What is ram memory latency and what is its importance?

Defining the concept of latency

Before we start, let's make it clear what latency is in very general terms. Roughly speaking, this is the time that elapses between a 'request' and its response, that is, the time that passes from when an action is performed (for example, clicking a button) until the response is received (for example, show a window). For example, when we consult the ping in a speed test or in an online game, we are looking at the latency of our network, that is, the time that elapses between sending a data packet and receiving its response.

Are you looking for the best guide to the best RAM memories on the market ? Click on the previous link and you will see the most outstanding models.

Latency times in RAM and how it is calculated

The 'CL' is normally said to be the 'latency' of RAM, but in reality it is only part of the total latency!

In general, what many people consider the actual measurement of RAM latency is the so-called CAS or CL latency.

CAS latency measures the number of clock cycles that pass from when a request to read data is made to when such information is available. So, yes it is a type of latency when measuring the time that elapses between a request and its response, but it is not a real indicator of the TOTAL latency of RAM. Why? Well, because with increasing frequency of RAM memory reduces the time it takes to perform a clock cycle. Remember that the frequency in Hertz (Hz) measures the number of times a cycle repeats every second, so the higher the frequency, the less time it will take to cycle. From here, we managed to get this formula:

time it takes to cycle ( ns ) × CAS latency (“ CL ”)

If it takes 1 nanosecond to cycle and it takes 15 cycles (CL15), the actual latency will be 15 nanoseconds (ns), but if we change this value by 0.7 ns and increase the CAS latency to CL17, the actual latency will be lower than 11.9ns.

In the second example, latency is lower because even if you have to do more cycles, it takes much less time to perform each one.

Now, the role of the RAM frequency here may not have been clear yet, so let's see how we go from the frequency announced by the manufacturers to the time it takes to cycle (ns), to so that the calculation can be carried out without problems.

The first thing we should know is that the most normal thing is that the memory is listed, for example, "DDR4 2133" instead of "DDR4 2133MHz". In the latter case, although we call it the 'frequency' of the RAM it is not really, because the frequency at which the RAM actually works is half, that is, in this case 1066.5MHz. As in DDR (DOUBLE data rate) memories 2 operations are performed per second and not one, the announced 2133MHz would actually be 2133MT / s (millions of transfers per second) and the frequency would be 1066.5MHz.

So when we talk about frequency in this article, we are not referring to the transfer rate, which is what is normally called “frequency”. In any case, they are equivalent: the higher the transfer rate, the more frequency.

So, once we have the frequency data, if we calculate its inverse (1 ÷ frequency) we will finally obtain the period of each cycle in seconds, in this case 0.0009376465 seconds or, better said, 9.38 nanoseconds. It would only be necessary to multiply it by the CL, and we would already have the total latency data. Modifying the previous formula, we can go directly to the result in nanoseconds like this:

(1 000 ÷ REAL frequency) × CAS latency (“ CL ”)

We know that this explanation can be quite messy… that's why we leave you this table that we have done with all the calculations already made for 180 different RAM combinations

DDR4 RAM memory latency

This table raises certain doubts. For example, two very common RAM combinations are 3000MT / s CL15 and 3200MT / s and CL16. Both have exactly the same latency according to our formula, that is, 10ns. However, there is one factor that we have omitted.

The RAM memory (specifically the Dynamic RAM or DRAM that we use in our computers, mobiles, etc.) is organized by various rectangular rows with 8 columns called "words". So, the latency calculation formula that we have used before refers to the delay there is when accessing the FIRST WORD, but we must have two more latencies in consideration: the latency of the fourth and the eighth word. To calculate it, use this formula:

Word N = × (1 ÷ actual frequency)

The Infinity Fabric has its best-known use in interconnecting the core groups or also called CCX, used in almost all Ryzen processors (except some like the 2200G and 2400G APUs). However, access to RAM also uses Infinity Fabric, so its frequency has an even greater impact on memory access latencies.

