RAM details. Determining the frequency of RAM Memory with high frequency

17.10.2023

RAM FAQ

Random access memory (RAM) is designed for temporary storage of data and commands necessary for the Central Processing Unit to perform operations. RAM transmits commands and data to the processor directly or through cache memory. Each RAM cell has its own individual address...

The most common types of memory are:

^ SDR SDRAM(designations PC66, PC100, PC133)
DDR SDRAM(designations PC266, PC333, etc. or PC2100, PC2700)
RDRAM(PC800)

Now for subsequent explanations, I’ll tell you about timings and frequencies. Timing- this is the delay between individual operations performed by the controller when accessing memory.

If we consider the composition of the memory, we get: its entire space is presented in the form of cells (rectangles), which consist of a certain number of rows and columns. One such "rectangle" is called a page, and the collection of pages is called a bank.

To access a cell, the controller sets the bank number, the page number in it, the row number and the column number, time is spent on all requests, in addition, quite a large cost is spent on opening and closing the bank after the read/write operation itself. Every action takes time, it is called timing.

Now let's take a closer look at each of the timings. Some of them are not available for configuration - access time CS# (crystal select) this signal determines the crystal (chip) on the module to carry out the operation.

In addition, the rest can be changed:

^RCD (RAS-to-CAS Delay) this is the delay between signals RAS (Row Address Strobe) And CAS (Column Address Strobe), this parameter characterizes the interval between accesses to the bus by the signal memory controller RAS# And CAS#.
CAS Latency (CL) this is the delay between the read command and the availability of the first word to be read. Introduced to set address registers to guarantee a stable signal level.
^ RAS Precharge (RP) this is the time of re-issuance (charge accumulation period) of the signal RAS#- after what time the memory controller will be able to issue a line address initialization signal again.

Note: the order of operations is exactly this (RCD-CL-RP), but often the timings are written not in order, but by “importance” - CL-RCD-RP.

^ Precharge Delay(or Active Precharge Delay; more often referred to as Tras) is the active time of the line. Those. the period during which a row is closed if the next required cell is in another row.
^SDRAM Idle Timer(or SDRAM Idle Cycle Limit) the number of clock cycles during which the page remains open, after which the page is forced to close, either to access another page or to refresh (refresh)
^ Burst Length This is a parameter that sets the size of the memory prefetch relative to the starting address of the access. The larger its size, the higher the memory performance.

Well, we seem to have understood the basic concepts of timings, now let’s take a closer look at memory ratings (PC100, PC2100, DDR333, etc.)

There are two types of designations for the same memory: one by the "effective frequency" DDRxxx, and the second by the theoretical bandwidth PCxxxx.

The designation "DDRxxx" historically developed from the sequence of names of the standards "PC66-PC100-PC133" - when it was customary to associate memory speed with frequency (unless a new abbreviation "DDR" was introduced in order to distinguish SDR SDRAM from DDR SDRAM). Simultaneously with DDR SDRAM memory, RDRAM memory (Rambus) appeared, on which cunning marketers decided to set not the frequency, but the bandwidth - PC800. At the same time, the width of the data bus remained 64 bits (8 bytes), that is, those same PC800 (800 MB/s) were obtained by multiplying 100 MHz by 8. Naturally, nothing has changed from the name, and PC800 RDRAM is the same the same PC100 SDRAM, only in a different package... This is nothing more than a sales strategy, roughly speaking, “to trick people”. In response, companies that produce modules began to write theoretical throughput - PCxxxx. This is how PC1600, PC2100 and the following appeared... At the same time, DDR SDRAM has an effective frequency that is twice as high, which means the number on the designation is higher.

Here is an example of notation correspondence:

100 MHz = PC1600 DDR SDRAM = DDR200 SDRAM = PC100 SDRAM = PC800 RDRAM
133 MHz = PC2100 DDR SDRAM = DDR266 SDRAM = PC133 SDRAM = PC1066 RDRAM
166 MHz = PC2700 DDR SDRAM = DDR333 SDRAM = PC166 SDRAM = PC1333 RDRAM
200 MHz = PC3200 DDR SDRAM = DDR400 SDRAM = PC200 SDRAM = PC1600 RDRAM
250 MHz = PC4000 DDR SDRAM = DDR500 SDRAM

As for ^ RAMBUS (RDRAM) I won’t write much, but I’ll still try to introduce it to you.

There are three types of RDRAM - Base, Concurrent And Direct. Base and Concurrent are practically the same thing, but Direct has significant differences, so I’ll tell you about the first two in general, and about the last one in more detail.

^Base RDRAM And Concurrent RDRAM Basically they differ only in operating frequencies: for the first, the frequency is 250-300 MHz, and for the second, this parameter is, accordingly, 300-350 MHz. Data is transmitted at two data packets per clock cycle, so the effective transmission frequency is twice as high. The memory uses an eight-bit data bus, which consequently gives a throughput of 500-600 Mb/s (BRDRAM) and 600-700 Mb/s (CRDRAM).

^ Direct RDRAM (DRDRAM) unlike Base and Concurrent, it has a 16-bit bus and operates at a frequency of 400 MHz. The bandwidth of Direct RDRAM is 1.6 Gb/s (taking into account bidirectional data transfer), which looks pretty good compared to SDRAM (1 Gb/s for PC133). Usually, when talking about RDRAM, they mean DRDRAM, so the letter "D" in the name is often omitted. When this type of memory appeared, Intel created a chipset for the Pentium 4 - i850.

The biggest plus Rambus memory means that the more modules, the greater the throughput, for example, up to 1.6 Gb/s per channel and up to 6.4 Gb/s with four channels.

There are also two disadvantages, quite significant:

1. The claws are gold and become unusable if the memory card is removed and inserted into the slot more than 10 times (approximately).

2. Overpriced, but many people find very good use for this memory and are willing to pay top dollar for them.

That's probably all, we've figured out the timings, names and denominations, now I'll tell you a little about various important little things.

