motherboards when the SSD does not see them.

The computer does not see the SSD due to cable or drivers

In addition to the problems described above, very often the motherboard does not see the SSD due to a problem cable or connector. In this case it will help cable replacement SATA to working order. Also, in many cases, the motherboard does not see due to a faulty SATA port, so you can solve this problem connecting to another port.

If you connect an SSD to a computer running on an HDD, you may encounter a situation where it does not see it. The system does not see the installed SSD due to old drivers. This problem can be solved by updates such drivers, like Intel Rapid Storage Technology Driver and AMD AHCI Driver.

SATA AHCI

AHCI is a required mode for the controller to work properly with your SSD. This mode allows the SATA controller to enable new functions, including increasing the speed of the SSD. Unlike the old IDE mode, AHCI mode provides the following advantages:

• AHCI mode support for hot swapping of connected drives in Windows;
• AHCI improves productivity when using NCQ technology;
• AHCI mode allows you to use a transfer speed of 600 MB/s (relevant for SSD drives).
• AHCI mode includes support for additional commands such as TRIM.

When installing Windows on a modern motherboard, it is not necessary to enable AHCI mode in the settings, since it is the default, but if you previously used an older Windows, for example, Windows XP, then you should switch the operating mode from IDE to AHCI. The figure below shows the BIOS settings of an MSI motherboard with AHCI mode enabled.

It is also worth noting that if you installed Windows 7 after XP, then after switching to AHCI mode, the BIOS firmware sees the installed seven in IDE mode, and subsequently you will get a blue screen. In this case, reinstalling Windows 7 in AHCI mode will help.

How to properly partition an SSD disk

Many PC users on forums often have this question: how to properly partition an SSD disk. The answer to this question is quite simple - there is no fundamental difference when partitioning disks between SSD and HDD. Therefore, if you have experience in partitioning HDDs, then you can also partition SDDs. The only point that needs to be taken into account is the capacity of the SSD and HDD, which is much higher for the latter. For example, the volume of the system disk must correspond to the size of the software installed on it and the free space for its proper functioning.

Let's sum it up

After reading this material, each of our readers will be able to see what the advantage of modern solid-state SSDs is over hard HDDs. Also in this material, our readers will find ways to solve problems related to SSDs. It is also worth noting that solid-state drives must be configured correctly in the operating system. For these purposes, we have an article “How to set up an SSD for Windows 7, 8 and 10”, which will help you configure the SSD correctly.

Video on the topic

The SSD market is gradually becoming more diverse. The capacity of SSD drives is growing, and at the same time the price per gigabyte of memory is falling. However, it is still premature to say that SSD drives have become popular. The main reason for this is their low (compared to traditional HDD drives) capacity and very high (again, compared to traditional HDD drives) cost per gigabyte of memory. Therefore, the presence of an SSD drive in a home desktop PC is rather an exception to the rule. Moreover, even in netbooks and laptops, SSD drives are still extremely rare. At the same time, it is already obvious that the future of data storage systems lies with SSD drives, which in the future will displace HDD drives from the market. When will this happen? Yes, in fact, as soon as they become comparable in capacity and cost to HDD drives. Then the latter will simply disappear as a class, since SSD drives have a number of undeniable advantages over HDD drives.
In this article we will look at some features of the functioning of modern SSD drives, which sometimes cause a lot of questions and confusion, we will talk about the features of their architecture, as well as possible options for using these drives in laptops, PCs and servers.

The relevance of switching to SSD drives

The performance of modern central processing units, which determine the computing capabilities of a PC, significantly exceeds the performance of traditional hard drives (HDDs). As a result, it is the data storage subsystems that in many cases become a bottleneck that hinders the growth of computer performance as a whole. The use of expensive solutions based on RAID arrays only partially solves the problem of imbalance in the performance of processors and HDD-based storage subsystems. And in the future, the imbalance between the performance of processors and HDDs will only increase, and we will inevitably come to the point where computer performance in many applications will no longer be determined by processor performance, but will rest on the weakest link - the data storage subsystem. Thus, since 1996, the average performance of processors has increased 175 times, while the performance of HDD disks (meaning selective reading of 20 KB blocks) has increased only 1.3 times.

