home Mts

New RAM standard - DDR4 RAM, specifications and features. DDR3 vs. DDR4. Theoretical differences ddr4 memory bandwidth

New generations of processors have stimulated the development of faster SDRAM (Synchronous Dynamic Random Access Memory) with a clock frequency of 66 MHz, and memory modules with such microcircuits are called DIMM (Dual In-line Memory Module).
For use with Athlon processors, and then with Pentium 4, the second generation of SDRAM chips was developed - DDR SDRAM (Double Data Rate SDRAM). DDR SDRAM technology allows data to be transmitted on both edges of each clock pulse, which provides the ability to double the memory bandwidth. With the further development of this technology in DDR2 SDRAM microcircuits, it was possible to transmit already 4 portions of data in one clock pulse. Moreover, it should be noted that the increase in performance occurs due to the optimization of the addressing and read / write process of memory cells, but the clock frequency of the memory matrix does not change. Therefore, the overall performance of the computer does not increase by two or four times, but only by tens of percent. In fig. shows the frequency principles of operation of SDRAM microcircuits of various generations.

There are the following types of DIMMs:

72-pin SO-DIMM (Small Outline Dual In-line Memory Module) - used for FPM DRAM (Fast Page Mode Dynamic Random Access Memory) and EDO DRAM (Extended Data Out Dynamic Random Access Memory)

100-pin DIMM - used for SDRAM (Synchronous Dynamic Random Access Memory) printers

144-pin SO-DIMM - used for SDR SDRAM (Single Data Rate ...) in laptop computers

168-pin DIMM - used for SDR SDRAM (less often for FPM / EDO DRAM in workstations / servers

172-pin MicroDIMM - used for DDR SDRAM (Double date rate)

184-pin DIMM - used for DDR SDRAM

200-pin SO-DIMM - used for DDR SDRAM and DDR2 SDRAM

214-pin MicroDIMM - used for DDR2 SDRAM

204-pin SO-DIMM - used for DDR3 SDRAM

240-pin DIMM - used for DDR2 SDRAM, DDR3 SDRAM and FB-DIMM (Fully Buffered) DRAM

244-pin Mini-DIMM - for Mini Registered DIMM

256-pin SO-DIMM - used for DDR4 SDRAM

284-pin DIMM - used for DDR4 SDRAM

To prevent the installation of the wrong type of DIMM-module, several slots (keys) are made in the textolite board of the module among the contact pads, as well as on the right and left in the area of the module fixing elements on the system board. For mechanical identification of various DIMMs, a shift in the position of two keys in the textolite board of the module, located among the contact pads, is used. The main purpose of these keys is to prevent the installation of a DIMM with an unsuitable voltage supply for memory chips into the socket. In addition, the location of the key or keys determines the presence or absence of a data buffer, etc.

DDR modules are labeled PC. But unlike SDRAM, where PC indicated the operating frequency (for example, PC133 - the memory is designed to operate at 133MHz), the PC indicator in DDR modules indicates the maximum achievable bandwidth, measured in megabytes per second.

DDR2 SDRAM

Name of the standard	Memory type	Memory frequency	Bus frequency	Data transfer per second (MT / s)
PC2-3200	DDR2-400	100 MHz	200 MHz	400	3200 MB / s
PC2-4200	DDR2-533	133 MHz	266 MHz	533	4200 MB / s
PC2-5300	DDR2-667	166 MHz	333 MHz	667	5300 MB / s
PC2-5400	DDR2-675	168 MHz	337 MHz	675	5400 MB / s
PC2-5600	DDR2-700	175 MHz	350 MHz	700	5600 MB / s
PC2-5700	DDR2-711	177 MHz	355 MHz	711	5700 MB / s
PC2-6000	DDR2-750	187 MHz	375 MHz	750	6000 MB / s
PC2-6400	DDR2-800	200 MHz	400 MHz	800	6400 MB / s
PC2-7100	DDR2-888	222 MHz	444 MHz	888	7100 MB / s
PC2-7200	DDR2-900	225 MHz	450 MHz	900	7200 MB / s
PC2-8000	DDR2-1000	250 MHz	500 MHz	1000	8000 MB / s
PC2-8500	DDR2-1066	266 MHz	533 MHz	1066	8500 MB / s
PC2-9200	DDR2-1150	287 MHz	575 MHz	1150	9200 MB / s
PC2-9600	DDR2-1200	300 MHz	600 MHz	1200	9600 MB / s

