If the memory timings do not match. Destroying myths about RAM

Greetings, dear readers! Today we will understand what timings in RAM mean and what this parameter affects. Indeed, suddenly, under this buzzword, they are trying to sell us another dummy - for example, how are megapixels in a cell phone camera without sane optics?

From this article you will learn:

A little materiel

To understand the timings - what they are and what they are for, you should go a little deeper into the mechanism of the RAM. A simplified diagram looks like this: RAM cells are arranged according to the principle of two-dimensional matrices, which are accessed by specifying a column and a row.

Memory cells are essentially capacitors that can be charged or discharged, thus writing one or zero (I think everyone has long been aware that any computing device works with binary code).

By changing the voltage from high to low, a row access (RAS) or column (CAS) pulse is sent. Clock-synchronized signals are first applied to the row, then to the column. When information is written, an additional tolerance pulse (WE) is applied. Memory performance directly depends on the amount of data transferred per clock cycle.

At the same time, there is one BUT: data is not transmitted instantly, but with some delay, which is also called latency. And as you know, nothing is transmitted instantly - even photons of light have a finite speed. What to say about the electrons trying to break through the layers of silicon?

What do timings mean

So, timing or latency is the amount of delay from receipt to execution of the command. There are several dozen types of them, as well as all sorts of sub-timings, but from a practical point of view, they are only of interest to engineers and other great hardware specialists.
For the average user, four types of timing are important, which are usually indicated when marking the RAM:

tRCD is the delay between RAS and CAS pulses;
tCL is the delay from issuing a read or write command to the CAS pulse;
tRP is the delay from processing a row to moving to the next one;
tRAS is the delay between row activation and start of processing.

Some manufacturers also indicate the Command rate - the delay between the selection of a specific chip on the memory module and the activation of the line.

Marking

The measure of timing is the memory bus cycle. In fact, these figures allow you to generally evaluate the performance of a RAM stick even before you buy it.

Usually timings are indicated on the nameplate along with the type of memory, frequency and other characteristics. For convenience, they are written as a set of numbers separated by a hyphen in the following order: tRCD- tCL- tRP- tRAS. For example, like this: 7-7-7-18.

However, not all manufacturers provide this information, so there is a possibility that by disassembling the computer and pulling out the memory module, you will not find the required data. How to find out the parameters of interest? In this case, programs will come to the rescue that allow you to get complete information about the hardware - for example, Speccy or CPU-Z.

And note that in the descriptions of goods in online stores, information about timings is often not provided.

Therefore, if you decide to get confused on hard and pick up an additional RAM bar with absolutely identical timings in order to activate the dual-channel RAM mode (why do you need it,), you will most likely have to go to a computer store and fool the seller (or find the information on the label yourself ).

Setting timings

Each stick of RAM is equipped with an SPD chip, which stores information about the recommended timings in relation to the system bus frequencies. Usually, with automatic settings, the computer sets the optimal latency value, thanks to which the RAM will show the best performance.

You can change the timings in BIOS. This is one of the favorite pastimes of overclockers and other computer sorcerers, who, with the help of all sorts of tricky settings, can significantly increase the performance of any hardware. If you don’t know what timings to set, it’s better not to touch anything, choosing automatic settings.

Naturally, when buying RAM, many are interested in the question of what will happen if different memory modules have different timings. In fact, nothing terrible will happen - you simply cannot run the RAM in dual-channel mode.

There are known cases of complete incompatibility of memory modules, the sharing of which provokes the appearance of the “blue screen of death”, but here, in addition to latency, many more additional parameters should be taken into account.

Going for a new memory bar, you may continue to wonder which timings are better. Naturally, those below. However, the difference in latency figures is reflected in the difference in the numbers on the price tag - other things being equal, a module with lower timings will cost more.

And if you have read my previous publications, then you probably still remember that I am indignant about the fossilized DDR3 in every way and I agitate everyone to focus on the progressive DDR4 standard when assembling a computer.

It will also be useful for you to read articles on this topic and how the processor frequency and RAM frequency correlate. For a deep dive, so to speak. To know everything.

With that, dear friends, I say to you “Until tomorrow”. Thank you for reading and sharing this post on social media.

In this study, we will try to find an answer to the following question - what is more important for achieving maximum computer performance, a high frequency of RAM or its low timings. And two sets of RAM produced by Super Talent will help us with this. Let's see how the memory modules look externally, and what characteristics they have.

⇡Super Talent X58

The manufacturer dedicated this kit to the Intel X58 platform, as evidenced by the inscription on the sticker. However, several questions immediately arise here. As everyone is well aware, in order to achieve maximum performance on the Intel X58 platform, it is highly recommended to use the three-channel mode of RAM. Despite this, this Super Talent memory kit consists of only two modules. Of course, for orthodox system builders, this approach may cause bewilderment, but there is still a rational grain in this. The fact is that the segment of top platforms is relatively small, and most personal computers use RAM in dual-channel mode. In this regard, the purchase of a set of three memory modules may seem unjustified to the average user, and if you really need a lot of RAM, you can purchase three sets of two modules each. The manufacturer indicates that the Super Talent WA1600UB2G6 memory can operate at 1600 MHz DDR with 6-7-6-18 timings. Now let's see what information is stored in the SPD profile of these modules.