You probably saw the By SPD option in the BIOS when setting the memory frequency, what does this mean? ^ SPD - Serial Presence Detect, this is a microcircuit on the module into which all the parameters for the operation of the module are programmed, these are the “default values”, so to speak. Now, due to the emergence of “noname” companies, they began to write the manufacturer’s name and date into this chip.

^ Register memory

Registered Memory This is memory with registers that serve as a buffer between the memory controller and the module chips. Registers reduce the load on the synchronization system and allow you to add a very large amount of memory (16 or 24 gigabytes) without overloading the controller circuits.

But this scheme has a drawback - the registers introduce a delay of 1 clock cycle for each operation, which means that register memory is slower than usual, all other things being equal. That is, the overclocker is not interested in it (and it is very expensive).

Everyone is now shouting about Dual channel - what is it?

^ Dual channel- double channel, this allows you to access two modules simultaneously. Dual channel is not a type of module, but a function integrated into the motherboard. Can be used with two (preferably) identical modules. It turns on automatically when there are 2 modules.

Note: to activate this function, you must install modules in slots of different colors.

Parity And ECC

Memory with Parity This is a parity-check memory that can detect certain types of errors.

^ Memory with ECC This is an error correction memory that allows you to find and correct the error of one bit in a byte. Mainly used on servers.

Note: it is slower than usual, not suitable for people who love speed.

^ What is DDR SDRAM memory :

DDR memory (Double Data Rate) ensures data transmission along the memory-chipset bus twice per clock cycle, on both edges of the clock signal. Thus, when the system bus and memory operate at the same clock frequency, the memory bus bandwidth is twice that of conventional SDRAM.

The designation of DDR memory modules usually uses two parameters: either the operating frequency (equal to twice the clock frequency) - for example, the clock frequency of the DR-400 memory is 200 MHz; or peak throughput (in Mb/s). The same DR-400 has a throughput of approximately 3200 Mb/s, so it can be designated as PC3200. Currently, DDR memory has lost its relevance and in new systems is almost completely replaced by the more modern DDR2. however, to keep afloat a large number of older computers that have DDR memory installed, its production is still ongoing. The most common 184-pin DDR modules are PC3200 and, to a lesser extent, PC2700 standards. DDR SDRAM can have Registered and ECC options.

^ What is DDR2 memory :

DDR2 memory is the successor to DDR and is currently the dominant memory type for desktop computers, servers and workstations. DDR2 is designed to operate at higher frequencies than DDR, is characterized by lower power consumption, as well as a set of new functions (prefetching 4 bits per clock, built-in termination). In addition, unlike DDR chips, which were produced in both TSOP and FBGA packages, DDR2 chips are produced only in FBGA packages (which provides them with greater stability at high frequencies). DDR and DDR2 memory modules are not compatible with each other not only electrically, but also mechanically: DDR2 uses 240-pin strips, while DDR uses 184-pin strips. Today, the most common memory operating at 333 MHz and 400 MHz, designated DDR2-667 (PC2-5400/5300) and DDR2-800 (PC2-6400), respectively.

^ What is DDR3 memory :

Memory of the third generation DDR standard - DDR3 SDRAM should soon replace the current DDR2. The performance of the new memory has doubled compared to the previous one: now each read or write operation means access to eight groups of DDR3 DRAM data, which, in turn, are multiplexed across the I/O pins using two different reference oscillators at four times the clock speed frequency Theoretically, effective DDR3 frequencies will be located in the range of 800 MHz - 1600 MHz (at clock frequencies of 400 MHz - 800 MHz), thus, the DDR3 marking depending on the speed will be: DDR3-800, DDR3-1066, DDR3-1333, DDR3-1600 . Among the main advantages of the new standard, first of all, it is worth noting significantly lower power consumption (supply voltage DDR3 - 1.5 V, DDR2 - 1.8 V, DDR - 2.5 V).

The disadvantage of DDR3 versus DDR2 (and, especially, compared to DDR) is its high latency. DDR3 DIMM memory modules for desktop PCs will have a 240-pin structure, familiar to us from DDR2 modules; however, there will be no physical compatibility between them (due to the “mirror” pinout and different locations of the connector keys).

^ What is ECC memory :

ECC (Error Correct Code) is used to correct random memory errors caused by various external factors, and is an improved version of the “parity control” system. Physically, ECC is implemented in the form of an additional 8-bit memory chip installed next to the main ones. Thus, modules with ECC are 72-bit (as opposed to standard 64-bit modules). Some types of memory (Registered, Full Buffered) are available only in the ECC version.

^ What is Registered memory :

Registered memory modules are used mainly in servers that work with large amounts of RAM. All of them have ECC, i.e. are 72-bit and, in addition, contain additional register chips for partial (or complete - such modules are called Full Buffered, or FB-DIMM) data buffering, thereby reducing the load on the memory controller. Buffered DIMMs are generally incompatible with non-buffered ones.

^ What is SPD:

Any DIMM memory module contains a small SPD (Serial Presence Detect) chip, in which the manufacturer records information about the operating frequencies and the corresponding delays of the memory chips necessary to ensure normal operation of the module. Information from the SPD is read by the BIOS during the computer self-testing stage even before the operating system boots and allows you to automatically optimize memory access parameters.
Types/types of RAM modules

There are quite a few types of random access memory (RAM). This article describes their characteristics so that you can have an idea of different RAM, because... Not everyone is able to distinguish between types of RAM...

FPM (Fast Page Mode) is a type of dynamic memory. Its name corresponds to the principle of operation, since the module allows faster access to data that is on the same page as the data transferred during the previous cycle. These modules were used on most 486-based computers and early Pentium-based systems around 1995.

EDO (Extended Data Out) modules appeared in 1995 as a new type of memory for computers with Pentium processors. This is a modified version of FPM. Unlike its predecessors, EDO begins fetching the next block of memory at the same time it sends the previous block to the CPU.