Today, the only way to solve this problem is to switch from HDD to SSD (Solid State Drive) based on flash memory. Such drives are capable of providing a level of performance that fully matches the performance of modern multi-core processors.

However, high performance is not the only advantage of SSD drives. They are also completely silent since they contain no moving parts, and, especially important for laptops, consume much less power compared to HDD drives. Thus, the power consumption of a regular 2.5-inch HDD in active mode is about 2.5-3 W and about 0.85-1 W in idle mode (Idle). If the HDD is not active, then after some time (depending on the settings) it goes into a low-power mode (Standby or Sleep) and when exiting this mode it takes about 1-2 seconds to spin up. Typical power consumption of an SSD (not server) in activity mode is about 0.15 W, and in idle mode - 0.06 W. Moreover, if configured correctly, the transition from activity mode to low-power mode occurs automatically if the disk is inactive for 25 ms. And these drives turn on almost instantly, since they simply have nothing to spin up. Note that in order for an SSD disk to automatically enter low power mode, it is necessary to activate the Device Initiated Power Management (DIPM) function in the registry, since the Host Initiated Power Management (HIPM) function is set by default, when the transition to low power mode is not controlled by the disk itself. and the operating system.

SSD drives are not inferior to traditional HDD drives in such a characteristic as mean time between failures (MTFB). So, if for a HDD the mean time between failures is about 300 thousand hours, then for SSD drives it is over a million hours.

It would seem that if the advantages of SSD drives are so obvious, why have they not yet become widespread? Unfortunately, SSD drives also have serious disadvantages. First of all, modern SSD drives are not comparable to HDD drives in terms of capacity. So, if the capacity of HDD drives (3.5 inches in size) reaches 3 TB, then the maximum capacity of SSD drives (2.5 inches in size) is only 512 GB. True, if we compare 2.5-inch SSD and HDD drives, their capacity is quite comparable.

The second disadvantage of SSD drives is their cost, which is several times higher than that of an HDD.

However, when it comes to the capacity of SSD drives, not everything is as bad as it might seem. SSD capacity is growing at a much faster rate than HDD capacity, and the day is not far when SSDs will surpass HDD capacity. As proof, here are some interesting statistics. In 2006, Intel, one of the leading players in the SSD market, produced NAND flash memory chips for SSD drives using the 90 nm process technology, with capacities of 1 or 2 Gbit. In 2009, Intel produced flash memory chips using the 34 nm process technology, and the capacity of the chips began to be 32 Gbit. In 2010, the company mastered the 25nm process for producing flash memory chips with a capacity of 64 Gbit. As you can see, the growth rate of capacity of flash memory chips for SSD drives is impressive: in fact, it doubles every year. So, soon SSD drives will surpass HDDs in volume.

It should also be noted that although the widespread use of SSD drives is still far away, it is incorrect to say that SSD drives are not bought at all. The statistics are as follows: in 2008, only 700 thousand SSD drives were sold in the world, in 2009 the sales volume was already 2 million units, and this year, according to forecasts, it will reach 5.9 million units. It is expected that by 2013 the market for SSD drives will reach 61.8 million units.

So, forecasts for sales volumes of SSD drives are very optimistic, but they do not answer the main question: what should users do today, when the capacity of SSD drives is not yet high enough and their cost is still very high? If we are talking about home users, then, of course, it makes no sense to throw away HDDs to install an SSD. However, it is still possible to increase computer performance by using SSD drives. The optimal solution is when a desktop PC uses a combination of one SSD drive and one or more HDD drives. You can install the operating system and all programs on an SSD disk (an 80 GB disk will be enough for this), and use the HDD disk for data storage.

