What is cache memory

Not all computer owners know what cache memory is, although it is actively used by absolutely all manufacturers, both processors and programs. Some users, who have only recently started to master a personal computer, sometimes on thematic forums in the global Network complain about the low speed of their electronic assistants. For example, if between running the shortcut of the office program and the appearance of its window takes 5 seconds, then this is considered a huge time. Or it is argued that those 10-15 seconds that are necessary to boot the operating system from a hard disk on magnetic plates is a waste of time. Surprisingly: just a decade ago, the launch of the program could take almost half a minute, and it was considered quickly. Obviously one thing is that the performance of computers has increased significantly and the cache memory of the processor played a significant role in this.

The RAM modules used in computer technology are based on DRAM ( random access memory) technology. Features of this technology in low cost, high reliability and ... relatively low speed. DRAM was used already ten years ago, however, even slower modifications of it. If at that time access to the memory cells was carried out with a delay of about 200 nanoseconds, now this value has overcome the boundary of 20 ns. It would seem - the speed should be just fantastic! However, in parallel with the improvement of DRAM, the bandwidth of the processor buses also increased, so the overall ratio did not change the way it could. So we came to the question of what a cache-memory is. How can I increase the performance of the computer's memory subsystem ? The answer is obvious - you can replace the morally obsolete DRAM with something more progressive. But the notorious experience of Intel with the expensive Rambus slats suggested that the replacement should not significantly increase the final cost.

Without this limitation, no one would ever have thought about what a cache memory is, because there would not be much sense in such a mechanism. It is enough to replace DRAM with a more advanced SRAM (random access memory) and the problem would be solved. But this would entail a significant increase in value. Therefore, a compromise variant was proposed that turned out to be so successful that it has been used since the days of the first 80286. To increase system performance between relatively slow RAM modules and a high-speed processor, high-speed memory blocks are located. Regarding the number of DRAM cells, their volume is extremely small, ranging from 8 kb (the first level of L1) to tens of megabytes (level L3). A special controller passes a bi-directional data stream through itself and copies them into a fast memory. With the next processor requests to DRAM, the controller checks to see if the required data is "in stock", and if they are found, the transfer to the processor is from the cache. As you can see, the principle of operation is quite simple. Complexities arise in the implementation methods: the developers have to decide which data to duplicate, how to update it, how to improve the efficiency of work, etc. But this is a very voluminous topic, so you can read more about it in specialized sources.

Thus, the answer to the question of what a cache-memory is, can be formulated as follows: the cache memory is a kind of buffer into which the controller writes / reads the data, increasing the performance of the memory subsystem.

And what is a cache for network applications ? The user's browser when browsing the Internet uses the same solution. When you first open any page, many of its elements (images, data) are stored in a folder on the disk, and on subsequent calls, these data are not downloaded from the slow Network, but are substituted from the folder, increasing the speed of surfing. Acceleration is particularly noticeable at low Internet speeds.