The relationship between the clock frequency and the data frequency is clear.

1. Memory and CPU-related data
Before we formally discuss the issue, we should first clarify the relationship between the core frequency of the memory, the clock frequency and the data frequency, and should be particularly familiar with it. This is the basis for discussing this issue. See table below: Table 1 memory names and various frequency lists

Visible from table 1, there is a fixed relationship between the core frequency, clock frequency, and data frequency. The proportional relationship between them is shown in table 2. Table 2. The proportional relationship between the core frequency, clock frequency, and data frequency of the memory

The core frequency is also called the particle frequency or the basic frequency; the clock frequency is also called the operating frequency, memory frequency, I/O frequency, DRAM, dramfrequency, time pulse or memory speed; The data frequency is also called the equivalent frequency or the transmission channel frequency. This article only uses the three names "core frequency", "clock frequency" and "Data Frequency". Be sure to clarify these names. In addition, the name of the CPU is also very different. For example, the frequency is also called Corespeed, FSB name is more, such as busspeed, bus frequency, FSB (and front-end bus mixing) and so on.

2. Basic principles of memory and CPU collocation
I think: Whether you use the CPU is Intel, or AMD, when matching memory must comply with three basic principles. The first is the frequency to synchronize: that is, the core frequency of memory to be equal to or slightly larger than the CPU FSB. Do not add to the memory it cannot withstand the high frequency, otherwise is the frequency "overload". After the frequency is "overloaded," the memory will refuse to work. In this way, the computer is to be blue screen. Of course, when you give it a frequency below the core frequency, it will be competent and enjoyable. The second is to match the bandwidth: you should try to make the memory bandwidth equal to the bandwidth of the CPU front-end bus, otherwise, the transmission capacity of the data will be constrained by the lower bandwidth; The third is the motherboard to regulate: because the above two conditions are sometimes impossible to meet at the same time. This depends on the motherboard to regulate, the main method of regulation is asynchronous set landlord. Because the first article is about the problem of life or death, this must be met. The second one can be handled flexibly. Also, when discussing how memory is paired with the CPU, it is important to make the following clear: Memory rejects work when the core frequency of ① memory is less than FSB. Performance is a blue screen. However, in any case, the core frequency of the memory is larger than the CPU FSB, the memory will be normal

Job. However, the system does not recognize its high frequency. Only admit that its core frequency equals FSB. For example, when you insert a ddr2-1066 into a FSB 200MHz board, the system will assume that the memory is ddr2-800. The front-end bus frequency (FSB) of the ②intel processor is 4 times times the FSB frequency. However, in the history of the front-end bus frequency and FSB is the same, so people often use the FSB to represent FSB (such as software cpu-z, it means that "fsb:dram" actually refers to "CPU FSB: Memory Clock frequency"). AMD's memory, before K8, also has a front-end bus, however, at that time the front-end bus frequency is twice times the FSB frequency. AMD's CPU does not have a front-end bus after K8. ③ when discussing the matching of memory with the Intel platform's CPU, you must know the CPU's FSB or front-end bus frequency. Know one on the line, because they are four times times the relationship between, at this time, do not care about the CPU frequency is how much. However, since the Intel I3/i5/i7, the reference CPU frequency is not the FSB and front-end bus FSB, but the basic frequency bf. However, in the detection software cpu-z, or the BF frequency is called FSB. In this case, how to match the memory, please refer to the article 6. ④ when talking about memory matching with the AMD platform CPU, you must first make it clear that the CPU is K8. K10 's? Or K8 before? Because these three kinds of CPU to accompany the memory method is very different. When discussing K8 CPU, just know the FSB or FSB; when discussing to K8 CPU with memory, just know the CPU frequency, do not mention the HT bus, because the memory with the HT bus independent, the same reason, when the memory for K10, if you do not overclock, just know the CPU

Or the motherboard supports what memory is available. Because the CPU of the K10 architecture is hooked to FSB with memory, the AMDCPU FSB is fixed at 200MHz. ⑤ when discussing the relationship between memory and CPU, it is calculated based on the parameters of memory and CPU, regardless of the motherboard. However, once you have chosen your collocation plan, you must also consider whether the motherboard supports it. Under normal circumstances, the frequency of the motherboard should be higher than the FSB frequency. ⑥ memory asynchronously is the frequency with which the memory is added to it to work. Generally refers to the reduction of frequency. However, the consequence of the frequency drop is slower speed and narrower bandwidth. All motherboards support asynchronous running of memory. However, the degree of support is different. ⑦ from the point of view of whether memory is paired with the CPU, there is no need to consider the amount of memory in memory. However, when the capacity exceeds a certain size, it is sometimes necessary to consider whether the motherboard is supported. ⑧ memory matches the CPU, it has nothing to do with the CPU being a few cores. These are the three principles and eight considerations that I think are memory matching.

