Amd Officially Benches Bulldozer

[informal]

I doubt that handbrake scales perfectly with cores.I searched for some tests on 2500K vs 2600K and 980x in handbrake .The results are all over the place. Somewhere 2600K is 10% faster,somewhere 20%. Look here : 980x vs i7 975,50% more cores and same clocks. 980x is 23% faster. 2500K vs 2600K,same number of cores but a bit higher clocks and SMT on for 2600K : 23% better performance. Here SMT adds as much as 50% more cores do in 980x vs 975 case.Strange. In case of AMD ,we have 43% better score (1090T vs 955BE) for 50% more cores. Pretty consistent speedup in case of K10.

edit: I have found SB-E vs 2600K results. 50% more cores,more cache,similar clocks and it gets 20% speedup. Different codecs though. It looks like the SMT already extracts maximum ILP from the code so additional cores scale less than linear when SMT is on. It would be great if we had SB-E with SMT off so we can compare it to 2500K. I suppose the scaling would be near 50% in that case.

As for "gaming" benchmark,it is really GPU bound at that point. Probably Thuban and 2600K would score around the same 80fps mark.