In the case of Intel, the bus they use performs at much higher frequencies, in most cases above 4000 MHz, but the important thing is not to see which bus reaches higher frequencies because then the reality could be different. Who has the leadership in memory access latencies, Intel or AMD?

Well the objective truth is that Intel Coffee Lake processors have lower memory access latencies than their AMD Ryzen competitors, as we showed you in the image performance tests. This is why people tend to prioritize high RAM frequencies less on Intel processors (from socket 1151), because despite having a similar impact on access latencies to RAM ( as we see the differences there are when changing to 3400MT / s of RAM are similar ), it has no impact on the speed of the ring bus (Infinity Fabric on AMD) on the one hand, and on the other with Intel at factory RAM speeds these latencies are similar to those of Ryzen with RAMs of 3200MT / so more.

And what RAM do I buy?

Once all these technical aspects have been explained and after talking about the importance of memory access latency in modern equipment, the million dollar question arrives: What CAS frequency and latency should I choose to make the best possible purchase?

As seen in the image above, there may be very large price differences between kits with higher speed and lower real latency (which we explained earlier ), something that, added to the extremely high prices of RAM, which has Up in price up to 40% more than a year ago or 160% more than two years ago, it presents a great dilemma in tight budgets where you should save on the chosen kit.

Here, our recommendation is that you look for the best balance between benefits and price. We believe that if you are going to use an AMD Ryzen processor or spend some money on Intel (say, for example, an i5 8600K or i7 8700K), you should try to set yourself at an ideal minimum of 3000 or 3200MT / s ( misnamed MHz ). However, if you are going to mount a much more basic but latest generation platform, a very balanced point in price and performance is the 2666MT / s. In fact, if you are going to mount Intel Coffee Lake with motherboards that are not Z370, you will not be able to raise the RAM of that frequency so it will be the perfect choice. As a final recommendation, if yours are the APUs, to use the full power of its integrated graphics, pay special attention to the RAM, with a minimum 2666MT / s (3000 or 3200 ideal) and always mandatory to use Dual Channel for which you will need 2 RAM modules or more.

Also remember that to use RAM more frequently, you will most likely have to make several adjustments to the BIOS of your motherboard, and check if it is possible to increase its frequency according to the specifications and compatibility of the motherboard.

How to know the latency of my RAM

To know the frequency and CL data of your RAM, all you have to do is download the CPUID CPU-Z application. Once inside the program, obtaining the data is as simple as accessing the "Memory" tab, and consulting the DRAM Frequency (RAM frequency) and CAS Latency (CAS latency). Once that data is obtained, you only need to apply the formulas or take a look at our table to find out the actual latency of your RAM.

Final words and conclusion on RAM memory latency

We know that, in general, this is a topic that can be quite difficult for many, whether they are beginners or not. There are many facts to keep in mind that are generally unknown. Therefore, we will summarize the conclusions we draw from the article in several points:

It could be said that latency, in the case of RAM memories, is the time that elapses from the request to access a data until it is accessed. The data “CL” of the RAM memory, also called CAS latency , is not an indicator of the actual latency of the RAM since it is also necessary to take into account the frequency at which it works ( half of the announced “MHz”: 2133, 2400, 3000…) and, in fact, it is a much more determining factor than the CL. The actual latency of RAM influences the performance of the processor, and more especially in games, which need to access RAM much more often. In AMD Ryzen the frequencies of RAM are also much more important than in the case of Intel Socket 1151 processors (CAS latencies matter basically the same), and especially if you are going to use a Ryzen processor for games, it is preferable to use RAM memory of frequency 3000MT / s or more. When buying, it is best to find a balance between performance and price due to the high current costs of RAM.

So now you know, we hope that our article has helped you to better understand the concept of latency in RAM memory. Do you have any doubts about this? Need advice on what RAM frequencies to buy? Do you have suggestions or criticisms about what we have explained to you? Don't hesitate to leave us a comment or open a discussion in our hardware forum!