Quote:
Originally Posted by
Asterox 
That's quite enough one test in programs that use only one CPU core, and one Multithread test both processors at the same CPU frequency.
No, it definitely isn't. Performance in different applications can vary wildly based on a variety of factors, such as how frequently a cache miss occurs. Cache misses in particular are what I'm getting at — if code executed with very few relative misses, the branch prediction improvements of Piledriver wouldn't translate to much performance improvement.
There are perhaps hundreds of factors that play into how well a CPU performs in one cycle — Piledriver only improves a handful of them. If the bottleneck lies in a region where Piledriver didn't improve over Bulldozer, say the L1 and L2 cache latencies, then you're not going to see a noteworthy performance improvement.
Also, it turns out that my memory has failed me — Trinity was more than 15% faster clock for clock in two applications in the THG review, not one. Still, Piledriver is definitely not a flat 15% improvement over Bulldozer per clock. There's also the potential that the cores used in Trinity may not be the same ones showing up in Vishera.
Quote:
Originally Posted by
erunion
Why do you people always say that? Its so amazingly, incredibly false.
It's just him, generally. The rest of the AMD shills generally keep their circlejerk to themselves down in the CPU section.