And again there is some discrepancy between the real and declared characteristics. The maximum JEDEC profile assumes the operation of modules at a frequency of 1333 MHz DDR with timings of 9-9-9-24. However, there is an extended XMP profile, the frequency of which coincides with the declared one - 800 MHz (1600 MHz DDR), but the timings are somewhat different, and for the worse - 6-8-6-20, instead of 6-7-6-18, which are indicated on the sticker. Nevertheless, this set of RAM worked without problems in the declared mode - 1600 MHz DDR with timings of 6-7-6-18 and a voltage of 1.65 V. As for overclocking, higher frequencies were not obeyed by the modules, despite the installation of increased timings and increase the supply voltage. Moreover, when the voltage Vmem was increased to the level of 1.9 V, instability was observed in the initial mode as well. Unfortunately, the heatsinks are very firmly glued to the memory chips, so we did not dare to remove them for fear of damaging the memory modules. It's a pity, the type of chips used could shed light on this behavior of the modules.

⇡Super Talent P55

The second set of RAM, which we will consider today, is positioned by the manufacturer as a solution for the Intel P55 platform. The modules are equipped with low-profile black heatsinks. The maximum declared mode assumes the operation of these modules at a frequency of 2000 MHz DDR with timings of 9-9-9-24 and a voltage of 1.65 V. Now let's look at the profiles wired into the SPD.

The most productive JEDEC profile assumes the operation of modules at a frequency of 800 MHz (1600 MHz DDR) with timings of 9-9-9-24 and a voltage of 1.5 V, and there are no XMP profiles in this case. As for overclocking, with a slight increase in timings, these memory modules were able to operate at a frequency of 2400 MHz DDR, as evidenced by the screenshot below.

Moreover, the system booted even at 2600 MHz DDR, but the launch of test applications led to a hang or reboot. As in the case with the previous Super Talent memory kit, these modules did not react in any way to an increase in the supply voltage. As it turned out, the better overclocking of the memory and the stability of the system was more facilitated by an increase in the voltage of the memory controller built into the processor. However, the search for the maximum possible frequencies and parameters at which stability is achieved in such extreme modes, we leave to enthusiasts. Next, we will focus on studying the next question - to what extent the frequency of RAM and its timings affect the overall performance of the computer. In particular, we will try to figure out what is better - to install high-speed RAM that works with high timings, or it is preferable to use the lowest possible timings, even if not at maximum operating frequencies.

⇡ Test conditions

Testing was carried out on a stand with the following configuration. In all tests, the processor was running at 3.2 GHz, the reasons for this will be explained below, and a powerful graphics card was required for tests in the game Crysis.

As mentioned above, we will try to find out how the frequency of RAM and its timings affect the overall performance of the computer. Of course, these parameters can simply be set in the BIOS and tested. But, as it turned out, with a Bclk frequency of 133 MHz, the operating frequency range of the RAM in the motherboard we used is 800 - 1600 MHz DDR. This is not enough, because one of the Super Talent memory kits reviewed today supports DDR3-2000 mode. And in general, more and more high-speed memory modules are being produced, manufacturers assure us of their unprecedented performance, so it will definitely not hurt to find out their real performance. In order to set the memory frequency to, say, 2000 MHz DDR, it is necessary to increase the frequency of the Bclk bus. However, this will change the frequencies of both the processor core and its third-level cache, which operates at the same frequency as the QPI bus. Of course, it is incorrect to compare the results obtained under such different conditions. In addition, the degree of influence of the CPU frequency on the test results may turn out to be much more significant than the timings and frequency of the RAM. The question arises - is it possible to somehow get around this problem? As for the processor frequency, within certain limits it can be changed using a multiplier. However, it is desirable to choose such a value for the bclk frequency so that the final frequency of the RAM is equal to one of the standard values 1333, 1600 or 2000. As you know, at present the base frequency of bclk in Intel Nehalem processors is 133.3 MHz. Let's see what the frequency of the RAM will be at different values of the bclk bus frequency, taking into account the multipliers that the motherboard we use can set. The results are shown in the table below.

	Frequency bclk, MHz
	133.(3)	150	166.(6)	183.(3)	200
Memory multiplier	RAM frequency, MHz DDR
6	800	900	1000	1100	1200
8	1066	1200	1333	1466	1600
10	1333	1500	1667	1833	2000
12	1600	1800	2000	2200	2400

As can be seen from the table, with a bclk frequency of 166 MHz, frequencies of 1333 and 2000 MHz can be obtained for RAM. If the bclk frequency is 200 MHz, then we get the coincidence of the RAM frequencies at 1600 MHz, as well as the required 2000 MHz. In other cases, there are no coincidences with the standard memory frequencies. So which bclk frequency do you prefer in the end - 166 or 200 MHz? The following table will help answer this question. Here are the CPU frequency values, depending on the multiplier and the bclk frequency. To evaluate the impact of timings, we need not only the same memory frequencies, but also the CPU, so that this does not affect the results.

