The 40% was just a general statement. I did point out that the 3770k was (mostly) a superior chip, but with the caveat that in some areas the 8350 beats it as well.
For example, C-Ray at Openbenchmarking.org.
http://openbenchmarking.org/result/1210227-RA-AMDFX835085
In C-Ray, the FX-8350 completes the test in 23.34s The i7-3770 takes 33.05s i.e., the 3770k is 30% slower.
It also beats it on the Linux compile by about 3-4%.
My point was not to say that the 8350 is faster than the 3770 though; the 3770 is superior in 95% of the cases.
On the i5 front though, that'st just not true. What bothers me is when I see people saying that an i5-2500 or i5-2400 is better\faster. That's just plain false in the majority of scenarios, but it's a belief fostered by the consumer oriented / marketing driven tech sites that focus on running game benchmarks at 1024x768 with a GTX 690 (and yes I understand their explanation of why they do that, and it's flawed - it winds up presenting a synthetic / unrealistic scenario as if it were meaningful).
Look at that same chart linked to above. Right off the bat, the 8350 smashes the i5-2400s - 100% faster in SMP NAS testing. 60% faster than the i5-2500k. 35% faster than the i5-3470 (noted : 15% slower than the 3770k).
Then we get to John the ripper (DVD ripping) - it beats the i7-3770k by almost 20%, more than twice as fast as the i5-2500k and nearly twice the speed of the i5-3470.
Look at the Linux compile time. 8350 = 82s. I5-2500k =116.14s. i5-3470 = 114.25. The 8350 is almost 40% faster than the i5-3470 here.
You can also go the other way, AMD loses in tests which involve FPU quite badly. But the performance is not 'less' overall - it's just different. People don't like different, so they tend to say it's imbalanced, when they should say - superior in some ways, inferior in others.
What it really comes down to is this : If you have a high load, heavily threaded application that does not do a lot of FPU operations (note: SSE is not FPU, which is why AMD does well on high load heavily threaded encoding / ray tracing), AMD wins vs the i5's. If you multitask a lot with applications that fit that description - AMD wins.
In my particular case, I run multiple VMs with Win 7 & Delphi / VS 2010, W2k8 + SQL Server 2008, and Linux/gcc, on top of my host OS. In that scenario, my iMac with an i5-2500S running VMWare Fusion doesn't hold a candle to my FX-8320 running VMware Workstation. And that's using the same exact VM files. Have Time Machine kick off on a USB drive in the background on the iMac and watch the cursor jump around (firewire and TB don't have that effect btw, no cpu interference there). On the AMD - not so, backups don't interfere.
Anyway, I'm done. YMMV