3. How frequency synchronization is achieved on Intel platforms
On the Intel platform, the memory frequency is synchronized with the CPU: the core frequency of the memory must be equal to or slightly larger than the FSB of the CPU. Otherwise, the "frequency overload", the memory will stop working, the main manifestation is the blue screen. Although I often hear different voices in this statement. However, I firmly believe that my inductive method is completely able to stand up to the feet. Because memory can withstand how much frequency, is entirely determined by the core frequency. Both the clock frequency and the data frequency are amplified by technical means on the basis of the core frequency. When the core frequency is equal, different types of memory have different clock frequencies and data frequencies.

For example, on Intel platforms, when FSB is 200MHz, in order to achieve frequency synchronization, you should choose DDR-400, ddr2-800, or ddr3-1600. Because the core frequency of these kinds of memory is 200MHz. This is a very important rule. However, in the past, people are saying that the memory synchronization condition "Memory clock frequency (instead of saying" memory core frequency ") is equal to or slightly greater than FSB frequency is synchronous". More ambiguous statements, such as "The frequency of memory is equal to or slightly greater than FSB, is synchronous." Many of the articles on the internet have been said so far. Quite chaotic. I think that in the SD and DDR era, it is perfectly correct to say that the clock frequency and the core frequency are equal in sync. However, after the advent of DDR2 and DDR3, it is completely wrong to say this again because of the different types of memory. Its core frequency is different from the clock frequency. Therefore, there are great differences between the two methods of discrimination. However, by habit, people are still willing to use clock frequency to compare with the FSB of CPU. I think it's not an option to tell if memory is synchronized by the clock frequency versus CPU frequency. However, this must be likened to an appropriate change. For example: "For SDRAM and DDR, memory is synchronized when the clock frequency of the memory is equal to the FSB of the CPU." For example, when FSB is 200MHz, it should be equipped with DDR-400 memory "; for DDR2, it should be said:" When the memory clock frequency is twice times the FSB, it is memory synchronization. For example, when FSB is 200MHz, it should be ddr2-800 memory "; for DDR3, it should be said:" When the Memory clock frequency is 4 times times the FSB, it is memory synchronization. For example, when FSB is 200MHz, it should be equipped with ddr3-1600 memory ". Isn't that too wordy?

However, in the detection software cpu-z, the ratio relationship between the CPU's FSB (which is called the FSB) and the clock frequency (known here as the DRAM) indicates whether the memory is synchronized. The relationship is shown in the following table: Table 3 The frequency relationship between the CPU's FSB and the memory clock frequency when memory is synchronized
The type of memory Fsb:dram (that is, the ratio of the CPU's FSB to the Memory clock frequency) Ddr1:1ddr21:2ddr31:4

Note: In the detection software cpu-z, is to use "FSB" to express the CPU FSB, this is the misuse of historical formation. The "DRAM" in the table above refers to the clock frequency.
Knowing the above principles, we should choose what kind of memory and CPU to match the problem will be solved. I think: to Intel's CPU before the I3/I5/I7 architecture, as long as you can know the FSB or front-end bus frequency (fsb=4 times FSB) One of the parameters, you can select the appropriate memory. Different types of memory have different methods of calculation. The specific calculation method is:

① when selecting DDR memory on Intel platforms:
Ddr-xxx three "x" =FSB value ÷2 (1) For example, when the Intel platform FSB is 533MHZ, if the DDR memory is selected, xxx=533÷2=266, the DDR-266 type of memory should be selected. This option also conforms to the principle of "core frequency of memory (133MHz) equals CPU FSB (133MHZ)". ② when DDR2 memory is selected on the Intel platform: three "x" =FSB values after ddr2-xxx (2)

For example, when the FSB on the Intel platform is 800MHz, if you choose DDR2 Memory, the XXX=FSB value = 800, so you should select the ddr2-800 type of memory. This option also conforms to the principle of "core frequency of memory (200MHz) equals CPU FSB (200MHZ)".
③ when DDR3 memory is selected on the Intel platform: four "x" =FSB values after ddr3-xxxx x2 (3)

For example, when the Intel platform FSB is 800MHZ, if you choose DDR3 Memory, you should select Four "x" is equal to 800x2=1600 memory, that should be selected ddr3-1600 type