DDR3 SDRAM

Name of the standard	Memory type	Memory frequency	Bus frequency	Data transmissions per second (MT / s)	Peak data rate
PC3-6400	DDR3-800	100 MHz	400 MHz	800	6400 MB / s
PC3-8500	DDR3-1066	133 MHz	533 MHz	1066	8533 MB / s
PC3-10600	DDR3-1333	166 MHz	667 MHz	1333	10667 MB / s
PC3-12800	DDR3-1600	200 MHz	800 MHz	1600	12800 MB / s
PC3-14400	DDR3-1800	225 MHz	900 MHz	1800	14400 MB / s
PC3-16000	DDR3-2000	250 MHz	1000 MHz	2000	16000 MB / s
PC3-17000	DDR3-2133	266 MHz	1066 MHz	2133	17066 MB / s
PC3-19200	DDR3-2400	300 MHz	1200 MHz	2400	19200 MB / s

Peak values are indicated in the tables; in practice, they may not be achievable.
For a comprehensive assessment of the capabilities of RAM, the term memory bandwidth is used. It takes into account the frequency at which data is transmitted and the width of the bus and the number of memory channels.

Bandwidth = Bus frequency x channel width x number of channels

For all DDRs - the number of channels = 2 and the width is 64 bits.
For example, when using DDR2-800 memory with a 400 MHz bus frequency, the bandwidth will be:

(400 MHz x 64 bit x 2) / 8 bit = 6400 MB / s

Each manufacturer gives each of its products or parts its internal production marking, called P / N (part number) - part number.
For memory modules from different manufacturers, it looks like this:

Kingston KVR800D2N6 / 1G
OCZ OCZ2M8001G
Corsair XMS2 CM2X1024-6400C5

On the site of many memory manufacturers, you can study how their Part Number is read.

Kingston part number	Description
KVR1333D3D4R9SK2 / 16G	16GB 1333MHz DDR3 ECC Reg CL9 DIMM (Kit of 2) DR x4 w / TS

We checked what new DDR4 RAM modules differ from the previously used DDR3 memory modules and how much more efficient they are than the previous generation of equipment.

The first DDR4 memory information appeared in 2008. Then it was assumed that it will hit stores within five years and very quickly gain more popularity than DDR3 memory.

It soon became clear, however, that the memory used has a very large potential for development, and a quick transition to a new standard does not make sense. Therefore, although several years ago computer companies showed their DDR4 memory models at various exhibitions, due to the lack of supporting platforms, there was no possibility of its practical application. Everything has changed over the past year. DDR3 memory has reached its limit.

The end of the DDR3 era

Although there are modules available on the market with DDR-2400, DDR-2800, and even faster, further acceleration was almost impossible. True, some manufacturers managed to get a higher clock frequency, but the creation of such memory modules on a mass scale was unprofitable and practically impossible.

Constant acceleration of the existing type of random access memory is impractical - power consumption increases significantly and fault tolerance decreases. The solution to these problems turned out to be just DDR4 memory modules, which have much more opportunities to increase performance, and at the same time consume much less energy.

Do we need new DDR4 memory modules

Yes, and not only because of the speed. The first memory models of the new type do not have higher performance than the modules of the previous generation.

When one technology reaches its limits, and the second is just entering the market, we cannot tell the difference in speed between them. In this case, it is not the productivity that is essential, but the prospects of the new technology. Therefore, already now, when upgrading a computer, it is worth considering the choice of a platform compatible with the new type of memory.

At the next replacement of components, we will be able to use the modules of the new generation. If after some time we understand that memory is too slow or insufficiently capacious, then we will not have problems buying more powerful components - otherwise the situation is in the case of DDR3 memory, which has reached its limit and will slowly leave the market.