Flash memory cell design

As we said, the main advantage of SSD drives is their higher performance compared to HDD drives, but no specific characteristics such as sequential and selective read and write speeds were provided. However, before moving on to considering the speed characteristics of SSD drives, as well as types of SSD drives, you need to familiarize yourself with the features of their architecture and the process of reading and writing information to these drives. Let's start with a brief description of the structure of a flash memory cell.

At its simplest level, a flash memory cell is n-channel MOSFET transistor with a so-called floating gate. Let us remember that the usual n-channel MOSFET transistor (structure n-p-n) can be in two states: open and locked (closed). By controlling the voltage between drain and gate, it is possible to create an electron conduction channel ( n-channel) between source and drain (Fig. 1). The voltage at which a conduction channel appears is called threshold. The presence of a conduction channel corresponds to the open state of the transistor, and the absence (when the transistor is not able to conduct current from source to drain) corresponds to a closed state.

Rice. 1. MOSFET transistor structure (open and closed state)

In the open state, the voltage between the drain and the source is close to zero, and in the closed state it can reach a high value. Of course, the transistor itself is not capable of storing information. Actually, a floating shutter is designed specifically for storing information (Fig. 2). It is made of polycrystalline silicon and is completely surrounded by a layer of dielectric, which provides it with complete absence of electrical contact with the elements of the transistor. The floating gate is located between the control gate and the substrate p-n-transitions. Such a gate is capable of maintaining a charge (negative) placed on it for an unlimited time (up to 10 years). The presence or absence of excess negative charge (electrons) on the floating gate can be interpreted as a logical one and zero.

Rice. 2. Floating gate transistor design and reading the contents of a memory cell

First, consider the situation where there are no electrons on the floating gate. In this case, the transistor behaves similarly to the traditional transistor already discussed. When a positive voltage (initialization of the memory cell) equal to the threshold value is applied to the control gate, a conduction channel is created in the gate area - and the transistor goes into the open state. If an excess negative charge (electrons) is placed on the floating gate, then even when a threshold voltage is applied to the control gate, it compensates for the electric field created by the control gate and prevents the formation of a conduction channel, that is, the transistor will be in the closed state.

Thus, the presence or absence of charge on the floating gate uniquely determines the state of the transistor (open or closed) when the same threshold voltage is applied to the control gate. If the supply of voltage to the control gate is interpreted as initializing the memory cell, then the voltage between the source and drain can be used to judge the presence or absence of charge on the floating gate.

That is, in the absence of a control voltage at the gate, regardless of the presence or absence of charge on the floating gate, the transistor will always be closed, and when a threshold voltage is applied to the gate, the state of the transistor will be determined by the presence of charge on the floating gate: if there is a charge, then the transistor will be closed and the output voltage will be high; if there is no charge, the transistor will be open and the output voltage will be low.

The closed state of the transistor (absence of a conduction channel) is usually interpreted as a logical zero, and the open state (presence of a conduction channel) as a logical one. Thus, when initializing a memory cell (applying a threshold voltage to the gate), the presence of charge on the floating gate is interpreted as a logical zero, and its absence as a logical one (see table).

The result is a kind of elementary memory cell capable of storing one information bit. In this case, it is important that the charge on the floating gate (if there is one) can be maintained indefinitely, both during initialization of the memory cell and in the absence of voltage on the control gate. In this case, the memory cell will be non-volatile. All that remains is to figure out how to place a charge on the floating gate (write the contents of a memory cell) and remove it from there (erase the contents of a memory cell).

Placing a charge on a floating gate is realized either by the injection of hot electrons (CHE-Channel Hot Electrons) or by the Fowler-Nordheim tunneling method (Fig. 3). Well, charge removal is carried out only by the Fowler tunneling method.