SDRAM (Synchronous DRAM) is a type of random access memory that works so fast that it can be synchronized with the processor frequency, excluding standby modes. The microcircuits are divided into two blocks of cells so that while accessing a bit in one block, preparations are in progress for accessing a bit in another block. If the time to access the first piece of information was 60 ns, all subsequent intervals were reduced to 10 ns. Starting in 1996, most Intel chipsets began to support this type of memory module, making it very popular until 2001.

SDRAM can operate at 133 MHz, which is almost three times faster than FPM and twice as fast as EDO. Most computers with Pentium and Celeron processors released in 1999 used this type of memory.

DDR (Double Data Rate) was a development of SDRAM. This type of memory module first appeared on the market in 2001. The main difference between DDR and SDRAM is that instead of doubling the clock speed to speed things up, these modules transfer data twice per clock cycle. Now this is the main memory standard, but it is already beginning to give way to DDR2.

DDR2 (Double Data Rate 2) is a newer variant of DDR that should theoretically be twice as fast. DDR2 memory first appeared in 2003, and chipsets supporting it appeared in mid-2004. This memory, like DDR, transmits two sets of data per clock cycle. The main difference between DDR2 and DDR is the ability to operate at significantly higher clock speeds, thanks to improvements in design. But the modified operating scheme, which makes it possible to achieve high clock frequencies, at the same time increases delays when working with memory.

^ RAMBUS (RIMM)

RAMBUS (RIMM) is a type of memory that appeared on the market in 1999. It is based on traditional DRAM but with a radically changed architecture. The RAMBUS design makes memory access smarter, allowing pre-access to data while slightly offloading the CPU. The main idea used in these memory modules is to receive data in small packets but at a very high clock speed. For example, SDRAM can transfer 64 bits of information at 100 MHz, and RAMBUS can transfer 16 bits at 800 MHz. These modules did not become successful as Intel had many problems with their implementation. RDRAM modules appeared in the Sony Playstation 2 and Nintendo 64 game consoles.

Memory: RAM, DDR SDRAM, SDR SDRAM, PC100, DDR333, PC3200... how to figure it all out? Let's try!

So, the first thing we must do is “smooth out” all the doubts and questions about the denominations in memory...

The most common types of memory are:

SDR SDRAM(designations PC66, PC100, PC133)
DDR SDRAM(designations PC266, PC333, etc. or PC2100, PC2700)
RDRAM(PC800)

Now for subsequent explanations, I’ll tell you about timings and frequencies. Timing- this is the delay between individual operations performed by the controller when accessing memory.

Now let's take a closer look at each of the timings. Some of them are not available for configuration - access time CS# (crystal select) this signal determines the crystal (chip) on the module to carry out the operation.

In addition, the rest can be changed:

RCD (RAS-to-CAS Delay) this is the delay between signals RAS (Row Address Strobe) And CAS (Column Address Strobe), this parameter characterizes the interval between accesses to the bus by the signal memory controller RAS# And CAS#.
CAS Latency (CL) this is the delay between the read command and the availability of the first word to be read. Introduced to set address registers to guarantee a stable signal level.
RAS Precharge (RP) this is the time of re-issuance (charge accumulation period) of the RAS# signal - after what time the memory controller will be able to issue the line address initialization signal again.

Note: the order of operations is exactly this (RCD-CL-RP), but often timings are written not in order, but by “importance” - CL-RCD-RP.

Precharge Delay(or Active Precharge Delay; more often referred to as Tras) is the active time of the line. Those. the period during which a row is closed if the next required cell is in another row.
SDRAM Idle Timer(or SDRAM Idle Cycle Limit) the number of clock cycles during which the page remains open, after which the page is forced to close, either to access another page or to refresh (refresh)
Burst Length This is a parameter that sets the size of the memory prefetch relative to the starting address of the access. The larger its size, the higher the memory performance.

Well, we seem to have understood the basic concepts of timings, now let’s take a closer look at memory ratings (PC100, PC2100, DDR333, etc.)

There are two types of designations for the same memory: one by the "effective frequency" DDRxxx, and the second by the theoretical bandwidth PCxxxx.

Here is an example of notation correspondence:

100 MHz = PC1600 DDR SDRAM = DDR200 SDRAM = PC100 SDRAM = PC800 RDRAM
133 MHz = PC2100 DDR SDRAM = DDR266 SDRAM = PC133 SDRAM = PC1066 RDRAM
166 MHz = PC2700 DDR SDRAM = DDR333 SDRAM = PC166 SDRAM = PC1333 RDRAM
200 MHz = PC3200 DDR SDRAM = DDR400 SDRAM = PC200 SDRAM = PC1600 RDRAM
250 MHz = PC4000 DDR SDRAM = DDR500 SDRAM

As for RAMBUS (RDRAM) I won’t write much, but I’ll still try to introduce it to you.

Base RDRAM And Concurrent RDRAM Basically they differ only in operating frequencies: for the first, the frequency is 250-300 MHz, and for the second, this parameter is, accordingly, 300-350 MHz. Data is transmitted at two data packets per clock cycle, so the effective transmission frequency is twice as high. The memory uses an eight-bit data bus, which consequently gives a throughput of 500-600 Mb/s (BRDRAM) and 600-700 Mb/s (CRDRAM).

Direct RDRAM (DRDRAM) unlike Base and Concurrent, it has a 16-bit bus and operates at a frequency of 400 MHz. The bandwidth of Direct RDRAM is 1.6 Gb/s (taking into account bidirectional data transfer), which looks pretty good compared to SDRAM (1 Gb/s for PC133). Usually, when talking about RDRAM, they mean DRDRAM, so the letter "D" in the name is often omitted. When this type of memory appeared, Intel created a chipset for the Pentium 4 - i850.

The biggest plus Rambus memory means that the more modules, the greater the throughput, for example, up to 1.6 Gb/s per channel and up to 6.4 Gb/s with four channels.

There are also two disadvantages, quite significant:

1. The claws are gold and become unusable if the memory card is pulled out and inserted into the slot more than 10 times (approximately).

2. Overpriced, but many people find very good use for this memory and are willing to pay top dollar for them.

That's probably all, we've figured out the timings, names and denominations, now I'll tell you a little about various important little things.

