With the increasing number of overclocking players, the asynchronous memory technology has also begun to flourish. However, the technological development has also brought about more confusion. Even the "prawns" cannot fully figure out the current asynchronous memory technology. It is precisely for this reason that memory asynchronous technology is necessary to solve the problem. If you are interested, you may go on to understand what the memory asynchronous technology is like!
1. What is memory synchronization and Asynchronization?
Memory synchronization means that the memory frequency runs at the same frequency as the CPU external frequency. That is to say, in the case of memory synchronization, the memory frequency = the CPU external frequency. For example, when the 520 MHz external frequency p4 is synchronized with the memory, the memory also runs on the MHz external frequency. Because the DDR memory is used, the memory frequency is 400 MHz × 2 = MHz (ddr400 ). The memory asynchronous technology is to make the memory frequency different from the CPU external frequency, the memory frequency = CPU external frequency × N/m (a specific ratio ). When the 520 MHz external frequency P4 166 is asynchronous in the memory, the memory can run at MHz, and ddr333 memory is enough. Of course, the memory can also run at 266 MHz. At this time, the system only needs to use ddr533 memory. That is to say, the ratio of N/m can be greater than or less than 1, that is, when the memory is asynchronous, the memory frequency can be higher or lower than the CPU external frequency.
2. Why does memory Asynchronization exist?
In earlier computer systems, CPU was dominant in the combination of memory and CPU. That is to say, when the CPU clock speed is 100 MHz, the memory frequency can only be 100 MHz, memory usage depends entirely on the CPU. With the rapid development of CPU technology, the frequency of CPU continues to increase, which causes users to upgrade the CPU as well as the memory, without doubt increasing the upgrade cost, this situation is not changed until the 694x chipset of via is released, and the memory and CPU external frequency can finally be run asynchronously.
Of course, This asynchronous running technology is not completely out of the limitations of CPU external frequency, but adopts the "± 33mhz" solution. That is to say, when the P3 processor runs at MHz, the memory can run at either MHz or MHz asynchronously. When the memory runs at a MHz frequency, the system will be able to achieve greater performance improvement. At that time, it was definitely a Leading memory technology. As for the current memory asynchronous technology, it has developed to a more advanced stage. The Asynchronous Operation of memory and CPU external frequency can even be set at a ratio of or.
In theory, with a higher memory frequency, you can get a larger data bandwidth, which will be of great help to improve the system performance. In addition, the memory asynchronous technology can be more flexible with memory. As the CPU external frequency continues to improve, the memory asynchronous technology can help upgrade users to save money for memory replacement. Asynchronous computing memory difference
With the development of technology, the asynchronous memory computing method has greatly changed. We can no longer use a simple "± 33mhz" to calculate the relationship between the memory and the CPU external frequency. At this stage, the memory asynchronous technology is divided into two camps: AMD and Intel. When using the memory asynchronous technology, especially the overclock, the calculation method is completely different.
For intel, the P4 520 processor is used as an example. With the asynchronous memory technology, although the external CPU frequency runs at 200 MHz, the memory can run at 166 MHz. When the CPU overclock reaches 300 MHz, the memory frequency must also be increased based on the proportion of the memory asynchronous technology. Now we have a formula of 200/166 = 300/X. According to the calculation formula, the memory running frequency X equals 250 MHz. To overclock P4 520 to MHz, ddr500 memory is required.
If the same situation occurs on the amd platform, the calculated results will be quite different. In particular, the memory Asynchronization technology on the K8 platform is not directly related to the CPU external frequency, because the CPU used on the K8 platform is integrated with the memory controller and only the CPU clock speed, that is, the most direct relationship between frequency X frequency doubling. Here we take sempron64 2500 + as an example. The actual frequency of this processor is 1.4 GHz, and the CPU clock speed is MHz external frequency × 7. The memory asynchronous technology can also be used to run at 166 MHz when the CPU external frequency runs at MHz. When the external frequency of the CPU exceeds 300 MHz, the memory frequency also increases.
