In the late evening of no one's time, have you been puzzled by what kind of memory you were buying? Have you ever dared to save money by not knowing memory and dare to buy a big deal for nights of lightning?
Today, a winter took everyone to learn about the simple knowledge of memory, so that you no longer worry when purchasing memory! This article is biased towards popular science, the old driver DIY players can click on the cross in the upper right corner, of course, interested advisers suggest It's not too late to read again.
What is memory?
Memory refers to memory, also known as main memory, which is the space that the CPU uses for direct addressing and storage. It is equivalent to a bridge, responsible for data exchange between processors and processors such as hard disks, motherboards, and graphics cards. We can think of memory as a data buffer, a high-speed buffer..
The reason why memory is called memory is that with respect to the external storage of hard disks, the software data we use are all installed and stored on the external memory. But when we run them, we need to transfer the data of these software to the memory in order to run. Smooth, because the speed of data exchange between CPU and memory is much faster than that of external memory.
What is the difference in memory size?
The memory capacity of this probably most people know, memory capacity and solid state capacity, are a parameter to explain how much data is stored, the greater the memory capacity, the more naturally stored data.
So what happens when the memory capacity is insufficient? I believe that many officials understand what the memory is, guess When the memory capacity is insufficient, the data of our running program cannot be called to run on the memory, and it will cause a noticeable click.
Because of the lack of space in the 'warehouse' of memory, there are already a lot of people inside. , Want to run other programs can only wait to run inside the program data first stop running out of memory.
So in 2018, which is generally more than 3GB in mobile phone memory, how much computer memory is generally enough?
One winter suggested that everyone guarantee Computer memory size is 8GB And above, the reason is that with the development of hardware, the development of PC software is getting faster and faster, and it can drain the performance of computer hardware more and more.
Many day-to-day office applications and browsers have added a lot of feature plug-ins, and they have become more and more memory when running them. This is especially true for those who like to play games, taking the example of “The Jedi Survival”, the hottest in recent years, as an example. , we have tested even 8GB of memory is also a bit stretched.
A wealthy supervisor can naturally add money to 16GB of memory, but now except for very professional work software and extreme computer application environments.
The general day-to-day environment is far from using 16 GB of memory. The continued increase in memory capacity will have a negligible increase in computer performance. Moreover, the memory is so expensive. Therefore, a winter adviser who has money needs no special needs. 16GB memory can be.
Memory frequency is what?
Memory and CPU, like the GPU core, have a frequency, which is the working clock speed of these electronic products. The memory frequency is calculated in units of MHz (megahertz).
Under the same architecture standards, The higher the memory frequency, the faster their clock speed, the more responsive the response, the better naturally.
But unlike CPUs and GPUs, The memory frequency we usually say is the equivalent frequency of memory. Instead of the actual frequency of the memory particles, what is the relationship between the memory equivalent frequency and the actual frequency of the particles? Also, what is the operating frequency of the memory? It is necessary to start with the operating mechanism of DDR memory.
Memory in DDR era can transmit data both at the rising and falling of the pulse, that is, twice the data is transmitted in one cycle, so the equivalent cycle speed (frequency) is twice the single transmission speed (particle core frequency), which is also called particle. The frequency of work.
Secondly, the memory equivalent transfer speed is related to the memory pre-read mechanism. What is the pre-read mechanism?
For example, if one winter runs two steps, each step is 1 meter, then my final running speed is 2m/s, where 2m/s is the equivalent frequency of memory. The data is transferred twice in the memory clock cycle, and the span is the memory's pre-read mechanism.
We can easily find out that if we have bigger spans, the final speed will be faster, and the memory will be the same. The big difference between each generation of memory in the DDR era is that the size of the pre-reading mechanism is different. In order to facilitate understanding, I've put it in order to use the actual particle core frequency of 200 as an example.
We can see that when the memory is pre-read different, the final equivalent speed will be very different. DDR pre-read 2 bits, DDR2 pre-read 4 bits, DDR3 pre-read 8 bits, and DDR4 16 bits.
So when the core frequency of the memory particles is the same, the equivalent frequency of the DDR is twice the core frequency of the particles, DDR2 is four times, DDR3 is eight times, and DDR4 is 16 times. Therefore, the arrival of DDR4 makes the memory equivalent frequency. A stair.
The memory equivalent frequency is the final speed of the memory, which reflects the memory performance. The higher the frequency, the better the performance is.
Single and dual channel is what?
A single memory is controlled by a 64-bit memory controller. Dual-channel memory means using two 64-bit memory controllers to control two separate memories.
The CPU can address these two memory addresses to read data, so that the theoretical bandwidth of the memory can be doubled, and the theoretical data access speed can be doubled correspondingly. It is just like the horse routing into a single lane into two lanes. The car can pass at the same time without waiting.
↑↑↑Single channel 2933MHz memory read and write performance
Dual channel 2133MHz memory read and write capability
It seems that the single-channel and double-channel really improve, but is it really necessary for the dual-channel?
In daily use, 64-bit memory bandwidth is sufficient, but there are some exceptions. For example, if you use a powerful nuclear display, the memory is used as the memory, and the memory bandwidth becomes urgent. , It is absolutely necessary to install dual channels in this case.