	Frequency bclk, MHz
CPU multiplier	133.(3)	150.0	166.(6)	183.(3)	200.0
9	1200	1350	1500	1647	1800
10	1333	1500	1667	1830	2000
11	1467	1650	1833	2013	2200
12	1600	1800	2000	2196	2400
13	1733	1950	2167	2379	2600
14	1867	2100	2333	2562	2800
15	2000	2250	2500	2745	3000
16	2133	2400	2667	2928	3200
17	2267	2550	2833	3111	3400
18	2400	2700	3000	3294	3600
19	2533	2850	3167	3477	3800
20	2667	3000	3333	3660	4000
21	2800	3150	3500	3843	4200
22	2933	3300	3667	4026	4400
23	3067	3450	3833	4209	4600
24	3200	3600	4000	4392	4800

As a starting point, we took the maximum processor frequency (3200 MHz) that it can show with a bclk base frequency of 133 MHz. It can be seen from the table that under these conditions, only with a frequency of bclk=200 MHz, exactly the same CPU frequency can be obtained. The remaining frequencies, although close to 3200 MHz, are not exactly equal to it. Of course, the CPU frequency could be taken as the initial one, and even lower, say - 2000 MHz, then it would be possible to get correct results with all three values of the bclk bus - 133, 166 and 200 MHz. However, we have abandoned this option. And that's why. Firstly, there are no Intel desktop processors with the Nehalem architecture with such a frequency, and they are unlikely to appear. Secondly, reducing the CPU frequency by more than 1.5 times can lead to the fact that it becomes a limiting factor, and the difference in results will practically not depend on the mode of operation of the RAM. Actually, the first estimates showed exactly this. Thirdly, it is unlikely that the user who buys a deliberately weak and cheap processor will be very concerned about choosing an expensive high-speed RAM. So, we will test with base frequency bclk - 133 and 200 MHz. The CPU frequency in both cases is the same and equals 3200 MHz. Below are screenshots of the CPU-Z utility in these modes.

If you paid attention, the QPI-Link frequency depends on the bclk frequency and, accordingly, they differ by 1.5 times. By the way, this will allow us to find out how the L3 cache frequency in Nehalem processors affects the overall performance. So, let's start testing.

The A-Data memory module with a clock frequency of DDR3-1333 sets the timings to 9-9-9-24, when the operating frequency is lowered to DDR3-1066, the timings are reduced to only 8-8-8-20.

Memory Bandwidth

Bandwidth- memory characteristic, on which performance depends and on which it is expressed as the product of the system bus frequency and the amount of data transferred per clock. However, the frequency of the memory module and the theoretical bandwidth are not the only parameters that are responsible for system performance. Memory timings also play an important role.

Bandwidth (Peak Data Rate)- This is a comprehensive indicator of the capabilities of RAM, it takes into account the data transfer frequency, bus width and the number of memory channels. The frequency indicates the potential of the memory bus per clock - at a higher frequency, more data can be transferred.

The peak indicator is calculated by the formula:

Bandwidth (B) = Baud rate (f) x bus width (c) x number of memory channels (k)

If we consider the example of DDR400 (400 MHz) with a dual-channel memory controller, the peak data transfer rate is:
(400 MHz x 64 bits x 2)/ 8 bits = 6400 MB/s

We divided by 8 to convert Mbps to Mbps (8 bits in 1 byte).

Bandwidth

For fast computer operation, the bandwidth of the RAM bus must match the bandwidth of the processor bus. For example, for a processor Intel core 2 duo E6850 with a 1333 MHz system bus and a bandwidth of 10600 Mb / s, you need to buy two RAM with a bandwidth of 5300 Mb / s each (PC2-5300), in total they will have a system bus bandwidth (FSB) equal to 10600 Mb / s.

At high data processing speeds, there is one minus - high heat generation. To do this, manufacturers have reduced the supply voltage of DDR3 memory to 1.5 V.

Dual channel mode

To increase data transfer speed and increase bandwidth, modern chipsets support dual-channel memory architecture.

If you install two absolutely identical memory modules, then dual-channel mode will be used. Best to use Kit- a set of two or more memory modules that have already been tested when working with each other. These memory modules are from the same manufacturer, with the same size and the same frequency.

When using two identical DDR3 memory modules in dual-channel mode, it can increase the bandwidth up to 17.0 GB / s. If you use RAM with 1333 MHz, then the bandwidth will increase to 21.2 GB / s.

Memory Timings

Timings, latency, CAS Latency, CL. Quite often, these parameters are not indicated in the product description, and in fact they characterize the speed of RAM. The smaller the value, the faster the RAM works. Try to choose RAM with the lowest timing and preferably from memory modules with the same amount of memory and operating clock speed. However, for example, memory modules with a clock frequency of DDR-800, 5-5-5-18 and DDR3-1066, 7-7-7-20 can be considered equivalent in terms of performance.

Timings

Timings- signal time delays. Timings are measured in nanoseconds (ns). The measure of timings is tact. In the description of RAM, they are indicated as a sequence of numbers (CL5-5-4-12 or simply 9-9-9-24), where the following parameters are indicated in order:

CAS Latency– the delay between the read command and the readability of the first word.

RAS to CAS Delay (RCD)- delay between RAS (Row Address Strobe) and CAS (Column Address Strobe) signals, this parameter specifies the interval between accesses to the bus by the RAS# and CAS# signal memory controller.

