I did but their note really relevant hence why I linked a better comparison, the OP is running the 920 at 2.9GHz and the 3770K at 3.9GHz, thats a 1GHz advantage to the newer chip.
However your link does show the SB i7 beating the Bloomfield i7 by a full 15% at the same clock speed which is at the high end of the BF-SB IPC jump (10-15%) so as the BF-IB jump is 15.5%-26.5% if thats also at the high end then IB is 26.5% faster at the same clock speed, factor in that the OP's IB has a 35% clock speed advantage over the BF and it really should be destroying it.
Hell using some basic napkin maths (obv not scientific) if the chip has +26.5% IPC in this specific scenario then its MHz are effectively worth 1.265x what the BF chips are, so a 3.9GHz IB has the effective power of a 4.9GHz BF chip (obviously a very rough calc and only applies to this specific scenario).
You still cannot even explain,why the time has gone down from 180 minutes to 50 minutes. None of your calculations account for that,as it makes the IB Core i7 a couple of times faster. So no it is not destroying especially since Turbo on first generation Core i7 CPUs was extremely conservative. That is why socket 1155 Core i7 CPUs could match them as they tended to have more aggressive Turbo.
Moreover you are on ignoring this page
you linked to:
http://media.bestofmicro.com/C/C/334812/original/lame.png
http://media.bestofmicro.com/C/7/334807/original/itunes.png
Look at the first 4 CPUs,one IB CPU ,two SB CPUs and a single SB-E CPU. There is hardly an IPC increase,especially when you consider the
margin of error too in these measurements.
The Core i7 3930K has less L3 cache per core than a Core i7 3820 and even then,it is more likely TH had an example which was not running at max Turbo,ie,probably a high VID example or insufficient cooling.
Even using
your best case scenario, that would 60% faster overall,which is not the same as going from 180 minutes to 50 minutes.
The only such massive performance improvements are seen in software which can use things like AVX and AES-NI,ie,new instruction sets.
Certainly not a scientific test - but a couple of posters have said that it is a single threaded process... looking at the OP screenshots, the CPU is maxed and running at least 6 Instances of the MP3 Convert...
Therefore the gains are going to be fairly pronounced e.g. 8 Threads (4 Physical / 4 Logical) all 1Ghz faster than the old I7.
However....
Based on rough calculations... Old I7 would do 5.55 MP3s/ per thread per minute (8000/180/8), 3770k would be 20 MP3s/per thread per minute (8000/50/8)... almost a 4 Increase in throughput?
It could be something as simple as a new version of Foobar was installed on the new rig, and is configured differently to make better use of SMP.....
You get it. Where is this massive increase in speed coming from?? If it is not from the usage of new instruction sets,it does make me think what is happening here.
I agree with the software version. If you look at HandBrake for example,the changes between the last three versions have been huge and significantly changed the way it interacts with CPUs.
This is also an important point when it comes to reviews. Using old versions of software can skew results easily and have no relevance to actual usage patterns.