One of DDR4 memory strengths is its energy efficiency. Currently, the vast majority of computers sold are laptops, tablets and convertible devices. The most important feature of such equipment is its performance and operating time without recharging, which, in turn, depends on energy consumption.

Is it worth changing the memory to DDR4

So far, this dilemma does not exist, because a platform is not yet available that supports both types of memory. Therefore, replacing memory will replace the entire platform. However, expect to see motherboards available soon that support both types of memory.

Is it worth then change DDR3 memory to DDR4? When choosing a new computer with a future upgrade in mind, it may be helpful to consider this option. Of course, then initially we will spend more, but this will make it easier to upgrade the computer in the future.

A new kind of memory modules

Minor changes have taken place in the appearance of the memory module. True, its length and thickness are the same as in the case of DDR3, but an experienced eye will notice that the new modules are a millimeter taller, and have 284 instead of 240 pins.

In addition, the contacts in the central part of the module are higher than those at the edges. As a result, installing memory will require less effort. The position of the indents in the module has also changed. This procedure makes it impossible for the memory to be placed in an inappropriate slot, such as for DDR3 memory modules.

High speed DDR4

Currently on the market you can find mainly DDR3 memory at frequencies of 1333 and 1600 M / s (millions of operations per second), and modules intended for enthusiasts reach frequencies of the order of 2400 or 2866 M / s.

In the case of DDR4, these parameters will be better, and operation at the level of 2400 M / s will become practically the standard. The JEDEC standard assumes so far the creation of DDR4 memory at speeds from 1600 to 3200 M / s, but modules of the 4166 M / s level have already been announced.

Large delays

An increase in memory speed always entails an increase in latency, expressed in cycles of time. The situation will be similar this time. The JEDEC standard stipulates that the standard CAS latency for DDR4-2400 memory will be 15 cycles (for DDR3-1600 it was 10 cycles).

Keep in mind, however, that delays do not actually change. To verify this, it is enough to make a simple calculation. The memory speed at 1600 M / s means its actual clock frequency of size 00 MHz. This means that one cycle lasts 1/800 000 0000 sec. In this case, the delay, expressed in 10 cycles, is 12.5 nanoseconds. After performing the appropriate calculations for 2400 M / s memory and 15 cycles, we will get an identical result.

Large memory capacity

The largest DDR3 modules have a capacity of 8 GB. In the case of DDR4 memory, 32GB capacity can be easily achieved. Assuming that a standard motherboard can accommodate four memory modules, this means that a computer equipped with 128GB of RAM will soon be a reality.

Performance difference

Even if there is a large difference in baud rate between the two modules, there will be little or no performance change in most programs.

It will only be possible to see the benefits of using faster memory when using the most demanding applications and games.

Less energy consumption

DDR3 requires 1.5V and DDR4 only 1.2V. According to the manufacturers, this voltage change should provide energy savings of 30%. In practice, the savings are somewhat lower, however, thanks to the use of several innovations, it was possible to achieve the desired floor.

Changing the type of signaling and the use of DBI, that is, the method of inverting the memory bus, helped. It consists in the fact that if in a particular data line most of the information is zeros, they are replaced by ones, and the dedicated controller perceives the data line as inverted.

This allows the transistors to turn off and on more rarely, which reduces power consumption and improves signal stability.

The advent of DDR4 RAM on the market shook the unshakable positions of its predecessor. It has higher technical characteristics and many users have a natural question, which RAM bar is better? Numerous tests and comparisons of 4th generation RAM with DDR3 show the difference between the two. When choosing a DDR3 memory module, keep in mind that it has no DDR4 compatibility.

A computer is one of the components that is responsible for its performance: the speed of information processing and the maximum amount of data being processed at the moment. Until 2015, the first positions were firmly held by the third generation DDR3 RAM, but with the advent of DDR4, the situation began to change towards the latest modification. The appearance of the fourth generation RAM caused a great stir in the computer equipment market, at the same time a natural question arose, which is better than DDR3 or DDR4 and is the appearance of the latest model a common marketing ploy?