Memory. This option also conforms to the principle of "core frequency of memory (200MHz) equals CPU FSB (200MHZ)". The following table is a calculation of the memory based on the parameters of the CPU, based on the above principles. This is the highest frequency computed based on the condition of the frequency synchronization. There are spaces that are asynchronous to lower frequencies. (I3/i5/i7 the CPU of the architecture how to select memory, see the 6 of this article) Table 4. FSB frequency and memory model in memory synchronization

Note ①: the "frequency ratio" in the table refers to the ratio of the FSB to the clock frequency in the case of synchronization, that is, the "Fsb:dram" value in Cpu-z. If the cpu-z is not showing this value, then the memory frequency is asynchronous.
According to Cpu-z in the "Fsb:dram" ratio can also be easily reversed to install in the computer, (or set landlord) is what type of memory. For example, when your CPU's FSB is 200MHZ and the installed memory is DDR2, if the cpu-z display is "fsb:dram=1:1.66", then it is certain that the memory you are installing is not in sync state. Because DDR2 memory is synchronized, its ratio is 1:2. So what kind of memory is installed? Can be calculated as follows: Because 200:x=1:1.66, so, x=333. This 333 is the clock frequency value of the real memory 333MHz, which can be concluded that this memory is ddr2-667.

Note that the pins and sockets of ②:DDR, DDR2 and DDR3 memory are different and cannot be used incorrectly. So, when choosing a motherboard, figure out what kind of memory the motherboard supports. At present, for the convenience of users, some motherboards have both DDR2 and DDR3 memory slots.

4. Calculating methods for bandwidth parity on Intel platforms
What is bandwidth? In the computer field, the bandwidth is also called frequency width, refers to the amount of data that can be transmitted in a fixed time, that is, the ability to transmit data in the transmission channel. Bandwidth is typically expressed in bits per second (bps), which is the number of bits that can be transmitted per second. However, in order to be simplified, it is usually expressed in MB/s. Bandwidth is calculated by: Bandwidth = frequency x Bus position/8 (4)

Because the number of transmit bits of the computer memory bus is 64 bits, for the frequency is 800MHz,, the channel, bandwidth =800mhzx64b÷8b/b=6.4gb/s. There is also a simple calculation method, that is, bandwidth = frequency x8b (5)

For example, when the frequency equals 800MHz, the bandwidth is 800mhzx8b=6400mb/s=6.4gb/s. The calculation results of different frequency and bandwidth relationships are given in table 1. The bandwidth match here is actually the match of the frequency. For our ordinary users, or by frequency to discuss the more intuitive, because the bandwidth is equal, that is, the frequency is equal. The principle of bandwidth matching is that the data bandwidth of the memory should be equal to the bandwidth of the FSB on the CPU front-end bus, or "the data frequency of the memory equals the FSB frequency of the CPU, which is the bandwidth matching". Only in this way can we guarantee the smooth transmission of data. If one side has less bandwidth than the other, the small-bandwidth party poses a "bottleneck". However, if the CPU is already selected, the memory bandwidth is larger, and there is no harm.

Under the condition of satisfying synchronization, for DDR2, it will automatically meet the requirement of equal bandwidth, that is, the bandwidth of the memory can be equal to that of the front bus; however, for DDR, if synchronization is required, the bandwidth of the memory is only 1/2 of the FSB bandwidth for DDR3,; If the synchronization requirements are met, the bandwidth of the memory is twice times the FSB bandwidth. This relationship is shown in the table below. Table 5 Memory synchronization, the bandwidth of the FSB with the bandwidth of the memory data channel the type of memory DDRDDR2DDR3
1:2

Bandwidth of FSB: bandwidth of the memory data channel 2:11:1

The synchronization requirement for memory is either raw or dead, so it must be met, if not met, to run asynchronously. However, the requirements for bandwidth are somewhat more flexible. If not, it can be solved with a dual-channel or asynchronous approach. In the DDR era, the use of a two-channel approach to resolving memory bandwidth is a "bottleneck" problem. However, if you do not compose a dual channel, use only one memory bar, as long as the frequency is synchronous, it can also work properly. On the AMD platform, its memory controller landlord into the CPU. That is, the CPU deals directly with memory. The problem of bandwidth is not so important anymore.

