>most tests will run entirely in cache, placing no load at all on the memory controller, Exactly. Like CPU-Z. Yet unlike CPU-Z, it is not an extremely biased

>most tests will run entirely in cache, placing no load at all on the memory controller, Exactly. Like CPU-Z. Yet unlike CPU-Z, it is not an extremely biased sse microbenchmark, but does instead provide decent coverage of a CPU's functionality. The benchmarks within emBench include common code like gzip or jpeg crunch/decrunch routines, which are frequent stress points in real world applications.