Of course, if you are interested, you can try to install dual channels. The price of two 4GB memory will not be much more expensive than a single 8GB memory, and the memory bandwidth will double. Although the performance of the computer will not double, the performance will definitely be improved. Promote.
So the two lanes are so cool, why not do four lanes?
Actually there are four channels of memory, but this requires motherboard and CPU support, and the platforms supporting 4 channels of memory are all expensive enthusiast-class platforms or server platforms such as X299/X399, so for most of our consumers, four channels Memory is not an issue for us, and we do not use it.
What is the memory timing?
The general memory will be marked with data such as '16-16-16-38'. We usually call it memory timing. What is the ghost of this timing?
The corresponding value of the timing is CL-tRCD-tRP-tRAS.
The CL is called CAS Latency, which translates to the clock period required for column addressing.
tRCD stands for RAS-to-CAS Delay and is the difference between the row addressing and column addressing clock cycles.
tRP refers to the RAS Precharge Delay, which means that the precharge requires clock cycles before the next storage cycle.
tRAS is called Row Active Delay and refers to the total time period from the start of operation to the end of addressing when the memory stores data of a row.
The four memory data is the player's most concern. After understanding the meaning, the lower the memory, the lower the four data, the better the memory performance and the more sensitive the response. The most important of these is the first value, CL.
And we generally say 'memory with a time of 16' is specifically a memory with a CL value of 16.
About XMP/DOCP/AXMP
As we all know, game players and DIY players only account for a small number of consumers, and most of them are consumers who do not understand computers. What they need most when using electronic products is to be able to use them stably.
Therefore, in order to stabilize the memory, the manufacturer usually sets the memory factory frequency to be very conservative, and the game DIY player can overclock the memory to improve performance.
However, the traditional way of overclocking the memory is very troublesome. To adjust the frequency, you need to adjust the voltage, further adjust the timing and other small parameters, and then verify the stability one by one through the memory pager. In order to make it easier for everyone to overclock the memory , Intel proposed the XMP memory standard in 2007.
With Intel XMP-certified memory, there are two or more frequency profiles in the SPD. Simply enable these default XMP files in the motherboard BIOS to automatically overclock the memory to 1600 or higher. Simply put, it's a stupid one-button overclocking thing.
Although the XMP standard was proposed by Intel, the CPU market has AMD in addition to Intel. In order to take care of AMD consumers, AMD and motherboard memory vendors have also collaborated on the XMP standard used on the AMD platform.
However, this is called differently in different manufacturers. For example, in the ASUS BIOS, this is called DOCP (Allows you to select a DRAM OC profile, and the related parameters will be adjusted automatically, allowing you to select a memory overclocking file, and The parameters will be automatically adjusted), and the MSI motherboard is more simple and crude. Add an AMD A word in front of the XMP, called A-XMP.
So how do I buy memory?
As an ordinary consumer, The memory capacity is the first to be ensured. The lack of memory capacity will seriously affect the daily use experience. The frequency and timing mentioned later just make the memory performance stronger. We must 'eat' first, and consider not eating well.'
Regarding the amount of memory capacity, the above winter has already mentioned that it is recommended to be between 8GB and 16GB, but if you have special needs, or if there is no money to spend more than you want to fill the motherboard slots, of course.
So what kind of memory stick should we choose after we meet the memory capacity?
The overall performance of the memory can be compared by the frequency divided by the CL value. This value can measure the overall performance of the memory. However, in daily use, the increase in frequency brings about a greater performance improvement than the decrease in the CL value. Therefore, we have given priority to the memory frequency in the limited budget.
General use of 2400MHz frequency memory bar is sufficient, for DIY enthusiasts in memory above 3000MHz frequency should consider other issues.
On single and dual channels of memory, Suggest that you still choose dual channel , For example, I want to buy 8GB of memory, you can choose two 4GB, so although the price is expensive dozens of pieces, but for the performance improvement is worth it.
In terms of timing, when the memory frequency is high enough, for example, I have decided to buy a 3200MHz memory module, then in all 3200MHz different brands of memory, we must begin to consider the timing, this time the memory is short. board.
For DIY players, the purchase of low-frequency game memory with XMP is more cost-effective, their lower frequency and lower price is naturally lower, but with the XMP function as long as the manual into the BIOS settings memory can go to high frequency.
But there are two things to remind here. The first XMP function requires motherboard support, so if you want to buy XMP standard memory first look at whether your motherboard supports, the second is that XMP is not 100% overclocking success.
Otherwise, manufacturers will set these memory settings to the default high frequency, so buying XMP standard low-frequency memory is more like a game of overclocking, but DIY players can still think about it. Ordinary consumers recommend buying the default high-frequency memory directly. This is more stable.
Maybe many netizens have to say that some winter is talking nonsense. Of course it is expensive.
If you think so congratulating you, you have already taken a solid step on the road to becoming a 'water fish'. As you can see in the picture above, everyone is 8GB memory sticks, for most of the friends who use memory by default. Is it more expensive than Kingston's memory performance of 74 blocks?
Write last
Although today's memory is flying expensively, the gold bar is a necessity for small partners who need to be installed. So a winter also wrote this article to let everyone learn as much as possible about memory so as to achieve a witty choice. The effect of purchasing cost-effective memory, but also help everyone save money as much as possible.
The next step is the actual operation, I hope everyone can buy a satisfactory memory.