5. How to control memory operating parameters on the Intel platform with the motherboard
① Motherboard automatic control function: In fact, according to the memory synchronization requirements of the selected memory (see table 4) can be used normally. However, there are exceptions to the situation. For example: Sometimes, according to the calculation results, should use ddr2-800 memory to synchronize, but, you do not have ddr2-800, only ddr2-667 memory, and installed up. What do we do? Depends on the motherboard control. As long as you do not select "Manual" in the "Memclockmode" in the BIOS, and do not change the default setting landlord "Auto" in the "Systemmemorymultiplier" item, the motherboard will automatically reduce the clock frequency added to memory according to the SPD data on the memory bar.

333MHz, so that the frequency of ddr2-667 memory is not overloaded. At this time cpu-z "Fsb:dram" is "1:1.66", not "1:2". After this adjustment, the speed is reduced and the bandwidth of the memory becomes smaller, but it can be run normally. Also, if memory synchronization needs to meet the memory is ddr2-800, and you install or set landlord ddr2-1066 memory, the motherboard will add to the memory frequency as ddr2-800 treatment, cpu-z to "Fsb:dram" Display or "1:2", will not display "1:2.66" But. Yes, if you are overclocking memory to 1066MHz, Cpu-z will display "1:2.66". ② Motherboard Manual Control method: If you want to manually set the landlord memory is also possible. This is done by selecting "Manual" (manual) in the "Setmemoryclock" entry in the BIOS, or by selecting "Auto" in the "Systemmemorymultiplier" key, and choosing the memory that you actually installed on the memory slot, and the landlord will succeed. Of course, you still need motherboard support, that is, if the motherboard has this kind of landlord project. There are several ways to manually set landlord options. Some are directly in memory chip name (such as ASUS P5E3WS Motherboard, directly with such as ddr3-1333), and some are digital (such as gigabyte GA-P31-S3G motherboard with 3.33 or 4.0. 3.33 means ddr2-667 memory), of course, some motherboards are expressed in fsb:dram ratio. 6. There is no FSB on Intel platform, how to match the memory? Intel Core I3/I5/I7CPU is putting the memory controller inside the CPU. Memory controller-controlled QPI (Quickpathinterconnect, fast-track interconnect) bus frequency is much higher than the traditional FSB (typically 18 times times the FSB), so the space for it to match the memory is also mostly. For example, in the past, when FSB was 133MHz, Fsb=4x133mhz=533mhz, the relationship was fixed. If DDR memory is selected, the highest is DDR-266;

If you choose DDR2 Memory, the highest is ddr2-533, if DDR3 memory, the highest is ddr3-1066. If you choose higher-frequency memory, the system is not recognized, it can only degrade them. The FSB actually became a "bottleneck". Since the QPI bus has been available, the choice of memory is much looser. At this time the Cpu-z "FSB" 133MHz, is not the traditional sense of FSB. It is called the base frequency (BF). Therefore, the synchronization relationship between it and the memory frequency is changed. For example, when "FSB" (base frequency) 133MHz, it is even possible to choose dde3-1600 memory. Yes, but it's not unlimited. To what extent? This depends on the CPU or motherboard requirements. However, one approximate estimate is that the data frequency of the memory is roughly 1/2 of the qpi frequency. How much depends on the memory controller and motherboard support to what extent, not the user can be left or right. 7

Landlord problems with dual-channel memory
First of all, it should be clear that the dual channel is designed to increase the bandwidth of the memory, because, in the DDR era, the CPU FSB has reached 800MHz, but, from a frequency synchronization point of view, only with DDR-400 memory. Obviously, the bandwidth of the memory is only 1/2 of the FSB bandwidth, and the memory becomes the "bottleneck" of bandwidth matching, because the CPU has half the time to wait for the memory to pass the data over. Therefore, the concept of double channel is proposed. That is, add a memory controller in the North Bridge and add another DDR-400 memory to make two DDR-400 memory parallel use,, the bandwidth matching problem can be solved.
However, since two of the memory in the dual channel is controlled by two memory controllers, the two channels are made up of only the memory bandwidth, and there is no improvement in the performance metrics (such as core frequency) for each memory. Therefore, when using two DDR-400 to compose a dual channel, its performance is not DDR-800. Widely circulated on the internet "with two ddr2-533 composed of two channels, used in the Fsb=1066mhz platform, is absolutely matching" the saying is wrong. Because the core frequency of ddr2-533 is 533mhz÷

4=166mhz, and Fsb=1066mhz when the FSB is 1066mhz÷4=266mhz. Two frequencies are not equal. In other words, such a match does not satisfy the synchronization condition. Hard to do so with the computer is to blue screen. Because the memory is composed of two channels, their core frequency does not increase.

