Let's imagine you come on time on the fourth night and begin with your standard hot dog. When you come, the waiter, who has gained confidence, has the hot dog ready for you.

"Gee, I'd really prefer a bratwurst right now; I didn't request this hamburger," you comment as you finish your hot dog and he places a hamburger on your plate. Because the waiter estimated incorrectly, you'll have to wait the whole 60 seconds while the kitchen makes your brat this time. This is referred to as a cache miss, which occurs when the cache controller fails to correctly load the cache with the data that the processor requires next. When a cache miss occurs, the system usually waits or, in the instance of a 233MHz Pentium computer, the system throttles down to 16MHz (RAM speed).

The L1 cache on most Intel CPUs, according to Intel, has a 90 percent hit ratio. (Slightly better CPUs, such as the Pentium 4, are available.) This indicates that the cache contains proper data 90% of the time, and as a result, the CPU works at maximum speed (233MHz in this case) 90% of the time. However, the cache controller estimates incorrectly 10% of the time, and the data must be fetched from the much slower main memory, causing the CPU to wait. This effectively reduces the system's speed to that of the RAM, which in this case was 60ns or 16MHz.

The CPU was 14 times quicker than the main memory in this example. In recent systems, memory rates have gone from 16MHz (60ns) to 333MHz (3.0ns) or faster, while CPU speeds have also increased to 3GHz and beyond. Memory is still 7.5 or more times slower than the processor, even in the most modern computers. The difference is made up by the cache.

L1 cache has always been integrated into the processor core, where it runs at the same speed as the core. This, combined with a 90 percent or higher hit ratio, makes the L1 cache essential for system performance.

Download