How does external cache help?

[Win_98]

I have one meg and I see no difference
when disabled. Strange.
I benchmarked still same score too.

[Steve R Jones]

Memory Caches
A memory cache, or CPU cache, is a memory bank that bridges main memory and the CPU. It is faster than main memory and allows instructions to be excecuted and data to be read at higher speed. Instructions and data are transferred to the cache in blocks, using some kind of look-ahead algorithm. The more sequential the instructions and data, the more the cache improves performance.

A level 1 (L1) cache is a memory bank built into or packaged within the chip. A level 2 cache (L2) is a group of memory chips on the motherboard. Increasing a Level 2 cache may speed up some applications and amount to nothing on others. Both types are used together. In PCs, the cache is static RAM (SRAM), while main memory is dynamic RAM (DRAM).

[alan]

if you disable your L2 cache and you see no difference in bench's then i would hazzard a guess that it aint working! because you will notice a difference either with L1 abled/disabled, L2 abled/disabled or any combination there of.
alan

[Win_98]

I have another question.
so how is celeron running at full speed of
128K cache better then a super7 board with cache that has 1meg?
so if more cache is better,
then is 2meg twice as fast as 1meg.

[SDT]

These are two different caches. The 128k (L1)cache is on the processor and is very very fast. The 1meg (L2)is on the mainboard and is not as fast as the 128k on the processor. The "on the processor" cache is preferred as it is as fast as the processor (faster than the cache on the mainboard).
Memory is much slower than either cache.

More cache is better. More "on chip" cache is better than more cache on the mainboard because of the speed of the cache and the short data path to the cpu for both data and instructions.

There is a logical limit to cache size as the "look ahead" algorythm becomes less and less accurate (effective) as the size gets larger. In other words if you are loading 10% of the instructions available to the processor into cache then the processor may have to go back to memory to get the next instruction. If you are loading 75% of the instructions into cache then the odds are greatly reduced that the processor will have to go back to memory for the next instruction. Bare in mind the there are a lot of instructions that are rarely used.
The same logic applies to the data moving through the cache. If you only processed data sequentially then you would only need room in the cache for the next 4 (I picked a random number. The real number is based on cpu speed and main memory latency) cpu cycles (assuming that the cpu can process the data in one cycle ...not necessairily so). This would allow the memory to feed the cache faster then the cpu could empty it so a larger cache would not make things faster. The same logic applies to non-sequential data & instructions... your accuracy at guessing the next block of data the the cpu will need reaches a limit and beyond that you would need to retrieve masses of data to get a hit in cache. That kind of mass retrieve process is inefficient and expensive so the developers don't do it. Therefore there is a logical limit to cache size beyond which the system just thrashes and you see very little performance gain.