DRAM

Every computer needs a place to briefly store data while it is working. This place is called working memory (RAM). DRAM is the technology with which this working memory is almost always built. The abbreviation stands for Dynamic Random Access Memory. "Random access" means: the computer can immediately access any location in memory without first working its way through other locations. "Dynamic" means that the stored data fades away by itself and must be constantly refreshed. If you switch off the power, everything is gone.

Why no computer runs without DRAM

A processor calculates incredibly fast, but has almost no room for data itself. It needs a temporary store it can draw on. An SSD or hard drive would be far too slow for this. DRAM sits exactly in this gap: faster than any permanent storage, cheaper than the tiny memory built directly into the processor.

How much DRAM a device has often determines its speed more than the processor's power does. When working memory is full, the operating system has to offload data to the hard drive. You notice this immediately: programs start sluggishly, switching between windows stutters. That's why almost every notebook's spec sheet lists something like "16 GB RAM".

In the AI industry, DRAM has become an economic bottleneck. Large language models must fit entirely into the memory of graphics cards, otherwise they don't run. A special, very fast form called HBM is stacked directly next to the compute chip. Demand for it is so great that memory chip prices now move the entire quarterly results of companies like Samsung, SK Hynix, or Micron.

One capacitor per bit

A DRAM cell stores exactly one bit, that is, a zero or a one. It consists of only two components: a tiny capacitor and a transistor acting as a switch. The capacitor holds a small electrical charge. If it is charged, that counts as a one; if it is empty, as a zero. This extreme simplicity is the reason billions of such cells can be packed onto a chip the size of a thumbnail.

The problem: the capacitor leaks. Its charge drains away within milliseconds. That's why the chip constantly reads out all cells and immediately writes them back. This refreshing is called refresh and happens thousands of times per second, completely unnoticed by the user. However, it does consume power, even when the computer isn't doing anything at all.

Its counterpart is called SRAM. There, a cell consists of six transistors and holds its state without refreshing. SRAM is significantly faster, but requires many times the area per bit and is correspondingly expensive. It is only used to build the small caches within the processor. Incidentally, a common misconception is that DRAM is a replacement for a hard drive: both store data, but only the hard drive retains it after the power is switched off.

From DDR5 to the memory crisis in the news

In everyday life, DRAM is usually encountered under the abbreviation RAM. When buying a notebook, you'll see something like "16 GB DDR5". DDR5 is the current generation of the connection technology with which the chips talk to the processor. In smartphones, the power-saving variant is called LPDDR, where LP stands for "low power".

DRAM regularly appears in business news because the market fluctuates strongly. Almost the entire world production comes from three manufacturers. If they all build new factories at the same time, prices collapse. If demand picks up, for example due to the AI boom, prices double within months. Such cycles determine billions in profits or losses.

Anyone who tinkers with their own computer also encounters DRAM. Additional memory modules are the cheapest upgrade of all for many desktop PCs. In modern notebooks and smartphones, however, the memory is often soldered in permanently. Whatever you choose at purchase then remains for the entire lifetime of the device.