Rice. 3. The process of writing and erasing an information bit into a floating gate transistor

When using the hot electron injection method, a high voltage is applied to the drain and control gate (a voltage above the threshold is applied to the control gate) to give the electrons in the channel enough energy to overcome the potential barrier created by a thin dielectric layer and tunnel into the floating gate region ( When reading, less voltage is applied to the control gate, and no tunneling effect is observed).

To remove charge from the floating gate (the process of erasing a memory cell), a high negative voltage is applied to the control gate, and a positive voltage is applied to the source region. This causes electrons to tunnel from the floating gate region to the source region (Fowler-Nordheim (FN) quantum tunneling).

The floating gate transistor we considered can act as a basic flash memory cell. However, single-transistor cells have a number of significant disadvantages, the main one being poor scalability. The fact is that when organizing a memory array, each memory cell (transistor) is connected to two perpendicular buses: the control gates are connected to a bus called the word line, and the drains are connected to a bus called the bit line (in the future, this organization will be considered using the example of NOR -architecture). Due to the high voltage in the circuit when writing using the hot electron injection method, all lines - words, bits and sources - must be located at a sufficiently large distance from each other to provide the required level of isolation, which, of course, affects the limitation of flash memory capacity.

Another disadvantage of a single-transistor memory cell is the effect of excessive charge removal from the floating gate, which cannot be compensated by the writing process. As a result, a positive charge is formed on the floating gate and the transistor always remains in the open state.

Other types of memory cells have also become widespread, such as the SST cell (Fig. 4), developed by Silicon Storage Technology, Inc. In the SST cell transistor, the shapes of the floating and control gates have been changed. The control gate is aligned at its edge with the edge of the drain, and its curved shape makes it possible to place a floating gate partially below it and at the same time above the source area. This arrangement of the floating gate makes it possible to simplify, on the one hand, the process of placing a charge on it by injection of hot electrons, and on the other hand, the process of removing the charge due to the Fowler-Nordheim tunneling effect.

Rice. 4. Structure of an SST memory cell

When the charge is removed, electron tunneling occurs not in the source region, as in the considered single-transistor cell, but in the control gate region. To do this, a high positive voltage is applied to the control gate. Under the influence of the electric field created by the control gate, electrons are tunneled from the floating gate, which is facilitated by its curved shape towards the edges.

By placing a charge on the floating gate, the drain is grounded and a positive voltage is applied to the source and control gate. In this case, the control gate forms a conduction channel, and the voltage between the drain and the source “accelerates” the electrons, giving them energy sufficient to overcome the potential barrier, that is, to tunnel to the floating gate.

Unlike a single-transistor memory cell, an SST cell has a slightly different memory array organization scheme.

Multi-level and single-level flash memory cells

All types of memory cells discussed so far are capable of storing only one bit of information per cell. Such memory cells are called single-level cells (SLC). However, there are also such cells, each of which stores several bits - these are multi-level cells, or MLC (Multi Level Cell).

As already noted when describing a single-transistor memory cell, the presence of a logical one or zero is determined by the voltage value on the bit line and depends on the presence or absence of charge on the floating gate. If a threshold voltage value is applied to the control gate, then in the absence of charge on the floating gate the transistor is open, which corresponds to a logical one. If there is a negative charge on the floating gate, which shields the field created by the control gate with its field, then the transistor is in the closed state, which corresponds to logical zero. It is clear that even if there is a negative charge on the floating gate, the transistor can be switched to the open state, but for this you will have to apply a voltage exceeding the threshold value to the control gate. Therefore, the absence or presence of charge on the floating gate can be judged by the threshold voltage value at the control gate. Since the threshold voltage depends on the amount of charge on the floating gate, it is possible not only to determine two limiting cases - the absence or presence of charge - but also to judge the amount of charge by the value of the threshold voltage. Thus, if it is possible to place a different number of charge levels on a floating gate, each of which has its own threshold voltage value, then several information bits can be stored in one memory cell. For example, in order to store 2 bits in one cell using such a transistor, it is necessary to distinguish between four threshold voltages, that is, to be able to place four different charge levels on the floating gate. Then each of the four threshold voltages can be assigned a combination of two bits: 00, 01, 10, 11.