You probably saw the By SPD option in the BIOS when setting the memory frequency, what does this mean? SPD - Serial Presence Detect, this is a microcircuit on the module into which all the parameters for the operation of the module are programmed, these are the “default values”, so to speak. Now, due to the emergence of “noname” companies, they began to write the manufacturer’s name and date into this chip.

Register memory

Everyone is now shouting about Dual channel - what is it?

Dual channel- double channel, this allows you to access two modules simultaneously. Dual channel is not a type of module, but a function integrated into the motherboard. Can be used with two (preferably) identical modules. It turns on automatically when there are 2 modules.

Note: To activate this function, you need to install modules in slots of different colors.

Parity and ECC

Memory with Parity This is a parity-check memory that can detect certain types of errors.

Memory with ECC This is an error correction memory that allows you to find and correct the error of one bit in a byte. Mainly used on servers.

Note: it is slower than usual, not suitable for people who love speed.

I hope that after reading the article you have understood the more popular “obscure concepts”.

- Faster, even faster, please speed up, at least a little, otherwise I’ll be…

– I can’t, dear Gamer, because I’ve reached my maximum clock frequency.

The dialogue of a Gamer, for whom every fraction of a second counts, could look something like this.

The clock speed of random access memory (RAM) is the second most important parameter after volume. The higher it is, the faster the data exchange between the processor and RAM occurs, the faster the computer works. RAM with low clock rates can become a bottleneck in resource-intensive games and programs. And if you don’t want to ask the capricious piece of hardware to speed up a little every time, always pay attention to this characteristic when purchasing. Today we’ll talk about how to find out the frequency of RAM based on the description in store catalogs, as well as the one installed on your PC.

How to understand what kind of “beast” a store offers

In the description of RAM modules on online store websites, sometimes not all of them are indicated, but only certain speed characteristics. For example:

DDR3, 12800 Mb/s.
DDR3, PC12800.
DDR3, 800 MHz (1600 MHz).
DDR3, 1600 MHz.

Some might think that this example is about four different planks. In fact, this can be used to describe the same RAM module with an effective frequency of 1600 MHz! And all these numbers indirectly or directly point to it.

To avoid any further confusion, let’s figure out what they mean:

12800 Mb/s is the memory bandwidth, an indicator obtained by multiplying the effective frequency (1600 MHz) by the bus width of one channel (64 bits or 8 bytes). Bandwidth describes the maximum amount of information that a RAM module is capable of transmitting in one clock cycle. I think it’s clear how to determine the effective frequency from it: you need to divide 12800 by 8.
PC12800 or PC3-12800– another designation for the throughput of a RAM module. By the way, a set of two strips intended for use in dual-channel mode has 2 times higher bandwidth, so its label may indicate PC25600 or PC3-25600.
800 MHz (1600 MHz)– two values, the first of which indicates the frequency of the memory bus itself, and the second - 2 times larger - its effective frequency. How are the indicators different? Computers, as you know, use DDR type RAM - with double the data transfer rate without increasing the number of bus cycles, that is, in 1 clock cycle not one, but two conventional pieces of information are transmitted through it. Therefore, the main indicator is considered to be the effective clock frequency (in this example, 1600 MHz).

The screenshot below shows a description of the speed characteristics of the RAM from the catalogs of three computer stores. As you can see, all sellers designate them differently.

Different RAM modules within the same generation - DDR, DDR2, DDR3 or DDR4 - have different frequency characteristics. Thus, the most common DDR3 RAM in 2017 is available with frequencies of 800, 1066, 1333, 1600, 1866, 2133 and 2400 MHz. Sometimes it is designated this way: DDR3-1333, DDR3-1866, etc. And this is convenient.

Not only the RAM has its own effective frequency, but also the device that controls it - the memory controller. In modern computer systems, starting with the Sandy Bridge generation, it is part of the processor. In older ones - as part of the north bridge components of the motherboard.

Almost all RAM can operate at lower clock speeds than specified in the specifications. RAM modules with different frequencies, provided that other parameters are similar, are compatible with each other, but can only function in single-channel mode.

If the computer has several RAM sticks with different frequency characteristics, the memory subsystem will exchange data at the speed of the slowest link (with the exception of devices). So, if the controller frequency is 1333 MHz, one of the strips is 1066 MHz, and the other is 1600 MHz, the transmission will proceed at a speed of 1066 MHz.

How to find out the frequency of RAM on a computer

Before learning how to determine the frequency indicators of RAM on a PC, let’s figure out how the computer itself recognizes them. It reads the information recorded in the SPD chip, which is equipped with each individual RAM stick. What this microcircuit looks like is shown in the photo below.

SPD data can also be read by programs, for example, the well-known utility, one of the sections of which is called “ SPD" In the screenshot below we see the already familiar characteristics of the speed of the RAM strip (field “ MaxBandwidth") - PC3-12800 (800 MHz). To find out its effective frequency, just divide 12800 by 8 or multiply 800 by 2. In my example, this figure is 1600 MHz.

However, in CPU-Z there is another section - “ Memory", and in it - the parameter " DRAMFrequency", equal to 665.1 MHz. This, as you probably guessed, is actual data, that is, the frequency mode in which the RAM actually operates. If we multiply 665.1 by 2, we get 1330.2 MHz - a value close to 1333 - the frequency at which the memory controller of this laptop operates.

In addition to CPU-Z, similar data is shown by other applications used to recognize and monitor PC hardware. Below are screenshots of the free utility:

RAM is used to temporarily store data necessary for the operation of the operating system and all programs. There should be enough RAM; if there is not enough, the computer starts to slow down.

The board with memory chips is called a memory module (or stick). Memory for a laptop, except for the size of the slots, is no different from memory for a computer, so when choosing, follow the same recommendations.