The role of two channels can be likened to this: The road is wide, can accommodate two cars, but only a car back and forth, this is not a waste? Buy another car, let them run in parallel on One road, freight volume does not increase by one times! This shows that the dual channel only solves the problem of freight volume (bandwidth). It's not the problem of doubling the speed of the car. That is: Two ddr2-533 composed of two channels, only to increase the bandwidth, and can not increase the frequency of memory, two ddr2-533 composed of two channels, the effect is not equal to a ddr2-1066.
Of course, if you must use ddr2-533 for Fsb=1066mhz platform, it is not impossible, the method is asynchronous, so that the frequency of adding to the ddr2-533 down. However, does the motherboard support it? If it is supported, then it needs to be run asynchronously, is this a "perfect match"?

8. How to configure memory before AMDCPUK8
Prior to K8, AMD memory also had a front-end bus (FSB), but there was a little difference with Intel that the front-end bus frequency was twice times the FSB (Intel was 4 times times). Because memory synchronization or not is considered FSB, so when the FSB with the core frequency of memory is synchronous. Therefore, the FSB is 133MHz, should be equipped with DDR-266 memory, FSB 166MHz, should be equipped with DDR-333 memory, FSB is 200MHz, should be equipped with DDR-400 memory. These are all consistent with Intel's platform. The difference is that the bandwidth is balanced when the frequency is synchronized on the AMD platform. As a result, there is no dual-channel problem with AMD platforms at the time. The matching method of memory in 9.AMDCPUK8 era
In AMD's K8 CPU, the memory controller is integrated into the CPU (see), the memory directly with the CPU. Front-end bus does not exist, and North Bridge connected to the bus renamed HT bus, this bus, in addition to no longer with the memory contact, the other role with the original front-end bus task is

The same.

Because of these changes, the memory-to-CPU collocation of the AMD platform has also changed greatly. Because at this time we are only concerned about the CPU frequency. No more thinking about FSB and FSB. Now the introduction of memory and AMD's CPU Collocation method is as follows: ① first calculates the memory controller's division multiple DIV (this calculation is automatically carried out by the computer, of course, we should also grasp): div=cpu Nominal Frequency ÷ You set the clock frequency of the landlord memory in the BIOS (6) This set landlord clock frequency is best consistent with the memory you actually installed. If not, the computer is calculated based on the landlord value you set. If the DIV evaluates to not an integer, the integer is entered, for example, 7.4 is entered as 8. If the calculated crossover factor is less than 5, then take 5. These methods are all AMD's design rules. ② know the memory divide multiple div, the actual clock frequency through the memory by the following formula: The actual clock frequency of the memory =cpu the actual frequency of the memory (7) Know the clock frequency of memory, you can know how to match the memory. For example, you calculate the clock frequency is 200MHz, and is the choice of DDR memory, it should be equipped with DDR-400; you calculate the clock frequency is 400MHz, if you want to match DDR2, you should choose ddr2-800; if you calculate the clock frequency is 400MHz, Want to match DDR3 memory, or should be with ddr3-800. Because no matter what type of memory, the data frequency is always twice times the clock frequency. Example 1:CPU is the AMD Fast Dragon (Athlon) 643200+, the frequency is 2.0GHz, if you intend to choose DDR-400 memory and in the BIOS to do such a set landlord when asked ① frequency division is how much? ②: What is the actual frequency of over-memory? Solution: ① because the clock frequency of the CPU equals 2000mhz;ddr-400 is equal to 400mhz/2=200mhz, so the frequency of the memory is Div=2000mhz÷200mhz=10;② because the CPU is not overclocked, so, Its frequency is still 2000MHz, so,

The actual bus frequency through the memory is =2000mhz÷10=200mhz, which is quite DDR-400. Therefore, it is possible to select DDR-400 memory. Of course, if the motherboard allows, you can also
To form a dual channel with two DDR400. To increase the amount of data transferred.

Example 2:CPU is the Flash Dragon 3200+, the main frequency is 1.8GHz, if the use of ddr2-800 memory, Q: What is the number of ① frequency multiplier? What is the actual clock frequency of the ② through memory? (3) If the frequency is over 2400MHz, how to choose the memory? Solution: ① divide coefficient div=1800mhz÷400mhz=4.5, should be taken up to 5;② through the memory of the actual clock frequency =1800mhz÷5=360mhz. Memory degradation is used for ddr2-720. Because the memory frequency is backwards compatible, the motherboard will support such memory. Therefore, it is also possible to match landlord. (3) If the frequency is over 2400MHz, then the clock frequency through the memory is 2400mhz÷5=480mhz, that is, the need to select ddr2-960 memory. Choose ddr2-1066. (however, because the frequency is too low, but also to see whether the memory controller or motherboard support) is that I based on the above principles of the results of the table, in order to simplify, in the table does not have the frequency division factor listed. A figure with a gray shading in the table indicates that the calculated crossover factor is 5 smaller. Therefore, when these clock frequencies are obtained, the crossover multiples are calculated as the divide factor equals 5. I believe that the table for reference, how memory with the CPU is very clear.