RAS Precharge Time (RP)– re-issuance time (charge accumulation period) of the RAS# signal – after what time the memory controller will be able to issue the line address initialization signal again.

DRAM Cycle Time Tras/Trc– total performance indicator of the memory module

If only one CL8 parameter is specified in the description, then it means only the first parameter - CAS Latency.

Many motherboards, when installing memory modules on them, do not set the maximum clock speed for them. One of the reasons is the lack of performance gain with an increase in clock frequency, because with an increase in frequency, operating timings increase. Of course, this may improve performance in some applications, but it may decrease in others, or it may not affect applications at all that do not depend on memory latency or bandwidth.

For example. The Corsair memory module installed on the M4A79 Deluxe motherboard will have the following timings: 5-5-5-18. If you increase the memory clock speed to DDR2-1066, the timings will increase and will have the following values 5-7-7-24.

The Qimonda memory module, when operating at a clock frequency of DDR3-1066, has working timings of 7-7-7-20, when the operating frequency increases to DDR3-1333, the board sets the timings of 9-9-9-25. As a rule, timings are written in SPD and may differ for different modules.

The main characteristics of RAM (its volume, frequency, belonging to one of the generations) can be supplemented by another important parameter - timings. What are they? Can they be changed in the BIOS settings? How to do it in the most correct way, from the point of view of stable computer operation?

What are RAM timings?

The RAM timing is the time interval during which the command sent by the RAM controller is executed. This unit is measured in the number of cycles that are skipped by the computing bus while the signal is being processed. The essence of the timings is easier to understand if you understand the design of RAM chips.

The RAM of a computer consists of a large number of interacting cells. Each has its own conditional address, at which the RAM controller accesses it. Cell coordinates are usually specified using two parameters. Conventionally, they can be represented as numbers of rows and columns (as in a table). In turn, groups of addresses are combined to make it "more convenient" for the controller to find a specific cell in a larger data area (sometimes called a "bank").

Thus, the request for memory resources is carried out in two stages. First, the controller sends a request to the "bank". It then asks for the "row" number of the cell (by sending a signal like RAS) and waits for a response. The waiting time is the RAM timing. Its common name is RAS to CAS Delay. But that is not all.

The controller, in order to refer to a specific cell, also needs the number of the "column" assigned to it: another signal is sent, such as CAS. The time while the controller is waiting for a response is also the timing of the RAM. It's called CAS Latency. And that is not all. Some IT professionals prefer to interpret the phenomenon of CAS Latency in a slightly different way. They believe that this parameter indicates how many single cycles should pass in the process of processing signals not from the controller, but from the processor. But, according to experts, in both cases, in principle, we are talking about the same thing.

The controller, as a rule, works with the same "line" on which the cell is located, more than once. However, before calling it again, it must close the previous request session. And only after that to resume work. The time interval between completion and a new call to the line is also timing. It's called RAS Precharge. Already the third in a row. That's all? No.

Having worked with the string, the controller must, as we recall, close the previous request session. The time interval between the activation of access to the line and its closing is also the timing of the RAM. Its name is Active to Precharge Delay. Basically, that's all.

Thus, we counted 4 timings. Accordingly, they are always written in the form of four digits, for example, 2-3-3-6. In addition to them, by the way, there is another common parameter that characterizes the computer's RAM. It's about the Command Rate value. It shows what is the minimum time the controller spends to switch from one command to another. That is, if the value for CAS Latency is 2, then the time delay between a request from the processor (controller) and the response of the memory module will be 4 cycles.

Timings: order of placement

What is the order in which each of the timings is located in this numerical series? It almost always (and this is a kind of industry "standard") is as follows: the first digit is CAS Latency, the second is RAS to CAS Delay, the third is RAS Precharge and the fourth is Active to Precharge Delay. As we said above, the Command Rate parameter is sometimes used, its value is the fifth in a row. But if for the four previous indicators the spread of numbers can be quite large, then for CR, as a rule, only two values are possible - T1 or T2. The first means that the time from the moment when the memory is activated until it is ready to respond to requests should be 1 cycle. According to the second - 2.

What are the timings talking about?

As you know, the amount of RAM is one of the key performance indicators of this module. The bigger it is, the better. Another important parameter is the frequency of the RAM. Here, too, everything is clear. The higher it is, the faster the RAM will work. What about timings?

For them, the rule is different. The smaller the values of each of the four timings, the better, the more productive the memory. And the faster, respectively, the computer works. If two modules with the same frequency have different RAM timings, then their performance will also differ. As we have already defined above, the values we need are expressed in cycles. The fewer of them, the faster the processor receives a response from the RAM module. And the sooner he can "take advantage" of such resources as the frequency of RAM and its volume.

"Factory" timings or your own?

Most PC users prefer to use those timings that are already set on the conveyor (or auto-tuning is set in the motherboard options). However, many modern computers have the ability to set the desired parameters manually. That is, if lower values are needed, they can usually be put down. But how to change RAM timings? And to do it in such a way that the system works stably? And perhaps there are cases in which it is better to choose increased values? How to set RAM timings optimally? Now we will try to answer these questions.

