Overviewing the Factors, Cache

First Things First Though

Of course, we have to set down some ground rules first. Let's clear up some of the anomalies going in right now before we discuss theory. The most important factor we haven't taken into consideration yet is chip cache - specifically level 2, or L2 as it's commonly called.

There are three types of processors out now. Older processors don't come with L2 cache. They have the L2 cache on the motherboard, running at system bus speeds (K6 + K6-2, old Pentiums). The cache is damn slow -100MHz at fastest. The speed of L2 cache on the motherboard is a severe system bottleneck. That's why you get a dramatic performance boost when you increase the system bus speed from 66MHz to 83MHz or 100MHz. Both Intel and AMD have moved the L2 cache closer to the CPU to overcome this bottleneck.

More recent processors have discrete (separate) L2 cache chips running along side the processor on the same package, but not on the same die (Pentiums IIs, non-Coppermine P3s, and Athlons). Currently the trend is moving towards chips that have on-die cache running at full processor speed. Cutting edge processors such as the Coppermine P3s, Celerons, K6-IIIs, and next generation Athlons all feature on-die cache.

Cache, baby!

So what's the deal? Well, of all the mainstream processors that keep the L2 cache on the motherboard right now, the K6-2 is the only one still in production. The L2 cache doesn't have to run at the same speed as the processor. L2 cache on the motherboard is designed to run at the system bus speed, which is usually either 66 or 100MHz. The L2 cache speeds will be far below the processor's speed, which is typically 200MHz or higher these days.

The only way this motherboard cache can limit overclocking potential is if you're trying to speed the system up by increasing bus speed. Clock speed refers to the number of MHz (Mega Hertz) or "cycles" per second your processor is running at. 1MHz = 1,000,000 cycles per second. 1GHz = 1,000,000,000 cycles per second. Computer instructions take a certain number of cycles to execute. Some need one, others might need as many as 14. Also, most computers nowadays can do multiple instructions per second, speeding them up significantly.