Computer Memory: Problem Never Resolved

Since the first computers, memories have been an headache for designers and manufacturers. PCs needed a processor with a fast attached volatile memory (RAM) for compute and caching, and a slower but much larger non-volatile magnetic memory (hard disk drive) to store, read and write data and files.

The two main forms of RAM are static RAM (SRAM) and dynamic RAM (DRAM). In SRAM, a bit of data is stored using the state of a six transistor memory cell. This form of RAM is more expensive to produce, but is generally faster and requires less power than DRAM and, in modern computers, is often used as cache memory for the CPU. Both of them have two disadvantages: they are volatile (information are lost when power is off) and price per gigabyte is very high.

Remember that the first operating system for IBM PC compatible was named MS-DOS (acronym for Microsoft Disk Operating System). Its main function was to manage the transfer of data between motherboard and HDD. Unfortunately, this later is not really a direct access memory, data being written into blocks of the same fixed size, several blocks being needed for a complete file. Today OS are doing much more but this management continues to be one of their fundamental function.

For decades, there was no other choice than HDD for primary storage for a computer. Price is affordable. Capacity increased regularly even if areal density stagnates these last years as well as main specs (access time and transfer rate). But the mechanic devices are not reliable enough. The industry was obliged to invent complicated and expansive disk arrays, as well as backup software copying data on external magnetic tape and disk drives, to protect the data in case of HDD crashes.

Just one example of technologies trying to find a place in storage: phase-change memory (PCM), with a first patent filed by Stanford R. Ovshinsky from Energy Conversion Devices in 1966, after exploring the properties of chalcogenide glasses as a potential memory technology. Fifty years later, after many tries by big companies (Intel, Samsung, STMicroelectronics, etc.), it never reached the commercialization stage but for optical discs only.

To avoid these OS problems, the dream would be to have one and only one fast reliable non-volatile silicon memory with big capacity and affordable price directly connected the processor. We are far to get that.

Storage is an enormous market. Consequently an incredible amount of companies and researchers have tried since decades to design a substitute of HDDs. Just one appears to be successful: NAND flash invented by Dr. Fujio Masuoka while working for Toshiba around 1980 and commercialized in 1987. As of 2012, there were attempts to use flash memory as the main computer memory but, up to know, it’s too slow.

It is pretty clear non-volatile memories where the data is close to CPU will dominate the market in the future. But which ones?

Today maybe the most promising memory type is 3D XPoint NVRAM, using PCM and recently announced by Intel and Micron, being not flash, not DRAM, but somewhere in between.