Now, we use the above two examples for example, to study how to get the results from the following table directly with memory: (Note: The font size is small, not the author's set landlord, is the journal editing function of the problem.) Just changed
Version, not yet perfect. Moon People too! ）

In the first example, the CPU frequency is 2GHz, in the BIOS set landlord memory is DDR-400, so from this table can be found through the memory clock frequency is 200MHz, that memory runs in DDR-400. This frequency matches exactly. The second example: CPU frequency is 1.8GHz, choose ddr2-800 Memory, from the table can be found in memory running frequency is 360MHz, quite ddr2-720 memory. It can also be seen from the table: if you want to make this memory run in ddr2-800, you only have to select 2.0GHZ or more CPUs. If the CPU's frequency is still 1.8GHz, even if you choose ddr2-1066 is not helpful, it will only degrade memory to ddr2-720. There are many users asked: I installed the memory frequency is very high, but in the cpu-z detected frequency is very low? The main reason is that the CPU you are installing is too low, and the motherboard does not support high frequency memory. Table 6AMD CPU and memory matching method list

The corresponding divider coefficients are not listed in this table and are therefore not conducive to overclocking. In fact, if you do overclocking, simply multiply the data in the table by the "CPU frequency after overclocking/overclocking CPU" to get the clock frequency added to the memory after overclocking. From this table we can clearly see how the memory should be chosen. For example, when the CPU frequency is 1600MHz with DDR-400 can be when the CPU frequency is 2000MHz with ddr2-800,;, is just right, when the CPU clock is 3200MHz, with ddr3-1066 is appropriate. I guess: the motherboard designer is to follow this idea to set the motherboard default memory frequency. This table is based on CPU and memory parameters under K8 conditions and has now evolved to K10 (seemingly without K9), so the situation has changed slightly. There is also an important factor to ignore, that is, in addition to motherboard support, the memory controller supports this frequency. If the controller is not supported, only motherboard support is not available. So, the above table is only a set of principles, if you use cpu-z for testing, the results are not consistent with the table, it is necessary to look for reasons from many aspects.

Matching method of CPU memory of 10.AMDK10 class
AMD's K10-Class CPU memory matching method and calculation method have changed a lot compared to K8. The order of calculation is: The first step: understand what kind of memory your selected CPU supports. For example, it could be ddr2-800, ddr3-1333, and so on, and landlord the memory in the "dramfrequency" key of the BIOS. Of course, you can also set landlord other models of memory. The system doesn't care what memory you plug in, but calculates it based on your landlord memory; the second step is to calculate the crossover factor of the memory divided by the memory set in the BIOS = The data frequency of the memory set in the BIOS ÷ FSB (8) memory data frequency value is how large, can be known from the memory model. For example, the ddr2-800 data frequency is the 800mhz;ddr3-1333 data frequency is 1333MHz. And so on By the formula (8), when the memory set in the BIOS is ddr2-800, 800mhz÷200mhz=4, that is, the division factor = 4, when the BIOS is set landlord FSB is ddr3-1333, 1333mhz÷200mhz=6.67, that is, the divisor coefficient equals 6.67.

The second step: the actual data frequency in memory based on the following formula is calculated by the actual data frequency = actual FSB x divide factor (9) Example 1: When the memory is set to ddr2-800 in the BIOS, if the FSB is over 230MHz, what is the frequency of the data through memory? A: Because the ddr2-800 data frequency is 800MHz, and AMD's CPU nominal FSB is 200MHz, so the crossover factor is equal to 800mhz÷200mhz=4. According to formula (9) can calculate the data frequency through memory is 230mhzx4=920mhz. That's pretty much ddr2-920. Poor physique ddr2-800 memory is unable to bear. It is best to install ddr2-1066 memory. The following table is a calculation of the data frequency and crossover factor for the landlord memory set in the BIOS. The memory given in the table is the actual model. However, in the list of motherboards, there may be landlord such as ddr2-1111, ddr2-1200, and so on. They are not included in this form and can be computed by the reader if such a landlord is encountered. Table 7. The dividing factor of AMDK10CPU

Note: The Fsb:dram in the table is the corresponding test result of the detection software cpu-z.

