The Future of Cache

[Deadrewski]

Hey guys, can anyone explain to me why we have so little cache per processor? Would being able to double or triple the amount of cache for a processor greatly increase performance, or would it go largely unnoticed? 

I do research in semiconductor physics, and I want to learn more about processor cache materials to see if I can start up some nanostructure semiconductor research to improve cache costs/size. Anything you know about cache would be really helpful! The wiki page on it is kinda vague.

[zlolslavez]

Theres little because its expensive to add up, and not a shitton is needed.

 

If you thought the volatile storage of RAM was expensive and fast, imagine what Cache Disks would be.

[SSL]

Because SRAM is made from transistors, and is therefore stupidly expensive compared to capacitor based DRAM. In addition, the larger a cache is the slower it is, so the bigger you make it the less benefit you get. It also requires space on the die that could be used for other things; memory access is only part of the performance equation.

[Deadrewski]

Really? The bigger it is, the slower it is? Hmm...well that's no good. What would you suggest would be the next big leap for CPUs, then? I mean, I already do research on nanowire semiconductors (like graphene) that will give us better conductivity, lower temperatures, and smaller transistors. Any other specific CPU related components that could use a mat sci boost? 

Thanks for the comments!