For an office computer, one 4 GB DDR4 stick with a frequency of 2400 or 2666 MHz is enough (costs almost the same).
RAM Crucial CT4G4DFS824A

For a multimedia computer (movies, simple games), it is better to take two 4 GB DDR4 sticks with a frequency of 2666 MHz, then the memory will work in a faster dual-channel mode.
RAM Ballistix BLS2C4G4D240FSB

For a mid-class gaming computer, you can take one 8 GB DDR4 stick with a frequency of 2666 MHz so that in the future you can add another one, and it would be better if it is a simpler running model.
RAM Crucial CT8G4DFS824A

And for a powerful gaming or professional PC, you need to immediately take a set of 2 DDR4 8 GB sticks, and a frequency of 2666 MHz will be quite sufficient.

2. How much memory is needed

For an office computer designed for working with documents and accessing the Internet, one 4 GB memory stick is sufficient.

For a multimedia computer that can be used to watch high-quality videos and undemanding games, 8 GB of memory is enough.

For a mid-range gaming computer, the minimum option is 8 GB of RAM.

A powerful gaming or professional computer requires 16 GB of memory.

A larger amount of memory may be needed only for very demanding professional programs and is not needed by ordinary users.

Memory capacity for older PCs

If you decide to increase the memory on your old computer, keep in mind that 32-bit versions of Windows do not support more than 3 GB of RAM. That is, if you install 4 GB of RAM, the operating system will see and use only 3 GB.

As for 64-bit versions of Windows, they will be able to use all the installed memory, but if you have an old computer or an old printer, then they may not have drivers for these operating systems. In this case, before purchasing memory, install the 64-bit version of Windows and check if everything works for you. I also recommend looking at the website of the motherboard manufacturer and seeing what volume of modules and total amount of memory it supports.

Please also note that 64-bit operating systems consume 2 times more memory, for example, Windows 7 x64 takes about 800 MB for its needs. Therefore, 2 GB of memory for such a system will not be enough, preferably at least 4 GB.

Practice shows that modern operating systems Windows 7,8,10 are fully operational with a memory capacity of 8 GB. The system becomes more responsive, programs open faster, and jerks (freezes) disappear in games.

3. Memory types

Modern memory is of the DDR SDRAM type and is constantly being improved. So DDR and DDR2 memory is already obsolete and can only be used on older computers. DDR3 memory is no longer advisable to use on new PCs; it has been replaced by the faster and more promising DDR4.

Please note that the selected memory type must be supported by the processor and motherboard.

Also, new processors, for compatibility reasons, can support DDR3L memory, which differs from regular DDR3 in reduced voltage from 1.5 to 1.35 V. Such processors will be able to work with regular DDR3 memory if you already have it, but processor manufacturers do not recommend this because -due to increased degradation of memory controllers designed for DDR4 with an even lower voltage of 1.2 V.

Memory type for older PCs

Outdated DDR2 memory costs several times more than more modern memory. A 2 GB DDR2 stick costs 2 times more, and a 4 GB DDR2 stick costs 4 times more than a DDR3 or DDR4 stick of the same size.

Therefore, if you want to significantly increase the memory on an old computer, then perhaps the best option would be to switch to a more modern platform by replacing the motherboard and, if necessary, a processor that will support DDR4 memory.

Calculate how much it will cost you; perhaps a profitable solution would be to sell the old motherboard with old memory and purchase new, albeit not the most expensive, but more modern components.

The motherboard connectors for installing memory are called slots.

Each memory type (DDR, DDR2, DDR3, DDR4) has its own slot. DDR3 memory can only be installed in a motherboard with DDR3 slots, DDR4 - with DDR4 slots. Motherboards that support old DDR2 memory are no longer produced.

5. Memory characteristics

The main characteristics of memory on which its performance depends are frequency and timings. Memory speed does not have as strong an impact on the overall performance of the computer as the processor. However, you can often get faster memory for not much more. Fast memory is needed primarily for powerful professional computers.

5.1. Memory frequency

Frequency has the greatest impact on memory speed. But before purchasing it, you need to make sure that the processor and motherboard also support the required frequency. Otherwise, the actual memory operating frequency will be lower and you will simply overpay for something that will not be used.

Inexpensive motherboards support lower maximum memory frequencies, for example for DDR4 it is 2400 MHz. Mid-range and high-end motherboards can support higher frequency memory (3400-3600 MHz).

But with processors the situation is different. Older processors with DDR3 memory support may support memory with a maximum frequency of 1333, 1600, or 1866 MHz (depending on the model). For modern processors that support DDR4 memory, the maximum supported memory frequency may be 2400 MHz or higher.

Intel 6th generation and higher processors and AMD Ryzen processors support DDR4 memory at 2400 MHz or higher. Moreover, their lineup includes not only powerful expensive processors, but also mid-range and budget-class processors. Thus, you can build a computer on the most modern platform with an inexpensive processor and DDR4 memory, and in the future change the processor and get the highest performance.

The main memory today is DDR4 2400 MHz, which is supported by the most modern processors, motherboards and costs the same as DDR4 2133 MHz. Therefore, purchasing DDR4 memory with a frequency of 2133 MHz today does not make sense.

You can find out what memory frequency a particular processor supports on the manufacturers’ websites:

By model number or serial number it is very easy to find all the characteristics of any processor on the website:

Or simply enter the model number in the Google or Yandex search engine (for example, “Ryzen 7 1800X”).

5.2. High Frequency Memory

Now I want to touch on another interesting point. On sale you can find RAM at a much higher frequency than any modern processor supports (3000-3600 MHz and higher). Accordingly, many users are wondering how this can be?

It's all about a technology developed by Intel, eXtreme Memory Profile (XMP). XMP allows memory to run at a higher frequency than the processor officially supports. XMP must be supported by both the memory itself and the motherboard. High-frequency memory simply cannot exist without support for this technology, but not all motherboards can boast of its support. These are mainly more expensive models above the middle class.

The essence of XMP technology is that the motherboard automatically increases the frequency of the memory bus, due to which the memory begins to operate at its higher frequency.

AMD has a similar technology called AMD Memory Profile (AMP), which was supported by older AMD processor motherboards. These motherboards usually also supported XMP modules.