However, the formula also includes the CPU clock speed, so there is a formula like 200 × 7/166 = 300 × 7/X. Note that the result on the left side of the equation is 8.433, but the memory controller of the K8 processor cannot process decimal places. Therefore, the integer "9" must be used in the calculation formula ". In this way, x = 300x7/9 = 233 MHz, the actual memory operating frequency becomes 233 MHz on the amd platform, and the ddr466 memory can be used together. When the K8 processor enters the E3 and E6 cores, the memory Asynchronization technology of the amd platform becomes richer. As a new memory controller, the K8 processor can even use ddr500 memory at the default frequency.
What is the impact of asynchronous memory on the system?
Theoretically, after the memory asynchronous technology increases the memory frequency, the corresponding data bandwidth will also be significantly improved, and the performance should be enhanced. However, from the test results when asynchronous technology began to appear, the performance improvement of memory asynchronous technology was not particularly obvious. Why? In fact, this is because the asynchronous running mode increases the memory bandwidth, but also increases the memory latency.
For example, if a processor runs at MHz, its clock cycle is 10ns. After the asynchronous memory technology is used, the memory can run at a MHz frequency with a clock cycle of 7.5ns. When the clock cycle of the 7.5 NS period ends, the clock cycle of the 10ns period has not ended. Therefore, the former needs to wait for the latter to complete a period before starting the next cycle, in this way, the memory delay is caused, and the performance loss caused by the delay also directly leads to a decline in the test performance. This situation is particularly serious when the NF2 chipset is developed. The NF2 motherboard can achieve optimal performance even only when the memory is synced. The memory Asynchronization technology even became a "chicken fault" on the amd platform at that time ". However, there is always a turning point. When the external frequency of the processor continues to increase, the memory technology also develops rapidly. Between the high-frequency memory and the CPU external frequency, the asynchronous latency becomes smaller and smaller, and the system performance is improved more and more obviously. This frees the internal frequency from the limitation of the CPU external frequency to a certain extent. The combination of memory and CPU can be more flexible and free, which leaves a lot of controllable space for users and greatly promotes the development of overclocking technology. Of course, for those who upgrade the CPU, you can also leave the previous memory, as long as you enable the memory asynchronous function of the motherboard, you can achieve a stable transitional upgrade.
Editor's note: currently, the asynchronous memory technology is mainly used for overclocking and upgrading. Because the technology maturity has been greatly improved, the system performance will be greatly improved. As long as you do not allow the memory to run in the limit overclocking status, the hardware will not be damaged. AMD memory division has a formula:
Memory frequency = extra frequency X Multiplier/memory Coefficient
Memory coefficient = standard external frequency X Multiplier/memory frequency setting value
The value of the memory coefficient is rounded up. For example, if the value is 5.1, the value is rounded up to 6.
For example, AM2 X-Dragon 3000 +, standard external frequency = 200, frequency doubling = 9, and DDR2 667 memory is inserted, so before the overclock, the memory coefficient = 200x9/333 = 5.4 (take 6), and the memory operating frequency = 200x9/6 = 300
When the overclock reaches 260 and the BIOS sets the memory to DDR2 533, the memory coefficient = 200x9/266 = 6.7 (take 7 ), memory operating frequency = 260x9/7 = 334
We usually talk about 533 of memory, 667 of memory or 800 of memory. The so-called 533,667 and 800 are the memory frequency. The memory clock speed is the same as the CPU clock speed. It is used to indicate the memory speed. It represents the maximum operating frequency that the memory can achieve. The memory clock speed is measured in MHz (MHz. The higher the memory frequency, the faster the memory can reach. The memory clock speed determines the maximum operating frequency of the memory. The memory frequency is generally the frequency of its own clock, that is to say, a memory of 800, its working frequency is 400. The operating frequency of the memory and the external frequency of the CPU are greatly affected. When the memory operating frequency is different from the external frequency of the CPU, an asynchronous working mode is formed, in the correct comparison value, asynchronous operations are faster than synchronization operations. Therefore, after overclock operations, you must check whether your CPU overclock is greater than the memory operating frequency. If it is too high, you need to adjust the comparison value, therefore, the memory frequency has a great impact on the overclocking.
The 357.2mhz in the figure indicates the memory operating frequency.
If the CPU frequency reaches 400, your DDR2-800 slip can work in sync. If you are a DDR2-667, it will be asynchronous, less than the CPU out-of-band to use.
Similarly, if the external frequency of the CPU is 200, you can use 667 to asynchronously press
1 to 1. 66. This is faster than synchronizing data.