Setting up timings

The factory timings are written in a dedicated area of the RAM chip. It's called SPD. Using the data from it, the BIOS system adapts the RAM to the configuration of the motherboard. In many modern BIOS versions, the default timing settings can be adjusted. Almost always this is done programmatically - through the system interface. Changing the values of at least one timing is available in most motherboard models. There are, in turn, manufacturers who allow fine-tuning of RAM modules using a much larger number of parameters than the four types mentioned above.

To enter the area of the desired settings in the BIOS, you need to enter this system (the DEL key immediately after turning on the computer), select the Advanced Chipset Settings menu item. Next, among the settings, we find the line DRAM Timing Selectable (it may sound a little different, but similar). We note in it that the timings (SPD) will be set manually (Manual).

How to find out the default RAM timing set in the BIOS? To do this, we find in the neighboring settings parameters that are consonant with CAS Latency, RAS to CAS, RAS Precharge and Active To Precharge Delay. Specific timings, as a rule, depend on the type of memory modules installed on the PC.

By selecting the appropriate options, you can set the timings. Experts recommend lowering the numbers very gradually. You should, after selecting the desired indicators, reboot and test the system for stability. If the computer is malfunctioning, you need to return to the BIOS and set the values a few levels higher.

Timing optimization

So, RAM timings - what are the best values for them to set? Almost always, the optimal numbers are determined in the course of practical experiments. The operation of a PC is connected not only with the quality of the functioning of the RAM modules, and not only with the speed of data exchange between them and the processor. Many other characteristics of a PC are important (up to such nuances as a computer cooling system). Therefore, the practical effectiveness of changing timings depends on the specific hardware and software environment in which the user configures the RAM modules.

We have already named the general pattern: the lower the timings, the higher the speed of the PC. But this is of course the ideal scenario. In turn, timings with reduced values can be useful when "overclocking" motherboard modules - artificially increasing its frequency.

The fact is that if you give the RAM chips acceleration in manual mode, using too large coefficients, then the computer may start to work unstably. It is quite possible that the timing settings will be set so incorrectly that the PC will not be able to boot at all. Then, most likely, you will have to "reset" the BIOS settings using the hardware method (with a high probability of contacting a service center).

In turn, higher values for timings can, by slowing down the PC somewhat (but not so much that the operating speed was brought to the mode that preceded "overclocking"), give the system stability.

Some IT experts have calculated that RAM modules with a CL of 3 provide about 40% less latency in the exchange of corresponding signals than those where CL is 5. Of course, provided that the clock frequency on both the other is the same.

Additional timings

As we have already said, in some modern models of motherboards there are opportunities for very fine tuning of the RAM. This, of course, is not about how to increase the RAM - this parameter is, of course, the factory one, and cannot be changed. However, the RAM settings offered by some manufacturers have very interesting features, using which you can significantly speed up your PC. We will consider those that relate to timings that can be configured in addition to the four main ones. An important nuance: depending on the motherboard model and BIOS version, the names of each of the parameters may differ from those that we will now give in the examples.

1. RAS to RAS Delay

This timing is responsible for the delay between the moments when rows from different areas of consolidation of cell addresses ("banks" that is) are activated.

2. Row Cycle Time

This timing reflects the time interval during which one cycle lasts within a single line. That is, from the moment of its activation until the start of work with a new signal (with an intermediate phase in the form of closing).

3.Write Recovery Time

This timing reflects the time interval between two events - the completion of the cycle of writing data to the memory and the beginning of the electrical signal.

4. Write To Read Delay

This timing shows how much time should elapse between the completion of the write cycle and the moment when data reading begins.

In many BIOS versions, the Bank Interleave option is also available. By selecting it, you can configure the processor so that it accesses the same "banks" of RAM at the same time, and not in turn. By default, this mode operates automatically. However, you can try to set a parameter of type 2 Way or 4 Way. This will allow you to use 2 or 4, respectively, "banks" at the same time. Disabling the Bank Interleave mode is used quite rarely (this is usually associated with PC diagnostics).

Setting timings: the nuances

Let's name some features related to the operation of timings and their settings. According to some IT specialists, in a series of four digits, the first one is the most important, that is, the CAS Latency timing. Therefore, if the user has little experience in "overclocking" RAM modules, experiments should probably be limited to setting values only for the first timing. Although this point of view is not generally accepted. Many IT experts tend to think that the other three timings are no less important in terms of the speed of interaction between the RAM and the processor.

In some models of motherboards in the BIOS, you can adjust the performance of RAM chips in several basic modes. In fact, this is setting timing values according to templates that are acceptable from the point of view of stable PC operation. These options usually coexist with the Auto by SPD option, and the modes in question are Turbo and Ultra. The first implies a moderate acceleration, the second - the maximum. This feature can be an alternative to manually setting timings. Similar modes, by the way, are available in many interfaces of the advanced BIOS system - UEFI. In many cases, as experts say, when you turn on the Turbo and Ultra options, the PC performance is sufficiently high, and its operation is stable at the same time.

Clocks and nanoseconds

Is it possible to express clock cycles in seconds? Yes. And there is a very simple formula for this. Ticks in seconds are considered to be one divided by the actual RAM clock speed specified by the manufacturer (although this figure, as a rule, should be divided by 2).