Purchasing more expensive memory with a very high frequency and a motherboard with XMP support makes sense for very powerful professional computers equipped with a top-end processor. In a middle-class computer, this will be wasted money, since everything will depend on the performance of other components.

In games, the memory frequency has a small impact and there is no point in overpaying; it will be enough to go for 2400 MHz, or 2666 MHz if the price difference is small.

For professional applications, you can take memory with a higher frequency - 2666 MHz or, if you want and have funds, 3000 MHz. The difference in performance here is greater than in games, but not dramatic, so there is no particular point in pushing the memory frequency.

Let me remind you once again that your motherboard must support memory at the required frequency. In addition, sometimes Intel processors become unstable at memory frequencies above 3000 MHz, and for Ryzen this limit is around 2900 MHz.

Timings are the delays between read/write/copy operations of data in RAM. Accordingly, the fewer these delays, the better. But timings have a much smaller impact on memory speed than its frequency.

There are only 4 main timings that are indicated in the characteristics of memory modules.

Of these, the most important is the first number, which is called latency (CL).

Typical latency for DDR3 1333 MHz memory is CL 9, for higher frequency DDR3 memory is CL 11.

Typical latency for DDR4 2133 MHz memory is CL 15, for DDR4 memory with a higher frequency is CL 16.

You should not purchase memory with a latency higher than specified, as this indicates an overall low level of its technical characteristics.

Typically, memory with lower timings is more expensive, but if the price difference is not significant, then memory with lower latency should be preferred.

5.4. Supply voltage

Memory may have different supply voltages. It can be either standard (generally accepted for a certain type of memory), or increased (for enthusiasts) or, conversely, reduced.

This is especially important if you want to add memory to your computer or laptop. In this case, the voltage of the new strips should be the same as the existing ones. Otherwise, problems are possible, since most motherboards cannot set different voltages for different modules.

If the voltage is set to a level with a lower voltage, then others may not have enough power and the system will not work stably. If the voltage is set to a level with a higher voltage, then the memory designed for a lower voltage may fail.

If you are building a new computer, then this is not so important, but to avoid possible compatibility problems with the motherboard and replacing or expanding memory in the future, it is better to choose sticks with a standard supply voltage.

The memory, depending on the type, has the following standard supply voltages:

DDR - 2.5 V
DDR2 - 1.8 V
DDR3 - 1.5 V
DDR3L - 1.35 V
DDR4 - 1.2 V

I think you noticed that there is DDR3L memory in the list. This is not a new type of memory, but regular DDR3, but with a reduced supply voltage (Low). This is the kind of memory needed for 6th generation Intel processors and higher, which support both DDR4 and DDR3 memory. But in this case, it is better to build the system on new DDR4 memory.

6. Marking of memory modules

Memory modules are marked depending on the type of memory and its frequency. The marking of DDR memory modules begins with PC, followed by a number indicating the generation and speed in megabytes per second (MB/s).

Such markings are inconvenient to navigate; it is enough to know the type of memory (DDR, DDR2, DDR3, DDR4), its frequency and latency. But sometimes, for example on ad sites, you can see markings copied from the strip. Therefore, so that you can get your bearings in this case, I will give the markings in a classic form, indicating the type of memory, its frequency and typical latency.

DDR - obsolete

PC-2100 (DDR 266 MHz) - CL 2.5
PC-2700 (DDR 333 MHz) - CL 2.5
PC-3200 (DDR 400 MHz) - CL 2.5

DDR2 - obsolete

PC2-4200 (DDR2 533 MHz) - CL 5
PC2-5300 (DDR2 667 MHz) - CL 5
PC2-6400 (DDR2 800 MHz) - CL 5
PC2-8500 (DDR2 1066 MHz) - CL 5

DDR3 - obsolete

PC3-10600 (DDR3 1333 MHz) - CL 9
PC3-12800 (DDR3 1600 MHz) - CL 11
PC3-14400 (DDR3 1866 MHz) - CL 11
PC3-16000 (DDR3 2000 MHz) - CL 11

PC4-17000 (DDR4 2133 MHz) - CL 15
PC4-19200 (DDR4 2400 MHz) - CL 16
PC4-21300 (DDR4 2666 MHz) - CL 16
PC4-24000 (DDR4 3000 MHz) - CL 16
PC4-25600 (DDR4 3200 MHz) - CL 16

DDR3 and DDR4 memory may have a higher frequency, but only top processors and more expensive motherboards can work with it.

7. Design of memory modules

Memory sticks can be single-sided, double-sided, with or without radiators.

7.1. Chip placement

Chips on memory modules can be placed on one side of the board (single-sided) or on both sides (double-sided).

This doesn't matter if you are purchasing memory for a new computer. If you want to add memory to an old PC, then it is advisable that the arrangement of chips on the new stick be the same as on the old one. This will help avoid compatibility issues and increase the likelihood of memory operating in dual-channel mode, which we will talk about later in this article.

Now on sale you can find many memory modules with aluminum radiators of various colors and shapes.

The presence of heatsinks can be justified on DDR3 memory with a high frequency (1866 MHz or more), since it heats up more. At the same time, ventilation must be well organized in the housing.

Modern DDR4 RAM with a frequency of 2400, 2666 MHz practically does not heat up and the radiators on it will be purely decorative. They can even get in the way, because after a while they become clogged with dust, which is difficult to clean out of them. In addition, such memory will cost slightly more. So, if you want, you can save on this, for example, by taking excellent Crucial 2400 MHz memory without heatsinks.

Memory with a frequency of 3000 MHz or more also has an increased supply voltage, but it also does not heat up very much and in any case there will be heatsinks on it.

8. Memory for laptops

Memory for laptops differs from memory for desktop computers only in the size of the memory module and is labeled SO-DIMM DDR. Just like for desktop computers, memory for laptops has types DDR, DDR2, DDR3, DDR3L, DDR4.

In terms of frequency, timings and supply voltage, memory for laptops does not differ from memory for computers. But laptops only come with 1 or 2 memory slots and have stricter maximum capacity limits. Be sure to check these parameters before choosing memory for a specific laptop model.