Example 2. Using AMD K10cpu, the landlord memory in the BIOS is ddr3-1333, motherboard is also supported, no overclocking. Excuse me: What is the frequency of data in memory? A: The above table can be found, in this case, the memory factor is 6.67, therefore, the actual data frequency in memory is 200mhzx6.67=1333mhz. As can be seen from this example: if Overclocking is not considered, the actual FSB and nominal FSB are equal, equal to 200MHz. Thus, the formula (8) shows that the actual data frequency of the memory is equal to the data frequency of the memory that you set landlord in the BIOS. So, the question of how to choose memory in K10 becomes very simple. It's simple enough to say that when you put memory into AMD's K10 class CPU, you set landlord in the BIOS what kind of memory, in memory frequency is you set landlord frequency. However, the large pre-problem is CPU memory controller and motherboard support. After you insert the CPU support memory bar, how the system runs this memory, you don't have to worry about it, just let it "Aout" or "byspd". However, if you want to manually set the landlord, you can set the BIOS "Memoryclockmode" key landlord to "Manual" (manual mode), and then use the above method to set landlord.

11. Memory Matching Method Example question and answer
Q 1: On the Intel platform, if the FSB is 200MHz, if I choose DDR memory, what should be the model? If you choose DDR2 memory, what model should I choose?

     a DDR-400 should be selected when DDR is selected according to the principle of memory core frequency equal to or not less than FSB;:, when choosing DDR2, you should choose ddr2-800. Q 2: What happens when I install ddr2-1066 memory on Intel platform Fsb=800mhz? A: Because the system provides a clock frequency of only 400MHz to memory, and ddr2-1066 requires 533MHz. Therefore, the system will treat ddr2-1066 as a good ddr2-800. Q 3: What happens when I install ddr2-800 memory on Intel platform Fsb=1066mhz? A: Because the system provides a clock frequency of up to 533MHz to memory, the ddr2-800 can withstand a clock frequency of 400MHz. Therefore, the memory needs to run asynchronously, that is, the clock frequency added to the memory down. This frequency reduction is what we call asynchronous. Async is done automatically by the motherboard. Of course, it can also be manually operated. Q 4: My CPU is AMD's K8, the frequency is 2600MHz. Motherboard also support ddr2-800, but, after I installed, cpu-z test results are: Frequency only 371.4MHz. Why is that? A: Let's calculate: 2600mhz/400mhz=6.5, in 7. 2600mhz/7=371.4mhz, quite ddr2-743. So, you said the situation is normal. On the AMD platform, the actual frequency of memory is often depressed. Q 5: My CPU is AMD's K8, the frequency is 2200MHz, the installation is ddr2-667. Why is the DRAM (Memory clock frequency) seen above the Cpu-z 314 or so? Rather than 667/2=333? A: Let's calculate: because you set landlord is ddr2-667 memory, so there is 2200mhz/333mhz=6.6, should be taken as 7. 2200mhz/7=314mhz. So, your computer is normal. On the AMD platform, not what frequency you set landlord, through memory frequency is you set landlord frequency. Q 6: What does "front-end bus: Memory" mean "CPU/9" when using cpu-z detection on AMD platforms? A: When the AMD platform uses CPU-Z to detect the frequency of the computer, such as in the "front-end bus: DRAM Frequency" entry is expressed as "CPU/9", where the "CPU" indicates the frequency of your CPU, "9" represents the frequency division factor. If your CPU's frequency is 1.8GHz, then, 1800mhz/9=200mhz. This means that your memory clock frequency is 200MHz. Q 7: What does "3:8" mean when the "front-end bus: Memory" item appears when using CPU-Z to detect a computer? Answer because AMD's CPU FSB areis 200MHz so there is "3:8=200:x" from this x=200:,,x8÷3=533, that is, your computer's memory clock frequency is 533MHz. So, this memory may be ddr2-1066 or ddr3-1066. Q 8: My AMD CPU frequency is 3000MHz, I choose ddr3-1333 memory. However, when using cpu-z detection, "Fsb:dram=3:8". What does that mean? A: On the AMD platform, the ratio is "3:8", that is, the hour-hand frequency is 533MHz, quite ddr3-1066. Your ddr3-1333 is being demoted to use. Why downgrade? This depends on the motherboard set landlord, may also be the default motherboard landlord is ddr3-1066, may also be CPU memory controller maximum support to 1066MHz. If not the memory controller does not support, as long as the motherboard support, you set its frequency landlord to 667MHz is also possible. Q 9: I use the AMDK8 CPU, the frequency is 2800MHz. What kind of memory should I choose? A: As you can see from table 6, in this case, you can choose ddr2-800. The choice of ddr2-1066 seems a little higher. Because your CPU frequency is relatively low, choosing DDR3 is not necessarily appropriate. However, it depends on whether the motherboard and memory controller are supported.