In order to be able to write 4 bits into one cell, it is necessary to distinguish 16 threshold voltages.

MLC cells are being actively developed by Intel, which is why memory technology based on MLC cells is called Intel StrataFlash.

Note that SLC memory cells provide higher read and write speeds. In addition, they are more durable, however, SSD drives based on them are more expensive, since with equal capacity of SSD drives based on MLC and SLC memory cells, the number of memory cells themselves in an MLC drive will be half as much (in the case of four-level cells memory). This is why SSD drives based on SLC memory cells are used mainly in servers.

Flash memory array architecture

The simplest flash memory cell we have considered, based on a floating gate transistor, capable of storing one bit of information, can be used to create non-volatile memory arrays. To do this, you only need to appropriately combine many cells into a single array, that is, create a memory architecture.

There are several types of flash memory architectures, that is, ways of combining memory cells into a single array, but the NOR and NAND architectures are the most widespread. Note that SSD drives use NAND-type memory organization, however, to better understand the features of this architecture, it is logical to first consider the simpler NOR architecture. In addition, the NOR architecture was the first architecture used in flash memory.

The NOR architecture (Fig. 5) involves a parallel way of combining memory cells into an array. As already noted, to initialize a memory cell, that is, to gain access to the contents of the cell, it is necessary to apply a threshold voltage value to the control gate. Therefore, all control gates must be connected to a control line called the Word Line. The contents of a memory cell are analyzed based on the signal level at the drain of the transistor. Therefore, the drains of the transistors are connected to a line called a bit line.

Rice. 5. NOR architecture

The NOR architecture owes its name to the logical operation “OR-NOT” (the English abbreviation is NOR). The logical NOR operation on multiple operands produces a value of one when all operands are zero, and a value of zero otherwise. In this case, we mean the principle of connecting transistors in general, and not specifically floating-gate transistors.

As an example, consider several transistors (without a floating gate) connected to a single bit line (Fig. 6). In this case, if at least one transistor is open, the output voltage on the bit line will be low. And only in the case when all transistors are closed, the voltage on the bit line will be high. We obtain the truth table of the input voltages at the gates of the transistors and the output voltage on the bit line, corresponding to the truth table of the logical function “NOR” (NOR). That is why this circuit for combining transistors is called NOR.

Rice. 6. Connecting transistors according to the NOR circuit

The NOR architecture provides random, fast access to any memory cell, but the processes of writing (using the hot electron injection method) and erasing information are quite slow. In addition, due to the technological features of producing flash memory chips with NOR architecture, the cell size is large, so such memory does not scale well.

Another common flash memory architecture is the NAND architecture (Figure 7), which is a logical NAND operation. The NAND operation produces a value of zero only when all operands are zero, and a value of one in all other cases. The NAND architecture involves a series connection of transistors, in which the drain of each transistor is connected to the source of the neighboring transistor, and in a series of several transistors connected in series, only one of them is connected to the bit line. Moreover, when considering the connection architecture, we are not talking specifically about floating-gate transistors.

Rice. 7. NAND architecture

Let's consider a group of such transistors connected in series (without a floating gate) (Fig. 8). If the control voltage at the gates of all transistors is equal to the threshold value, then all transistors are in the open state and the output voltage (voltage on the bit line) will be low, which corresponds to logical zero. If the input voltage on at least one transistor is low (below the threshold value), that is, if at least one transistor is in the off state, then the voltage on the bit line will be high, which corresponds to a logical one. We obtain the truth table of the input voltages at the gates of the transistors (voltages on the word line) and the output voltage on the bit line, corresponding to the truth table of the logical function “NAND” (NAND). That is why this circuit for combining transistors is called NAND.