That is, for example, if we want to know the clocks that form the timings of DDR3 or 2 RAM, then we look at its marking. If the number 800 is indicated there, then the actual RAM frequency will be 400 MHz. This means that the duration of the cycle will be the value obtained by dividing one by 400. That is, 2.5 nanoseconds.

Timings for DDR3 modules

Some of the most modern RAM modules are DDR3 chips. Some experts believe that such indicators as timings are of much lesser importance for them than for chips of previous generations - DDR 2 and earlier. The fact is that these modules, as a rule, interact with sufficiently powerful processors (such as, for example, Intel Core i7), whose resources allow you to access RAM less often. In many modern chips from Intel, as well as in similar solutions from AMD, there is a sufficient amount of their own analogue of RAM in the form of L2- and L3-cache. We can say that such processors have their own amount of RAM, capable of performing a significant amount of typical RAM functions.

Thus, working with timings when using DDR3 modules, as we found out, is not the most important aspect of "overclocking" (if we decide to speed up PC performance). Much more important for such microcircuits are just the same frequency parameters. At the same time, DDR2 RAM modules and even earlier technological lines are still installed on computers today (although, of course, the widespread use of DDR3, according to many experts, is more than a steady trend). And therefore, working with timings can be useful to a very large number of users.

(English) we went through the basic concepts and characteristics of RAM. In this article, we want to touch on topics that often cause controversy, and try to understand the following myths and statements:

All DDR3 memory is the same
Just need to add more RAM
There are only a few DIMM manufacturers
DDR-3200 support means any RAM can be used
When installing different modules, RAM runs at the speed (timings) of the slowest DIMM
Cheaper to buy two sets of DIMMs than one large and expensive set
RAM runs faster when all slots are occupied
RAM faster than 1600 MT / s does not give a performance boost
8 GB is enough for the next ten years
You will never be able to use 16 GB of memory
I'm not using all available memory, so extra memory won't speed things up.
64-bit OS allows you to use any amount of RAM
1.65V RAM can damage Intel processors
Dual-channel mode doubles the data transfer rate, i.e. the RAM works twice as fast

Myths about RAM | All DDR3 memory is the same

This topic alone deserves a separate article, but we will try to discuss it briefly and outline a few theses.

Consider the Kingston Fury line of RAM, which does not come equipped with the XMP profile and instead uses plug and play technology. The modules are reasonably priced, look nice, come with colorful heatsinks, and are aimed at older system users looking to upgrade their RAM. But since this memory is based on PnP, it will only work with certain chipsets: H67, P67, Z68, Z77, Z87 and H61 from Intel, along with AMD A75, A87, A88, A89, A78 and E35. You can also add Z87 and Z97 here. The list of chipsets is taken from the company's website.
The chips themselves are also different:

Most RAM manufactured today uses high-density 4Gb memory chips, while older DDR3 uses lower-density 2Gb chips. Older memory controllers can only handle low density chips. One of our editors recently discovered that none of the P55 motherboards wanted to work with his 8 GB modules. And if you install memory with different characteristics, then the module may not be determined or lose stability.
Memory chips are produced by many companies that adhere to their own specifications. Each line of chips is tested or binned, and according to the quality of the chip, it is marked and allocated to different series.

Most enthusiast motherboards are designed to support unbuffered memory without the use of error correcting code (ECC). ECC is typically used in servers and professional workstations where data integrity is critical, and buffered (registered) DIMMs are used exclusively in servers requiring ultra-high memory capacity. The combination of technologies in high-end platforms allows some enthusiasts to use ECC on their motherboards.

There is also RAM that is too fast for your processor, but if installed in the system, it may run at a slower speed on basic settings.

We generally recommend checking with RAM manufacturers who spend a lot of time testing memory on various motherboards. Motherboard manufacturers also provide Qualified Vendor Lists (QVLs) of RAM that they have tested on a particular board. But usually these lists indicate a small number of manufacturers whose memory was in the laboratory. Therefore, it is better to check with the memory manufacturer's list. you can find a lot of useful tips and tricks on RAM modules for hotel platforms and motherboards, as well as information about their speed and compatibility with various processors.

Myths about RAM | Just need to add more RAM

JEDEC is an association of electronic device manufacturers and developers that sets industry standards for widespread adoption among its members. Since some RAM manufacturers have exceeded the JEDEC maximum of DDR3-1600 CAS 11 (and later CAS 9) and offer tighter timings and higher data rates, mixing different RAM modules has not been as easy as originally thought.

Simply put, mixing RAM modules from different sets does not guarantee stable operation, even if you have two identical sets of the same model line. We'd like to add that DIMMs that don't work well together can often, but not always, be made to work by adjusting voltages and/or timings. For article "DDR3 memory: how to improve system performance?" two companies instead of single sets of 32 GB RAM at 2400 MT / s sent us a pair of identical sets of modules in a 2 x 8 GB configuration. Initially, they did not work together, but with the help of minor adjustments, we achieved a positive result.

What is the problem? After all, the modules have the same frequencies, timings and voltage.

DRAM mainly consists of memory chips soldered to a printed circuit board. During the production process of a certain RAM model, the manufacturer may use up a certain batch of printed circuit boards, and then switch to new PCBs from another production batch, which, as a result, may affect a number of characteristics.

