There is one beef I have with all these performance averages. They always factor in both the single and multi-threaded results from a multi-threaded programs like Cinebench and PovRay. These programs are optimized for multi-threaded and will be used as such. Thus they should be factored in as such. Including the single thread results in equal measures is ridiculous!
This cancels out some benchmarks where multi-core processors are on a home ground in favor of more single thread oriented CPUs. The same should be said about H.264 and X264 encoding benchmarks where the 1st and 2nd pass are weighted equally despite the 2nd pass having much more impact due to the fact that it takes much longer time than the 1st pass. F.ex if you have 10.000 frames of video it will take FX-8350 around 170s to complete Pass 1 but 660s for pass 2 for a total of 830s according to Xbitlabs results. i5-3570 will take around 180s for pass 1 and 830s for pass 2 for a total of 1010s.
If we weigh both parts equally the FX-8350 is 5,5% faster in pass 1 and 20,5% in pass 2 or 13% for both. If we however correct in terms of real world application the difference for both passes taken for the same video will turn out to be around 18%. This is statistically significant difference and in many cases it's much greater.
For example if we take results from Hardware Info: http://uk.hardware.info/r...eg-to-x264-video-encoding
Equal weight: 5,8% in favor of FX-8350
Real weight: 13,4% in favor of FX-8350
Add to this SysMark and PCMark which are arbitrarily weighted and these weights not being made known to us but seem to favor Intel disproportionately. As a result we have an average that is a wet dream of any Intel marketing person but does a disservice to consumers as it relaxes the pressure on Intel and strips AMD of it's advantages.
I am not knocking on Intel, they do great processors. But reviewers need to make an effort in being more evenhanded. I get the feeling they are still angry about their over-hyped expectations being Bulldozed over by a power hungry beast that didn't perform to expectation but still was a solid 10% improvement on average over X6 1100T despite losing in few titles and benchmarks to it's older brother.
It was a radical change and did not meet the unrealistic expectations we set for AMD. Many people fail to grasp just how much of an engineering brilliance the Intel Core architecture is. Especially the Sandy Bridge which made great strides in terms of power consumption. Compare the Bulldozer to the 1st generation of 32nm Core i7 965 and they trade blows at a similar power consumption level. The Sandy Bridge completely changed that and is probably the best revamp of any processor architecture ever made. Expecting AMD to copy that with it's limited resources is a delusional ignorance of the highest degree.
The fact that they remain within a spitting distance of Intel (about 15% below i7-3770) is incredible. I know few here remember what happened to the other competitors of Intel, but consider the last Cyrix processor "Samuel" with maybe 2/3 of the performance of the Duron and Celeron budget CPUs of AMD and Intel. That is a truly big difference. In reality 15% is almost not recognizable for a user.
Currently with the release of FX-8350 AMD is closing in on Intel in terms of Performance. I was personally hoping for better luck in reducing power consumption but all in all it's a nice step forward. I will be getting one in my own platform and playing with it soon and I am looking forward to it. I am also hoping for AMD to score big with Steamroller next year as it will address the two places where this architecture suffers the most at the moment, i.e. power efficiency and single thread execution. Depending on how well Intel deliver on Haswell it could be an interesting fight.