9. Memory operating modes

The memory can operate in Single Channel, Dual Channel, Triple Channel or Quad Channel mode.

In single-channel mode, data is written sequentially to each module. In multi-channel modes, data is written in parallel to all modules, which leads to a significant increase in the speed of the memory subsystem.

Single-channel memory mode is limited only to hopelessly outdated motherboards with DDR memory and the first models with DDR2.

All modern motherboards support dual-channel memory mode, while three-channel and quad-channel modes are supported only by a few models of very expensive motherboards.

The main condition for dual-channel mode operation is the presence of 2 or 4 memory sticks. Three-channel mode requires 3 or 6 memory sticks, and four-channel mode requires 4 or 8 memory sticks.

It is desirable that all memory modules are the same. Otherwise, dual-channel operation is not guaranteed.

If you want to add memory to an old computer and your motherboard supports dual-channel mode, try to choose a stick that is as identical in all respects as possible. It is best to sell the old one and buy 2 new identical strips.

In modern computers, memory controllers have been moved from the motherboard to the processor. Now it is not so important that the memory modules are the same, since the processor will still be able to activate dual-channel mode in most cases. This means that if in the future you want to add memory to a modern computer, you will not necessarily need to look for exactly the same module; you just need to choose the one that is most similar in characteristics. But I still recommend that the memory modules be the same. This will give you a guarantee of its fast and stable operation.

With the transfer of memory controllers to the processor, 2 more modes of dual-channel memory operation appeared - Ganged (paired) and Unganged (unpaired). If the memory modules are the same, the processor can work with them in Ganged mode, as before. If the modules differ in characteristics, the processor can activate the Unganged mode to eliminate distortions in working with memory. In general, the memory speed in these modes is almost the same and makes no difference.

The only downside to dual-channel mode is that multiple memory modules are more expensive than one of the same size. But if you are not very strapped for money, then buy 2 sticks, the memory speed will be much higher.

If you need, say, 16 GB of RAM, but you can’t afford it yet, then you can buy one 8 GB stick so that you can add another one of the same kind in the future. But it is still better to purchase two identical strips at once, since later you may not be able to find the same one and you will encounter a compatibility problem.

10. Memory module manufacturers

One of the best price/quality ratios today comes from the memory of the impeccably proven Crucial brand, which has modules from budget to gaming (Ballistix).

Competing with it is the well-deserved Corsair brand, whose memory is somewhat more expensive.

As an inexpensive but high-quality alternative, I especially recommend the Polish brand Goodram, which has bars with low timings at a low price (Play line).

For an inexpensive office computer, simple and reliable memory made by AMD or Transcend will be sufficient. They have proven themselves to be excellent and there are practically no problems with them.

In general, the Korean companies Hynix and Samsung are considered leaders in memory production. But now modules of these brands are mass-produced in cheap Chinese factories, and among them there are a lot of fakes. Therefore, I do not recommend purchasing memory from these brands.

An exception may be Hynix Original and Samsung Original memory modules, which are manufactured in Korea. These strips are usually blue, their quality is considered better than those made in China and the guarantee for them is slightly higher. But in terms of speed characteristics, they are inferior to memory with lower timings from other quality brands.

Well, for enthusiasts and fans of modding there are affordable overclocking brands GeIL, G.Skill, Team. Their memory is distinguished by low timings, high overclocking potential, unusual appearance and costs a little less than the well-promoted Corsair brand.

There is also a wide range of memory modules on sale from the very popular manufacturer Kingston. Memory sold under the budget Kingston brand has never been of high quality. But they have a top-end HyperX series, which is deservedly popular, which can be recommended for purchase, but is often overpriced.

11. Memory packaging

It is better to purchase memory in individual packaging.

It is usually of higher quality and is much less likely to be damaged in transit than memory that comes loose.

12. Increase memory

If you are planning to add memory to an existing computer or laptop, then first find out what the maximum memory capacity and total memory capacity is supported by your motherboard or laptop.

Also check how many memory slots are on the motherboard or laptop, how many of them are occupied and what kind of memory sticks are installed in them. It's better to do it visually. Open the case, take out the memory sticks, examine them and write down all the characteristics (or take a photo).

If for some reason you don’t want to get into the case, you can view the memory parameters in the program on the SPD tab. This way you won't know if the stick is single-sided or double-sided, but you can find out the memory characteristics if there is no sticker on the stick.

There is a base and effective memory frequency. The CPU-Z program and many similar ones show the base frequency, it must be multiplied by 2.

Once you know how much memory you can increase, how many free slots are available, and what kind of memory you have installed, you can begin to explore the possibilities of increasing memory.

If all memory slots are occupied, then the only way to increase memory is to replace existing memory sticks with new ones of larger capacity. And old planks can be sold on an advertisement site or exchanged at a computer store when purchasing new ones.

If there are free slots, then you can add new memory sticks to the existing ones. In this case, it is desirable that the new strips be as close as possible to the characteristics of those already installed. In this case, you can avoid various compatibility problems and increase the chances that the memory will work in dual-channel mode. To do this, the following conditions must be met, in order of importance.

The memory type must match (DDR, DDR2, DDR3, DDR3L, DDR4).
The supply voltage for all strips must be the same.
All planks must be single-sided or double-sided.
The frequency of all bars must match.
All strips must be of the same volume (for dual-channel mode).
The number of strips must be even: 2, 4 (for dual-channel mode).
It is desirable that the latency (CL) matches.
It is desirable that the strips are from the same manufacturer.

The easiest place to start choosing is with the manufacturer. Choose from the online store catalog strips of the same manufacturer, volume and frequency as those installed in yours. Make sure that the supply voltage matches and check with your consultant whether they are single-sided or double-sided. If the latency also matches, then generally good.

If you were unable to find strips from the same manufacturer with similar characteristics, then choose all the others from the list of recommended ones. Then again look for strips of the required volume and frequency, check the supply voltage and check whether they are single-sided or double-sided. If you are unable to find similar planks, then look in another store, catalog or ad site.