DDR4 development history

JEDEK started developing the fourth generation RAM back in 2005, when the most modern modification was DDR2. Already at that time, the company's engineers realized that the second generation of RAM would not be able to meet the requirements of rapidly developing processors and other PC components. Even the announced release of the third generation RAM will not be able to fully cope with the task. To solve the problem, it is not enough to simply increase the data processing speed as it was done in DDR3. Consideration should be given to parameters such as power consumption and volume, which affect the throughput of the device.

Attention! For working with specialized programs: packages for volumetric design, photo or video editors, the main parameter for choosing RAM is its bandwidth, that is, the speed of information processing.

In 2015, with the appearance on the market of Socket LGA1151 platforms, PC users had the opportunity to make a comparative analysis of the third and fourth generation RAM in the same conditions.

Specifications

Before saying that DDR3 or DDR4 is better and comparing them, you should familiarize yourself in detail with their technical characteristics and capabilities, as well as their advantages and disadvantages. This approach will allow you to correctly and accurately determine the future of memory modules and identify a promising sample.

DDR3

The main characteristics for random access memory, regardless of its generation, are the following characteristics:

Frequency. RAM of the third model is available with frequencies of 1066 MHz, 1333 MHz and 1600 MHz, and the latest modification has 1866 MHz. By overclocking the memory, its frequency can be increased to 2400 - 2666 MHz. The maximum value of this parameter during overclocking, which was obtained in laboratory conditions, is 4620 MHz.
Voltage. Power consumption ranges from 1.5 to 1.8 V. The latest version of DDR3L is capable of operating at low voltages of 1.25 to 1.35 V. The L index stands for Low Power.
Downtime. To determine the performance of a memory strip, one of the important parameters is timings or latency (CL), that is, the delay in transferring information. DDR3 1600 MHz has a latency of 9 clock cycles, it takes 1 second to get the time value. divided by 1600 million clock cycles and we get 0.625 ms per 1 clock cycle. The result is multiplied by 9 clock cycles and we get 5.625 ns. Then we multiply by 2 (the number of data streams) and the delay time is 11.25 ns.

Advice. The value of the latency can be determined from the marking of the RAM after the letters CL. Accordingly, the lower the value, the higher the performance of the device.

DDR4

4th generation RAM has higher parameters technical characteristics, due to which it bypasses its predecessor.

Comparison of DDR3 and DDR4

Based on the technical specifications, it can be seen that the latency of DDR4 is higher than that of its predecessor. However, when reading data linearly or storing them due to practically unchanging timings, this difference is compensated for, and the RAM of the fourth model wins. When working in multi-threaded mode, due to lower latency, DDR3 wins within the statistical error. When compressing large files (1.5 GB and more), DDR4 has 3% less time spent on operation than DDR3. The third generation RAM specification calls for the use of Vddr voltage. When performing energy-consuming operations, it increases due to built-in converters, thereby there is an abundant radiation of heat. The DDR4 module receives the required voltage from an external power supply (Vpp).

The Pseudo-Open Draid technology is implemented in the RAM of the fourth model, it allowed to completely eliminate the current leakage, which was observed in the previous version, which uses the Series-Stub Terminated Logic. The use of this interface for data input and output has reduced energy consumption by up to 30%. As for the memory capacity of the DDR4 bracket, the minimum value is 4 GB, and for DDR3 it is optimal, since the maximum is 8 GB. The structure of the third generation random access memory allows to accommodate up to 8 memory banks with a line length of 2048 bytes. The latest modification of RAM has 16 banks and a line length of 512 bytes, which increases the speed of switching between lines and banks.

From a comparison of DDR3 and DDR4, we can conclude that the latest generation of RAM bypasses its predecessor in almost all respects, but this difference is hardly noticeable to the average user. DDR3L 1600 MHz in combination with Intel Core i5 is almost as good as DDR4. It is recommended to install fourth-generation RAM for modern games or work in specialized programs that require a large amount of memory and high data processing speed.

