The AMD Thoroughbred is the processor core that took AMD to the .13 micron transistor size. The only differences between the Thoroughbred and the Barton is that the Barton is basically just a Thoroughbred with  256k more L2 cache, giving it 512k L2 cache total, and the Barton has a bigger core.  This however makes the CPU more expensive to manufacture. With this in mind, will the Barton be worth the higher price tag?  Well since the Barton 2500+ is at the $100 level it may be worth it, but being that the only benefit is the cores more L2 cache would it really be worth the upgrade.  Let's look at what the cache can give us.

Processor cache is located in the processors core and provides the processor a place to fetch important data without having to go to main memory.  If the processor had to fetch everything from main memory, your system's performance would be greatly affected because it would have to travel the entire pipeline every time it needed data.  In some cases, it can take up to 100 clock cycles to travel though the considerably slower FSB, memory controller, and back.  The reason that processors use a multi level cache is because it takes longer to populate larger caches so having a small very low latency cache (L1) and having a larger higher latency cache (L2) creates that best performance for today's CPU's.

There are two ways cache's work, temporal locality and spatial locality.  Temporal locality means that the data being used is likely to be accessed again in the near future and spatial locality means that data near what is being used is likely to be accessed in the near future.  Let's look at how both of these principles work. We will write a program that takes advantage of both of these principles and look at what part of the program exhibits each principle. In the program below it will count down from 1000 to 0 in a loop and then display a message. (Line numbers are just for reference.)

Like we mentioned before temporal locality means that the data being used is likely to be accessed again in the near future. In the example above, lines 8 through 12 would be used over and over again until the loop completes.  So that data would exhibit temporal locality.  

Like we mentioned above spatial locality means that the data near what is being used is likely to be accessed in the near future. In line 14 of the program above, it is waiting to display the final message after the loop completes. So this exhibits Spatial locality

Now why are these principles important?  The best answer to that is only programs that use these principles will have any improvement in a larger cache. Some programs use one or both of these principles. For example, office applications reuse a lot of data while movie encoding applications just process data and move on. This means that office applications would be more temporal locality and encoding applications would be more spatial locality.

The main focus of this review is to see how well your multimedia applications will benefit from the larger cache. Let's get to testing and see what we come up with.