Q 10. My amd8650cpu matches the ddr2-800 memory bar. Why did cpu-z show the memory frequency 480MHz when I put FSB over 240MHz? A: The crossover factor is 800mhz/200mhz=4. So, the data frequency of your memory is 240mhzx4=960mhz, which is quite ddr2-960. Its hour-hand frequency is exactly 960mhz/2=480mhz. So, Cpu-z's display is completely correct. When selecting memory for AMD's CPU, be sure to figure out if this CPU is K8, K10? or K8 before. Because there are three types of CPUs that have different methods of matching memory. Q What is the frequency of 11:HT related to the frequency of memory? A: Because AMD's memory controller is inside the CPU, it accepts CPU control directly. Although the memory bus and the HT bus originate from the CPU, each of them has its own task (see the drawings in this article), which is not related to each other. Therefore, it is not necessary to consider the frequency of the HT bus when matching memory. However, when using integrated graphics cards that require shared memory as video memory, there is some correlation between the two. However, you do not need to consider these when considering matching memory. Q 12: Is there a need to consider the matching of memory bandwidth to CPU bandwidth for AMD platforms? A: Because the memory is directly involved with the CPU on the AMD platform, the bandwidth of the CPU is the bandwidth of the memory. Therefore, when matching memory, the bandwidth of the CPU and the memory bandwidth matching problem can no longer be considered separately. As described in this article, the option of memory is available. However, in the K8 era, if it is ddr2-800 memory, CPU frequency must be increased to 2.8Ghz, in order to use it when ddr2-800, in fact, this is the bandwidth matching problem. Q 13: What is MCT? What is a DCT?

A: MCT (Memorycontroller) is a memory controller that controls memory reads and other operations, which determines the K8/K10 memory performance. Its control function is suitable for any type of memory. and DCT (Dramcontroller) is "Memory Interface Controller", also known as DRAM controller. It determines the type of memory that can be used. It can be upgraded to accommodate different types of memory. In other words, the memory controller supports what type of memory is determined by it. Q 14: What does dctsmode mean in the BIOS that supports the AMDK10 motherboard? A: Dctsmode (DRAM controller mode) refers to the two operating modes of the two independent 64bit unit memory controllers installed in the K10. There are Gangedmode and Ungangedmode. The latter is the default mode. The former mode is the same as the K8 dual channel mode, the memory of the two channels is logically connected as a whole, constitute 128bit, this mode is called "gangedmode". If two controllers start at the same time, they all have the same 128bit of data transmission as each time pulse cycle. This is not a dual channel, nor a single channel, but two single channel running simultaneously, called "ungangedmode" mode. The benefit of this pattern is that the consistency of the two memory requirements is not high.
Answer hmhding Netizen: Because the following words "more than" 1000 words. Not add up. So put it here to answer you:
The purpose and effect of the first problem control is to let the frequency of memory be added to the ddr2-667 when FSB is 200MHz:. , when the memory is added to the frequency is 166MHz instead of 200MHz. Of course, there is no blue screen problem. The second problem: when the FSB is 200MHz, with two DDR400 to form a dual channel, is to solve the bottleneck of bandwidth. In fact, with a DDR400 can also. A third question: your FSB is 100mhz,fsb:dram=3:4. From this, you can deduce that the DDR400 you are using has been reduced to DDR266 use. As for why it is not dropped to DDR200, I can not tell. Is the motherboard not supported? Please take a look at your motherboard manual. Thank you.

The reply to your second letter (just a few words, people always say more than 1000 words, not affixed): from this paragraph you are fully aware of why the memory frequency only downgraded to DDR266, rather than downgrade to DDR200 reason. Because the meaning of the "host/dramclockratio option" is actually the data frequency/FSB of the memory. The motherboard defaults to 2.66, so the data frequency of the memory is equal to 100mhzx2.66=266mhz. I made this decision according to GA-81 's instructions, and you can take a closer look at your instructions.

The relationship between the clock frequency and the data frequency is clear.