What is L2 Cache?

Level 2 or L2 cache is part of a multi-level storage strategy for improving computer performance. The present model uses up to three levels of cache, termed L1, L2 and L3, each bridging the gap between the very fast computer processing unit (CPU) and the much slower random access memory (RAM). While the design is evolving, L1 cache is most often built into the CPU, while L2 cache has typically been built into the motherboard (along with L3 cache, when present). However, some CPUs now incorporate L2 cache as well as L1 cache, and a few even incorporate L3 cache.

The job of CPU cache is to anticipate data requests, so that when the user clicks on a frequently used program, for example, the instructions required to run that program are at the ready, stored in cache. When this happens, the CPU can process the request without delay, drastically improving computer performance. The CPU will check L1 cache first, followed by L2 and L3 cache. If it finds the needed bits of data, this is a cache hit, but if the cache doesn’t anticipate the request, the CPU gets a cache miss, and the data must be pulled from slower RAM or the hard drive which is slower still.

Since it is the job of CPU cache to hold bits of data, you might wonder why there is more than one level of cache. Why have L2 cache at all, much less L3, when you can just make L1 cache bigger?

The answer is that the larger the cache, the longer the latency. Small caches are faster than large caches. To optimize overall performance, the best result is obtained by having the smallest, fastest cache most immediate to the CPU itself, followed by a slightly larger pool of L2 cache, and an even larger pool of L3 cache. The idea is to keep the most frequently used instructions in L1, with L2 cache holding the next most likely needed bits of data, and L3 following suit. If the CPU needs to process a request that isn’t present in L1 cache, it can quickly check L2 cache, then L3.

Cache design is a key strategy in the highly competitive microprocessor market, as it is directly responsible for improved CPU and system performance. Multi-level cache is made from more expensive static RAM (SRAM) chips versus cheaper dynamic RAM (DRAM) chips. DRAM and synchronous DRAM (SDRAM) chips are what we normally refer to simply as RAM. SRAM and SDRAM chips should not be confused.

When looking at new computers check out the amounts of L1, L2 and L3 cache. All else being equal, a system with more CPU cache will perform better, and synchronous cache is faster than asynchronous.