Comparison of DDR 3 and DDR 4 RAM: video

So the Intel Haswell-E processors came out. the site has already tested the top-end 8-core Core i7-5960X, as well as the ASUS X99-DELUXE motherboard. And, perhaps, the main feature of the new platform is support for the DDR4 RAM standard.

The beginning of a new era, the DDR4 era

About SDRAM standard and memory modules

The first SDRAM modules appeared back in 1993. They were released by Samsung. And by 2000, SDRAM, due to the production facilities of the Korean giant, completely replaced the DRAM standard from the market.

SDRAM stands for Synchronous Dynamic Random Access Memory. Literally it can be translated as "synchronous dynamic random access memory". Let us explain the meaning of each characteristic. Dynamic memory is because, due to the small capacity of the capacitors, it constantly requires updating. By the way, in addition to dynamic memory, there is also static memory, which does not require constant data update (SRAM). SRAM, for example, is at the heart of cache memory. Besides dynamic memory, memory is also synchronous, as opposed to asynchronous DRAM. Synchronicity means that the memory performs each operation for a known number of times (or ticks). For example, when requesting any data, the memory controller knows exactly how long it will take to get to it. The synchronicity property allows data flow to be controlled and queued. Well, a few words about "random access memory" (RAM). This means that at a time you can access any cell by its address for reading or writing, and always for the same time, regardless of location.

SDRAM memory module

If we talk directly about the construction of memory, then its cells are capacitors. If there is a charge in the capacitor, then the processor regards it as a logical unit. If there is no charge - as a logical zero. Such memory cells have a flat structure, and the address of each of them is defined as the row and column numbers of the table.

Each chip contains several independent memory arrays, which are tables. They are called banks. You can work with only one cell in a bank at a time, but there is a possibility of working with several banks at once. The recorded information does not have to be stored in one array. Often, it is split into several parts and written to different banks, and the processor continues to read this data as a whole. This recording method is called interleaving. In theory, the more such banks there are in the memory, the better. In practice, modules with a density of up to 64 Mbit have two banks. With a density from 64 Mbps to 1 Gbps - four, and with a density of 1 Gbps and higher - already eight.

What is a memory bank

And a few words about the structure of the memory module. The memory module itself is a printed circuit board with chips soldered on it. As a rule, on sale you can find devices made in DIMM (Dual In-line Memory Module) or SO-DIMM (Small Outline Dual In-line Memory Module) form factors. The first is intended for use in full-fledged desktop computers, and the second is intended for installation in laptops. Despite the same form factor, memory modules of different generations differ in the number of contacts. For example, the SDRAM solution has 144 pins for connecting to the motherboard, DDR - 184, DDR2 - 214 pins, DDR3 - 240, and DDR4 - already 288 pieces. Of course, in this case we are talking about DIMMs. Devices made in the SO-DIMM form factor, of course, have fewer contacts due to their smaller size. For example, a DDR4 SO-DIMM memory module is connected to the motherboard through 256 pins.

DDR (bottom) has more pins than SDRAM (top)

It is also quite obvious that the volume of each memory module is calculated as the sum of the capacities of each unsoldered chip. Memory chips, of course, can differ in their density (or, more simply, in volume). For example, last spring Samsung launched a mass production of chips with a density of 4 Gbps. Moreover, in the foreseeable future, it is planned to release memory with a density of 8 Gbps. Also, memory modules have their own bus. The minimum bus width is 64 bits. This means that 8 bytes of information are transferred per cycle. It should be noted that there are also 72-bit memory modules in which the "extra" 8 bits are reserved for the ECC (Error Checking & Correction) technology. By the way, the bus width of a memory module is also the sum of the bus widths of each individual memory chip. That is, if the memory bus is 64-bit and eight chips are soldered on the strip, then the memory bus width of each chip is 64/8 = 8 bits.

To calculate the theoretical bandwidth of a memory module, you can use the following formula: A * 64/8 = PS, where “A” is the data transfer rate and “PS” is the desired bandwidth. As an example, we can take a DDR3 2400 MHz memory module. In this case, the throughput will be equal to 2400 * 64/8 = 19200 MB / s. This number is meant in the marking of the PC3-19200 module.