The same thing can happen with solder. The manufacturer may start using a different type that has slightly modified conductive properties.

Also, the crystals themselves can be different. During the production process, the chips are binned, that is, sorted according to their quality.

Let's look at this concept from a theoretical point of view. In one production batch, there may be, say, 1000 memory chips that are split or binned. 200 chips can be classified by a manufacturer as entry-level chips, 350 slightly better, 300 even better, and 150 first-class chips. They then sell these chips to different memory module manufacturers.

If you buy DDR3-1866 memory modules from several companies, then most likely you will get different PCBs, solder with different conductive properties, and, quite possibly, chips of different levels from different manufacturers.

The memory chips themselves are produced by several different companies, which only exacerbates the compatibility problem. You probably already understand why mixing different RAM modules often causes problems.

We also noticed that most of the new RAM lines use 4 Gb chips, while the old lines use 2 Gb.

Myths about RAM | There are only a few DIMM manufacturers

This is both a myth and a delusion. There are several memory chip companies and many manufacturers of RAM modules. There are RAM modules made by one or more companies for other companies. For example, AMD Radeon RAM is made by Patriot and VisionTek.

Myths about RAM | DDR-3200 support means any RAM can be used

To use expensive 3200 MT/s memory, you need a processor that can handle such a high transfer rate. Otherwise, the memory will only work in 1333, 1600, or 1866 modes.

In the days of Intel LGA 775 processors, overclocking of the CPU and RAM was performed primarily through the FSB (system bus). Let's say you have a Q6600 processor and your motherboard supports 1066MHz FSB. In this case, the processor will operate at a native frequency of 2.4 GHz, and the memory at a speed of 1066 MT / s. If you want to overclock the processor by increasing the FSB frequency to 1333, then it will operate at a frequency of 3 GHz, and the memory at 1333 MT / s. In other words, the memory speed was limited by the FSB frequency limit. The memory controller was located in the chipset, more often in the northbridge of the motherboard, and also worked at the FSB frequency.

Today the memory controller has moved to the CPU. So the CPU is the main driver of memory operation at the advertised frequencies. Processors based on the Haswell architecture are designed for DDR3-1600 memory, and mid-range and high-end non-K-series chips, as a rule, can work quite stably with memory up to 1866 - 2133 MT / s. The K-series processors are overclockable and their controllers support enthusiast-oriented modules with increased data rates.

AMD's current FX processor line supports "up to 1866 MT/s per DIMM channel". However, you may run into problems running memory in 1866 mode on entry-level and sometimes mid-range processors. This is partly due to the fact that the memory controller of the FX processors is optimized for DDR3-1333 (according to the BIOS and Kernel Programming Guide). Like any other processor, FX chips can be overclocked to run at speeds even higher than DDR3-1866, but this will negatively affect stability.

Myths about RAM | When installing different modules, RAM runs at the speed (timings) of the slowest DIMM

Suppose you have a DDR3-1600 CAS 9 module and you add another module, but already 1866 CAS 9. This may cause the RAM to run at the default settings set by the motherboard, i.e. 1333 CAS 9 or 10 (many AMD motherboards use 1066 by default). Or both modules will work in 1600 CAS 9 (10 or even 11) mode if DOCP, EOCP, XMP or AMP technologies were enabled before installing the DDR3-1866 module.

But you can also set the options manually. Typically, in such scenarios, we would try 1866 mode at 10-10-10-27, increasing the voltage a little, about + 0.005 V. Depending on the results, you can adjust the voltage of the memory controller.

Myths about RAM | Cheaper to buy two sets of DIMMs than one large and expensive set

Even if you buy two identical sets, there is no guarantee that they will work together. RAM modules sold as a set have been tested for compatibility. Manufacturers do not guarantee the performance of mixed kits, even if they use the same memory module models.

Customers often do this with high speed modules and rely on XMP for setup. When XMP is enabled, the motherboard can read the profile of two sticks of RAM and set the secondary timings accordingly, but the tRFC timings for running two modules can be set to 226, while a bunch of four modules will need a value of 314. This problem is difficult to detect, since users rarely go to the settings of secondary timings.

Myths about RAM | RAM runs faster when all slots are occupied

Two sticks of RAM give less load on the memory controller than four. It takes less power, the memory controller needs less voltage to run smoothly, and the RAM is usually slightly faster, although it's not noticeable. The same goes for 3- and 4-channel motherboards. Users are often misled into thinking that four DIMMs (often sold as quad-channel sets) always operate in quad-channel mode, even though dual-channel motherboards cannot in principle do so.

Myths about RAM | RAM faster than 1600 MT / s does not give a performance boost

The validity of this statement depends on several factors. For processors with an integrated graphics core or APU, this is completely wrong, since the video core uses system memory, and the faster it is, the better!

Most RAM tests measure read, write, and copy speed. Many gaming tests when changing RAM 1600 to 2133 show an increase in frame rate from 3 to 5 FPS. This is because in most games, RAM is primarily used as a conduit for transmitting information to the GPU, as well as a buffer for frequently accessed data. The fact remains, RAM can boost FPS quite a bit. Since the price difference between 1600 and 2133 memory is not always large, sometimes buying faster RAM can be justified.

