Grimes,
Executable x86 code is a series of instructions defined by x86 standard. These instructions are many and they are complex, and the CPU and place a limit to how fast the CPU can run, so they came up with a much simpler instruction set with fewer instructions defined, but these instructions can run GHz speeds.
Decoder is placed between these 2 worlds that translates from external x86 instruction set to internal instruction set.
Executable code often runs in loops, where same instructions are executed and re-executed. Each time through the loop, decode has to decode from one world to another.
Trace cache is like a reservoir of instructions that have already been decoded as program ran through them the first time, and there is a likelyhood that they will be executed again. If they are going to be executed again, instead of re-decoding them, they will be executed from this reservoir called trace cache.
Intel has this feature in Pentium 4, AMD does not currently have it implemented in any CPUs, but it was mentioned as a possible enhancement for the next core - K9.
Joe
News 
Market Data 
Discover 
AI
