Quote Originally Posted by ugotd8 View Post
Interesting though, when running linpack with AVX enabled I am now getting the huge GFlops numbers, but each thread is only at %50 load per task manager. Load wattage says it's ~115W on the proc.
Technically there are only 4 cores, but with HT on, load is calculated in a rather strange way. Load is calculated by how much the front end (fetch/decode) is being taxed because typically instructions coming in is equal to how loaded the entire chain is.

HT takes advantage of the fact that most applications can use every execution unit the chip has by faking another core and fetching instructions from another thread. However, linpack is designed to max out the execution units with each thread. This means you have two threads competing for the exact same resources so you can only serve each thread at 50% utilization, but in reality your core is maxed out because pretty much everything is being put to use.