Rice. 8. Connecting transistors using the NAND circuit

In a floating-gate NAND transistor circuit, a group of transistors connected in series is connected at both ends with regular transistors (without a floating gate), which isolate the group of transistors from both ground and the bit line and connect the entire group of transistors to the bit line when they are initialized.

Compared to the NOR architecture, this architecture, due to the peculiarities of the manufacturing process (combining the drains and sources of adjacent transistors and a much smaller number of conductors), allows for a more compact arrangement of transistors, and therefore is highly scalable. Unlike the NOR architecture, where information is written using the hot electron injection method, in the NAND architecture, recording is carried out using the FN tunneling method, which allows for faster writing than for the NOR architecture.

Naturally, the question arises: how in the NAND architecture can you access a single memory cell (read the contents of the cell)? After all, if at least one of the transistors in such a series-connected group is in the closed state (which can be interpreted as the presence of a charge on the floating gate of the corresponding transistor), then the voltage on the bit line will be high regardless of the state of the remaining cells. To access a single cell, it is not enough to simply apply a threshold voltage to the gate of the transistor corresponding to that cell and measure the voltage on the bit line. It is also necessary that all other transistors are in the open state. To do this, a threshold voltage value is applied to the gate of the transistor corresponding to the memory cell, whose contents must be read, and a voltage exceeding the threshold value is applied to the gates of all other transistors and is sufficient to form a conduction channel even in the presence of a charge on the floating gate, but insufficient for the effect quantum tunneling of charges. In this case, all these transistors go into the open state and the voltage on the bit line is determined by the presence or absence of charge on the floating gate of the transistor corresponding to the memory cell being accessed.

Logical structure of NAND flash memory

As we have already noted, SSD drives use flash memory organized like NAND, so in the future we will focus on looking exclusively at NAND flash memory.

Despite the fact that flash memory allows access to read, write and erase a single cell, to more efficiently use elementary memory cells, they have been combined into arrays with a four-level structure. At the lowest level is the elementary memory cell, and the elementary cells combined into an array that holds 4 KB of data are called a memory page. 128 such pages form a memory block of 512 KB in size (sometimes a memory block includes 64 pages), and 1024 blocks form a 512 MB array. Thus, the logical structure of combining cells into arrays is quite simple. A page is like a cluster (sector) on a hard drive and represents the minimum size of data that flash memory can handle. However, there are fundamental differences between a hard disk cluster and a flash memory page when performing read, write, and delete operations. So, if in a hard disk a cluster can be read, written and deleted, then in flash memory, read and write operations are possible in 4 KB pages, and erasing data is only possible in 512 KB blocks. Moreover, once information is written to a page, it cannot be overwritten until it is cleared (deleted).

Features of data recording operations in SSD drives

So, as we have already noted, writing and reading data in NAND flash memory is possible in 4 KB pages, and erasing data is only possible in 512 KB blocks. In general, the process of writing information to SSD drives is very different from the similar process with HDD drives. This, for example, is due to the fact that the performance of SSD drives changes over time, and the speeds of sequential and selective access to flash memory differ from each other. In order to explain these phenomena, let's take a closer look at the recording processes on HDD and SSD drives.

In the case of HDD hard drives, the smallest unit of information that the hard drive management system operates on is called a sector or block. In HDD, the sector size is 4 KB (in new models) or 512 bytes. To address sectors (blocks) on the disk, the LBA (Logical Block Addressing) method is used, in which each block addressed on the hard disk has its own sequence number - an integer starting from zero (that is, the first block LBA = 0, the second LBA = 1, etc.). The number of LBA blocks on a disk is determined by the number of cylinders, tracks, sectors and read/write heads. So, the LBA block number is calculated using the formula:

LBA = [(Cylinder x No_of_heads + Heads) x Sectors/track] + )