The appearance of the 24 and 48 GB DIMMs and SO-DIMMs brutally put an end to a tradition, that of the configurations of RAM in powers of two. However, this tradition has already come to an end, since such amounts do not correspond to the classic evolution of this key component. What is this change in trend due to?
There was a time when RAM was a precious and scarce commodity, when applications had to make sure to use what was right and necessary at all times, and when any increase in capacity was seen as manna from heaven. Today, these limitations have disappeared and system memory exceeds 95% of requirements even in the most modest systems.
RAM has always evolved in powers of two.
It should be borne in mind that the cells of the RAM memory are distributed in matrices and therefore by cells and columns of bit cells. Although in truth today what a few decades ago was a single bit, today it is many bytes. Let’s not forget that current DDR memories can communicate up to 64 bits of information, or 8 bytes, per memory channel and simultaneously.
So with Moore’s Law, each node jump causes the transistor to decrease by 30%, so it goes from bar 10 to bar 7. So if we scale a memory, we can see that whenever there has been a major manufacturing process, the norm is to double the RAM storage capacity. All this due to the fact that we can place twice as many transistors in the same space and, therefore, store twice as much information. Thanks to this, RAM always increased in powers of two, however, this ended with the appearance of new memory modules.
Why do new memories use such strange configurations?
The appearance of 24 and 48 GB memories has completely broken a tradition that has lasted for years, however, there are several reasons why this happens and the first of them is related to the fact that RAM usually uses much less advanced manufacturing nodes than processors and the problem they faced when adopting 16nm nodes that they are now facing. This problem is easy to explain, although the density increases in the same way, the cost per mm² is not the same, but increased at each node, that is:
- We can install more transistors per area with each new jump in the manufacturing process.
- Regardless of capacity, the cost per area has increased, so a chip of the same size is more expensive.
Thus, the manufacturers, instead of opting for the rise in the price of RAM, chose another path which was to keep them within the margins, even if it meant reducing the storage capacity. In any case, the days when the amount of memory was a problem due to lack of memory are over, today it is the opposite due to the lack of optimization of many applications in its use. But that’s what happens when something is in excess.