How does one read information from memory directly? First, the address signal is sent to the corresponding row (Row), and only then information is read from the required column (Column). The information is read into the so-called amplifier (Sense Amplifiers) - a mechanism for recharging capacitors. In most cases, the memory controller reads at once a whole data packet (Burst) from each bit of the bus. Accordingly, when writing, every 64 bits (8 bytes) are divided into several parts. By the way, there is such a thing as Burst Length. If this length is 8, then 8 * 64 = 512 bits are transmitted at once.

Memory modules and chips also have such characteristics as geometry, or organization (Memory Organization). The geometry of a module shows its width and depth. For example, a chip with a density of 512 Mbit and a width (width) of 4 has a chip depth of 512/4 = 128M. In turn, 128M = 32M * 4 banks. 32M is a matrix with 16000 rows and 2000 columns. It can store 32 Mbps of data. As for the memory module itself, it is almost always 64-bit. The depth is easily calculated using the following formula: the volume of the module is multiplied by 8 to convert from bytes to bits, and then divided by the bit width.

Timing values can be easily found on the marking

It is necessary to say a few words about such characteristics of memory modules as timings (delays). At the very beginning of the article, we talked about the fact that the SDRAM standard provides for such a moment that the memory controller always knows how long this or that operation is being performed. Timings indicate the time required to execute a specific command. This time is measured in memory bus clock cycles. The shorter the time, the better. The most important are the following delays:

TRCD (RAS to CAS Delay) - the time it takes to activate the bank line. The minimum time between the activation command and the read / write command;
CL (CAS Latency) - the time between the issuance of the read command and the start of data transfer;
TRAS (Active to Precharge) - line active time. The minimum time between the activation of the line and the command to close the line;
TRP (Row Precharge) - time required to close a row;
TRC (Row Cycle time, Activate to Activate / Refresh time) - time between activation of rows of the same bank;
TRPD (Active bank A to Active bank B) - the time between activation commands for different banks;
TWR (Write Recovery time) - the time between the end of the recording and the issuance of the command to close the bank line;
TWTR (Internal Write to Read Command Delay) - the time between the end of writing and the read command.

Of course, these are not all delays existing in memory modules. You can list a dozen more different timings, but only the above parameters significantly affect memory performance. By the way, only four delays are indicated in the labeling of memory modules. For example, with parameters 11-13-13-31, CL is 11, TRCD and TRP are 13, and TRAS is 31 measures.

Over time, the potential of SDRAM reached its ceiling, and manufacturers were faced with the problem of increasing the speed of RAM. This is how the DDR standard was born.1

DDR coming

The development of the DDR (Double Data Rate) standard began back in 1996 and ended with an official presentation in June 2000. With the advent of DDR, the outdated SDRAM was referred to simply as SDR. How is DDR different from SDR?

After all SDR resources were depleted, memory manufacturers had several ways to solve the problem of improving performance. One could simply increase the number of memory chips, thereby increasing the capacity of the entire module. However, this would negatively affect the cost of such solutions - this venture was very expensive. Therefore, the JEDEC manufacturers' association took a different path. It was decided to double the bus inside the chip, and also to transfer data at twice the frequency. In addition, DDR provided for the transfer of information on both edges of the clock signal, that is, twice per clock. This is where the abbreviation DDR - Double Data Rate originates.

Kingston DDR Memory

With the advent of the DDR standard, such concepts as the real and effective memory frequency appeared. For example, many DDR memory modules ran at 200 MHz. This frequency is called real. But due to the fact that data transmission was carried out on both edges of the clock signal, manufacturers for marketing purposes multiplied this figure by 2 and allegedly received an effective frequency of 400 MHz, which was indicated in the marking (in this case, DDR-400). At the same time, the JEDEC specifications indicate that it is completely incorrect to use the term "megahertz" to characterize the level of memory performance! Instead, "Millions of transmissions per second over one data output" should be used. However, marketing is a serious matter, and few people were interested in the recommendations specified in the JEDEC standard. Therefore, the new term never caught on.

