What is L2 Cache?

I searched this forum and the internet for the answer but the wiki article just confused me.

So I was hoping you guys can help a computer newb like me out and give me a better understanding of what it is. .

 

Cache is a smaller bit of memory that sits between the processor and the system memory (RAM) that runs much faster than the RAM. The CPU can access cache memory much faster than RAM, which in turn is much faster than the hard drive. But each one is smaller than the one before it, so it stores less data. Make sense?

Analogy time: A semi can haul a lot of stuff, but it's not too fast. It's like your hard drive. A sedan can haul much less stuff, but it's a lot faster than the semi. It's like your RAM. A motorcycle can only hold very few things, but it's very, very fast. It's like the cache.

 

Wow that was a quick reply. Thanks, that makes sense.

So would the L2-Cache difference between a TL-50 and TL-52 make that much of a difference?

 

Basically, it'd depend on what you ask of your processor.. there are lots of things that determine how "fast" a computer is. but yes, given that all other things are identical, a larger cache=better performance.

 

It can make some difference. It really depends on how different programs run. Sometimes they can take advantage of additional cache, sometimes not. I don't know how much the difference in cache size is, but it probably won't make as much of a difference as the clock speed will. It depends on what you need the CPU speed for. Most people don't notice the difference between a mid-grade CPU and a top-end one.

 

L2 cache is a slower but bigger level of cache. L1 cache is quickest to access, but the size of it is usually a few hundred k. L3 cache is even bigger, and even slower.

 

L1 size is 32-64KB, not "a few hundred k".
(If memory serves, Intel's Core CPU's have 32KB L1 cache, and AMD's chips have 64KB.

And yes, cache can make a big difference.
However, a better analogy than Pitabred's (sorry ) might be your desk, versus going to the library. If you need to look something up, you have to go to the library, find the book you need, borrow it, bring it home and finally read the stuff you need.
Now, if you're clever, you don't return it to the library immediately. You put it on your desk, so you can easily reach it the next time you need to look something up in that book. Sooner or later, you'll have filled up your desk. Then you return some of the books you hopefully won't need again soon.

That's pretty much exactly what goes on with the cache. When the CPU requests data, it first looks in the cache, and if it isn't there, it has to order it all the way from the RAM, which takes ages. And once it's got the data, it puts it in the cache, because the odds are very good that it'll be needed again soon.

So the cache size corresponds to the size of your desk. The bigger it is, the more books you can have on it, and the less time you need to waste running to the library and back.

And to take the analogy even further, for some tasks, you might need to look up only a few things in each of several hundred books. Then your desk (cache) won't help you much, because almost as soon as you get a book, you're done with it and won't need it again.

In other cases, you only need one or two books, so if there's room for them on the desk, you only have to go to the library once.

So it depeds a lot on the program, whether or not a larger cache helps performance.

 

And yet, an even better analogy would be using food and cooking:

HDD = supermarket
RAM = basement freezer
L3 = fridge->bottom drawer/rack
L2 = fridge->top drawer/rack
L1 = cutting board
CPU = you're cooking

 

LOL, you guys are analogy crazy

If only someone would invent the equivalent of the internet or wikipedia in Jalf's analogy. Having a practically unlimited L1 cache. Now, that would be awesome

 

That's pretty much what computers with no cache are. (The RAM basically becomes the L1 cache then). And the problem is obviously that such big caches have to work much slower. So we'd then want to add a smaller L0 cache which was, say, 512KB, and while we're at it, we might as well ad a L-1 cache of, say, 64KB. And then we're back where we started.