The best option is always to sell all the old memory and buy 2 new identical sticks. If the motherboard does not support the brackets of the required volume, you may have to buy 4 identical brackets.

13. Setting up filters in the online store

Go to the “RAM” section on the seller’s website.
Select recommended manufacturers.
Select the form factor (DIMM - PC, SO-DIMM - laptop).
Select the memory type (DDR3, DDR3L, DDR4).
Select the required volume of slats (2, 4, 8 GB).
Select the maximum frequency supported by the processor (1600, 1866, 2133, 2400 MHz).
If your motherboard supports XMP, add higher frequency memory (2666, 3000 MHz) to the selection.
Sort the selection by price.
Consistently look through all items, starting with the cheapest ones.
Select several strips that match the frequency.
If the price difference is acceptable to you, take sticks with a higher frequency and lower latency (CL).

Thus, you will get the optimal price/quality/speed ratio of memory at the lowest possible cost.

14. Links

RAM Corsair CMK16GX4M2A2400C16
RAM Corsair CMK8GX4M2A2400C16
RAM Crucial CT2K4G4DFS824A

This tab describes the data SPD- a mechanism used to determine the presence and characteristics of memory modules. Stands for serial presence detect, consistent determination of availability. The word serial indicates the type of bus used, I2C - it is just serial. Tire I2C included in SMBus, developed by Intel, so if you disable detection of devices on the SMBus bus in CPU-Z, then SPD data will not be displayed. If you look at the memory module, you can see a small chip, different from memory chips, which has eight legs. This is the so-called SPD chip. In essence, this is an ordinary “flash drive” - a flash memory chip similar to those that store the BIOS of a motherboard and video cards (and other various peripherals).

Almost all motherboards set timings and frequencies based on SPD data, so errors in this data can lead to the system being unable to start. Problems especially often arise with modules designed for enthusiasts. Sometimes the frequencies and timings hardwired into the SPD are intended for use at higher voltages, which makes it impossible to boot at a standard voltage and you need to find a regular module, set the required voltage in the BIOS and then plug in the original modules. At least Corsair had this problem. Another example is when the manufacturer writes on the sticker the frequencies and timings and voltages at which the memory can be operated, but in order to boot, it writes safe frequencies in the SPD, greatly overestimated, or overestimated timings. And then newbies have questions: why did you buy DDR2-1066 memory, but it is defined as DDR2-800?

And now, in fact, the data that we can see on this tab. First group Memory Slot Selection:

combo box for selecting a module. Allows you to select the memory module for which SPD information is displayed.
on the right there is a field with the name of the memory type, in our case - DDR2.
Module Size- module size in megabytes.
Max. Bandwith- maximum throughput. In this case, PC2 means DDR2 memory, and the number after that means the maximum bandwidth in megabytes. The actual DDR bus frequency is indicated in brackets. The bandwidth is calculated using the formula: Freq * 64 * 2 / 8, where 64 is the memory bus width in bits (for all modules SDRAM it is equal to 64 bits), 2 means DDR technology, which doubles the bandwidth, and dividing by 8 converts bits to bytes (1 byte has 8 bits). Yes, for DDR2-800 with a real frequency of 400 MHz we get: 400*64*2/8= 6400MB/s, which is what CPU-Z shows.
Manufacturer- name of the memory module manufacturer. Usually not filled in Noname(unnamed) manufacturers.
Part Number- batch number. Likewise, not filled in Noname.
Serial Number- serial number of the module. Unnamed manufacturers make one firmware, so the concept of serialization does not exist at all.
Correction- presence of an error correction module. It does not occur on regular memory, and such a module is easy to distinguish by the “extra” memory chip. If a regular module has 4 or 8 chips on one side, then this one has 5 or 9. It is located in the middle. On some modules you can see the space on the board for this chip.
Registered- presence of register memory. Not of interest to enthusiasts.
Buffered- presence of buffered memory. Again, not of interest to enthusiasts.
SPD Ext.- availability of SPD extensions. SPD is developed by the organization JEDEC, which is involved in the adoption of standards in the field of memory. But the company NVIDIA proposed using bytes not used by the standard (and there are many of them) for high-speed profiles, where not only the main and additional timings will be specified, but also the voltage. She named her standard EPP - enhanced performance profile(Enhanced Performance Profile). Following her Intel added to its chipsets support for similar profiles with the name XMP - extreme memory profile(extreme memory profile). Profiles have been made for beginners who cannot overclock and set the necessary settings themselves, so they are not recommended for enthusiasts. The memory module supports either EPP or XMP, but the point here is not so much that both algorithms use adjacent bytes. The main reason is, of course, political. The memory must receive the blessing of either one company or another to proclaim support for the profile. It is technically possible to support both, but of course it will not be approved.
Week/Year - week and year of release.

Next group - Timings Table- table of timings for different frequencies. Column labels indicate the number of the table created according to the standard JEDEC, or profile EPP/XMP, if there is one.

Frequency- memory frequency. As stated, it may differ from what is written on the label, which is usually normal if the memory can operate at the frequency stated by the manufacturer.
CAS# Latency- minimum time between issuing a read command ( CAS#) and the start of data transfer (read delay).
RAS# to CAS#- the time required to activate a bank row, or the minimum time between the signal to select a row (RAS#) and the signal to select a column ( CAS#).
RAS# Precharge- time required to precharge the bank (precharge). In other words, the minimum time for closing a line, after which a new bank line can be activated.
tRAS- the minimum time the line is active, that is, the minimum time between the activation of the line (its opening) and the issuance of a precharge command (the beginning of closing the line).
tRC- minimum time between activation of lines of one bank. Is a combination of timings tRAS+tRP- the minimum time the line is active and the time it closes (after which you can open a new one).
Command Rate- the time required for the controller to decode commands and addresses. Otherwise, the minimum time between issuing two commands. Only used in advanced profiles.
Voltage- voltage used. JEDEC uses only the standard value, so this field will only differ in advanced profiles.