Also in the DDR standard a dual-channel memory mode appeared for the first time. It could be used if there was an even number of memory modules in the system. Its essence is to create a virtual 128-bit bus by interleaving modules. In this case, 256 bits were sampled at once. On paper, dual-channel mode can double the performance of the memory subsystem, but in practice the speed gain is minimal and not always noticeable. It depends not only on the RAM model, but also on timings, chipset, memory controller and frequency.

Four memory modules operate in dual channel mode

Another innovation in DDR is the presence of the QDS signal. It is located on the printed circuit board along with the data lines. QDS has been useful when using two or more memory modules. In this case, the data arrives at the memory controller with a small time difference due to the different distance to them. This creates problems when choosing a clock signal for reading data, which QDS successfully solves.

As mentioned above, DDR memory modules were implemented in DIMM and SO-DIMM form factors. In the case of DIMMs, the number of pins was 184. In order for DDR and SDRAM modules to be physically incompatible, for DDR solutions the key (cut in the area of the contact pad) was located in a different place. In addition, DDR memory modules operated at 2.5 V, while SDRAM devices operated at 3.3 V. Accordingly, DDR had lower power consumption and heat dissipation compared to its predecessor. The maximum frequency of DDR modules was 350 MHz (DDR-700), although the JEDEC specifications provided only 200 MHz (DDR-400).

DDR2 and DDR3 Memory

The first DDR2 modules went on sale in the second quarter of 2003. Compared to DDR, RAM the second generation did not receive significant changes. DDR2 used all the same 2 n -prefetch architecture. Whereas the internal data bus used to be twice as large as the external data bus, it is now four times wider. At the same time, the increased performance of the chip began to be transmitted over the external bus at a doubled frequency. Specifically the frequency, but not twice the transmission rate. As a result, we got that if the DDR-400 chip worked at a real frequency of 200 MHz, then in the case of DDR2-400 it operated at a speed of 100 MHz, but with twice the internal bus.

Also, DDR2 modules received a larger number of pins for attaching to the motherboard, and the key was moved to another place for physical incompatibility with SDRAM and DDR strips. The operating voltage has been reduced again. If the DDR modules worked at 2.5 V, then the DDR2 solutions operated at a potential difference of 1.8 V.

By and large, this is where all the differences between DDR2 and DDR end. At first, DDR2 modules in the negative direction were characterized by high latencies, which is why they were inferior in performance to DDR strips with the same frequency. However, the situation soon returned to square one: manufacturers reduced latencies and released faster sets of RAM. The maximum DDR2 frequency reached the effective 1300 MHz mark.

Different key positions for DDR, DDR2 and DDR3 modules

Moving from DDR2 to DDR3 followed the same approach as moving from DDR to DDR2. Of course, the data transfer on both ends of the clock signal was preserved, and the theoretical throughput has doubled. DDR3 modules retained the 2 n -prefetch architecture and got an 8-bit prefetch (in DDR2 it was 4-bit). At the same time, the internal bus became eight times larger than the external one. Because of this, once again, when changing generations of memory, its timings increased. The nominal operating voltage for DDR3 has been reduced to 1.5V, which makes the modules more energy efficient. Note that, in addition to DDR3, there is DDR3L memory (the letter L stands for Low), which operates with a voltage reduced to 1.35 V. It is also worth noting that DDR3 modules turned out to be neither physically nor electrically compatible with any of the previous generations of memory.

Of course, DDR3 chips received support for some new technologies: for example, automatic signal calibration and dynamic signal termination. However, in general, all changes are predominantly quantitative.

DDR4 is another evolution

Finally, we got to the brand new DDR4 memory. The JEDEC Association began developing the standard back in 2005, but it was not until the spring of this year that the first devices appeared on the market. During the development, the engineers tried to achieve the highest performance and reliability, while increasing the energy efficiency of the new modules, according to a JEDEC press release. Well, we hear that every time. Let's take a look at the specific changes made to DDR4 versus DDR3.

In this picture, you can trace the evolution of DDR technology: how the indicators of voltage, frequency and capacity have changed.