In addition, the WinRAR archiver takes data from RAM and compresses it in RAM before writing it to disk. When changing from DDR3-1600 memory to 2400, the speed increase in tests using WinRAR can reach 25%. There are many other memory-intensive applications: video editing, image manipulation, CAD, and so on. Even a small speed advantage can save you time if you're working on these types of applications.

If you use your PC in office single-tasking, like taking notes, then browsing the web, then watching a video, then you definitely don't need faster RAM. If you prefer to multitask, for example, you have a bunch of browser tabs open at the same time, while working with large spreadsheets or watching a video in a window, or working with images and performing a virus scan in the background, then faster memory can bring certain benefits.

You can test this yourself by running some similar applications with 1600 MT/s memory and then with faster RAM. After downloading several applications, run a benchmark such as SiSoftware Sandra and archive a large file with WinRAR at the same time. While these tasks are running, go through the open Windows windows, then check Sandra's results and backup time.

Myths about RAM | 8 GB is enough for the next ten years

If you really don't like multitasking, then 8 GB will be enough. But this does not apply to gamers and enthusiasts. Five years ago, 2 GB was enough, then 4 GB, and so on.

Another fact: computer manufacturers often skimp on RAM. For example, when 2 GB seemed enough, they installed 1 GB. Today, 6 - 8 GB of RAM is considered the norm and 16 GB is also not uncommon, so it is unlikely that the level of 8 GB will last long as a standard. Games use more and more RAM. If you're building a new system and want to keep it up to date in a few years, we recommend 16 GB of RAM.

Myths about RAM | You will never be able to use 16 GB of memory

This misconception is a continuation of the previous one, but is more relevant to users of memory-intensive applications, as well as those who work with large amounts of files and data. The more RAM you have, the more data it can hold for instant re-access, instead of going to a file on your hard drive or the network to re-download.

Many people use more than 20 GB of memory in the system at the same time almost every day, and this is becoming the norm among members of the Tom's Hardware forum, who often discuss the possibility of maximizing the performance of their 8 and 16 GB RAM kits.

Remember also that manufacturers do a lot of research and contact with software developers and users. So there is certainly a reason why modern motherboards are designed to support 32GB, 64GB, and 128GB (or more) of RAM.

Myths about RAM | I'm not using all the RAM, so the extra memory won't speed things up.

In some situations, increasing the amount of RAM can speed up some processes. Many programs adjust the amount of data stored in memory based on the amount of RAM available, so more RAM saves time by putting more frequently accessed data in RAM (rather than on the hard drive). This can be especially useful when you are working on projects with a variety of images or videos, CAD, GIS, virtual machines, etc. Another advantage of a large amount of RAM is the ability to create a RAM disk to load games, applications and other data. Such a drive has its hidden drawbacks, but many users are delighted with this feature.

Myths about RAM | 64-bit OS allows you to use any amount of RAM

Many people think that you can use an infinite amount of RAM with a 64-bit operating system, but this is not the case. As an example, here are the limits on the amount of RAM in Windows 7:

RAM limits in Windows 7
	x86 (32-bit)	x64 (64-bit)
Windows 7 Ultimate	4 GB	192 GB
Windows 7 Enterprise	4 GB	192 GB
Windows 7 Professional	4 GB	192 GB
Windows 7 Home Premium	4 GB	16 GB
Windows 7 Home Basic	4 GB	8 GB
Windows 7 Starter	2 GB	does not exist

And in Windows 8:

RAM limits in Windows 8
	x86 (32-bit)	x64 (64-bit)
Windows 8 Enterprise	4 GB	512 GB
Windows 8 Professional	4 GB	512 GB
Windows 8	4 GB	128 GB

Myths about RAM | 1.65V memory can damage Intel processors

For its processors, Intel recommends 1.50V memory at a certain data rate. For Haswell, this is DDR3-1600. What's confusing, however, is that Intel also certifies RAM (even DDR3-1600) that runs at 1.60 and 1.65 volts. Keep in mind that 1.60 - 1.65 V is considered normal for DDR3-2133 and higher RAM.

Most memory with lower data rates (such as DDR3-1333 and 1600) use 1.50V or less. We recommend that you refrain from buying RAM with these speeds if it is 1.65 V, as this may mean that the manufacturer used the cheapest and poor-quality memory chips. Why would RAM with good chips need 1.60-1.65 V at all? To further save yourself from future problems, we would recommend not buying DDR3-1866 memory that exceeds 1.50V, unless it has low timings (CL7 or CL8).

Myths about RAM | Dual-channel mode doubles the data transfer rate, i.e. the RAM works twice as fast

This is another misconception. When you install two sticks in dual channel mode, the memory controller does not see the RAM as two separate 64-bit devices, but as one 128-bit device. Theoretically, this should double the bandwidth, but in practice, the speed gain is 20-50 percent on Intel processors and slightly less on AMD chips.

This article was written with the participation of many members of the forum, but there are too many to list them all. We would also like to thank the wonderful employees of companies such as Corsair, G.Skill and Team Group, whose knowledge and experience in this area has helped us a lot.

As always, comments and constructive criticism of the article are welcome.