Since the 11th generation of Core processors, Intel has followed the pace of AMD Ryzen and introduced a "memory crossover mechanism". The latest 12th generation Core platform supports both DDR5 memory and DDR4 memory, and for the D4 part, the memory crossover mechanism is still retained. When using "high-frequency memory", or "memory overclocking", if you do not understand this mechanism in depth, you cannot get the corresponding high frequency, or increase the probability of overclocking failure.
What is Memory Crossover?
The frequency of memory operation is limited by the memory controller. The memory controller is an area of cpu control memory that is responsible for controlling the work of memory. Modern processor cache structure is complex, in order to improve the read and write performance of external access memory, it can be said that it is a technical problem that affects the whole body. So AMD and Intel have all gone down a similar path: adding a crossover. The original 1:1 synchronous crossover has been changed to an asynchronous crossover mechanism that automatically switches to 1:2 at high frequencies.
What is the specific operating mode of memory crossover?
In CPU-Z, you can see the difference between the Gea1 and Gear2 modes
Under the memory crossover mechanism, there are "Gear1 mode" and "Gear2 mode". Gear1 mode is 1:1, the ratio of memory controller frequency and memory operating frequency is 1:1, the two work synchronously, memory latency is low, and the performance is maximized. The Gear2 mode is 1:2, and the frequency of the memory controller is only half the memory frequency, which can reduce the pressure on the memory controller and make it easier for the memory to get higher frequencies.
Why is 3600MHz the tipping point for crossover?
First of all, we need to know a concept, how is the frequency of memory structured? Memory frequency = Ratio multiple× base frequency (100MHz) × reference coefficient (Reference Clock, there are 1.0 and 1.33).
The basic frequency is the frequency of the coordination of the entire computer platform, and like the long jump rope, everyone must jump together to not be tripped. The various accessories of the computer platform must also ensure that a same frequency is synchronized, usually the basic frequency in the computer is 100MHz. (Z Series chipsets can be fine-tuned)
Ratio multiples are somewhat similar to CPU multipliers, in a skipping rope cycle, you jump several times by yourself, others are still normal rates, as long as the timing is right, you will not trip. (In a computer, the right time is to ensure that it is an integer multiple.) ）
The reference coefficient is an asynchronous operation technique that allows memory to operate at 1.33 times the fundamental frequency, maintaining synchronous operation with other devices, and can also work normally at 1.0 and complete synchronization with other devices. (K series CPU two free switching, non-K series CPU only support 1.33)
Take the 11th generation Core processor, which is equipped with a memory controller constitution that can only be multiplied by 27-30 times. Plate factories generally adopt a conservative strategy, and the default is 27 times is the upper limit. This means that up to 27 × 100MHz × 1.33 = 3600MHz can be used. Therefore, in memory Gea1 mode, the operating frequency is up to 3600MHz.
The mechanism of the 12th generation Core processor is the same, at this time the smart you will definitely come to ask, is this not the 11th and 12th generation Core processor DDR4 memory frequency can only support 3600MHz? What about 4000MHz high frequency or higher? At this time, the Gear 2 mode has a place to play.
In Gear2 mode, which high frequencies cannot be used?
First of all, the situation is somewhat complicated.
The reference coefficient of the K series CPU in the Intel camp can be selected as 1.33 and 1.0, and the non-K is only 1.33; the Z series motherboard can move the base frequency because of the Clock Gen, and most of the B series motherboards do not have Lock Gen and cannot move the base frequency; so there are four combinations of [K-U with Z board, non-K-U with Z board, K-U with B board, non-K-U with B board].
With this formula Ratio x 2 x 100MHz (fixed) x 1.33 (non-K processors can only be 1.33) = (the frequency that can be exceeded), and there is a premise that Ratio must be an integer.
There are only a few integers, while the final frequency exceeds 3600MHz, we will substitute the integers into the formula one by one, and it is easy to get the frequency distribution that meets the requirements in gear2 mode, as shown in the following figure. Like 3700, 3900, 4100, 4300, 4900, 4933MHz... None of the frequencies can be operated.
11th Generation Core with and without K processor, with Z590 motherboard, can support the memory frequency (gray means not supported).
11th Generation Core with and without K processor, with B560 motherboard, can support the memory frequency (gray means not supported).
Gear2 will simply affect the memory controller, will not affect the entire CPU, with the memory frequency increases, latency and bandwidth performance are significantly climbed. The existence of Gear2 mode can make the high-frequency memory that cannot be supported easier to run, and will not be limited by the constitution and performance of the memory controller.
However, we also need to make a point clear, memory overclocking wants to exceed to a higher frequency, in addition to figuring out the new memory division mechanism of the new platform, it also requires that the memory particles you choose are good enough, otherwise overclocking will also be difficult.