One of the first DDR4 prototypes. Oddly enough, these are laptop modules

As an example, consider an 8GB DDR4 chip with a 4-bit wide data bus. Such a device contains 4 groups of banks with 4 banks in each. Inside each bank there are 131 072 (2 17) lines with a capacity of 512 bytes each. For comparison, we can give the characteristics of a similar DDR3 solution. This chip contains 8 independent banks. Each bank contains 65,536 (2 16) lines, and each line contains 2048 bytes. As you can see, the length of each line of the DDR4 chip is four times less than the length of the DDR3 line. This means that DDR4 will scan banks faster than DDR3. At the same time, switching between the banks themselves is also much faster. Immediately, we note that for each group of banks, an independent selection of operations (activation, reading, writing or regeneration) is provided, which makes it possible to increase the efficiency and memory bandwidth.

The main advantages of DDR4: low power consumption, high frequency, large amount of memory modules

The JEDEC Solid State Technology Association has unveiled the official final version of the Synchronous DDR4 ( Double Data Rate 4).

Its introduction is to provide a new level of RAM performance, its reliability and power consumption reduction.

DDR4 memory incorporates a host of modern advances that will allow this new type of memory to gain widespread acceptance in computing devices, from home appliances to servers and even more powerful computing systems.

DDR4 has set the performance level per slot at 1.6 billion hops per second, with the potential to peak at 3.2 billion / s in the future.
The minimum operating frequency of DDR4 memory is 2133 MHz to 4266 MHz, which is 1000 MHz higher than its predecessor (1333 MHz and 1666 MHz in the previous generation standard).
For memory with a frequency of 2133 MHz (the lowest frequency for DDR4 memory), the maximum bandwidth will be 2133 * 8 = 17,064 MB / s.
For memory with a frequency of 4266 MHz (the highest frequency defined in the standard), the maximum bandwidth will be 4266 * 8 = 34 128 MB / s.
Operating voltage lowered: 1.1V - 1.2V versus 1.5V in DDR3.
Assumed process technology - 32 and 36 nm.

The DDR4 architecture allows 8n prefetching of data per clock cycle (8n prefetch) with two or four selectable groups of memory blocks. This allows devices to conduct independent activation, read, write and update operations through separate memory blocks.

All of the above features, as well as a number of smaller changes and innovations, have significantly improved memory efficiency. DDR4.

DDR4 module has 284 connections, while standard DDR3 modules only have 240 contacts.

V SO-DIMM versions will be presented 256 contacts while DDR3 SO-DIMMs have only 204 pins.

In the DDR4 specifications, for the first time, there was a description of working with memory in multichip packaging. The standard allows a column (stack) of eight crystals. Moreover, all crystals are "hung" on common signal lines. This was done not because it is better this way (although it really simplifies the steps to expand the memory space), but for the reasons that, in general, the ideology of DDR4 memory operation is a point-to-point connection of modules with controllers.

There will be many channels, not two or four, so each of them needs to provide the highest possible performance without overloading the exchange mechanisms. In the same vein, the possibility of independent simultaneous work two or four memory banks. For each group of banks, all basic operations, such as reading, writing and regenerating, are architecturally allowed at the same time.

According to iSuppli's forecast, by 2014 the level of penetration into the DDR4 memory market will be 12%, by 2015 - 56%. However, manufacturers may be quick to start implementing the new standard, prompted by the desire to raise prices for their products, which are now at extremely low levels. Micron, for example, announced the development of the first fully functional module back in May and plans to begin mass production at the end of this year. Samsung has already demonstrated 284-pin PC4-17000 memory (2133 MHz). We just have to wait for their support from Intel and AMD. Intel plans to start supporting DDR4 in early 2014 in high-performance 4-socket server systems based on Haswell-EX processors, while regular users will probably have to wait until 2015, since neither Haswell's 22nm processors nor the following 14nm Broadwell processors support DDR4 is not available.

DDR4 is just one of the first steps towards widespread adoption of next generation memory. Applications for DDR4 memory include servers, laptops, desktops and consumer electronics. Initially, DDR4 will appear in server systems and after that it will start mass production of such memory for consumer computers.