Actually, the base MBA 2012 model is 47% to 24% (Turbo Boost taken into account) slower than the Pro model. The max upgrade of the Air model is 45% to 12.5% (again, Turbo Boost taken into account) slower than the i7 Pro model once again, so it's actually quite slow. Turbo Boost does not kick in except for short intervals because the thermal profile of the MacBook Air is not adequate to sustain the load for a longer period of time.
Not to mention Intel HD 4000 in the air is 80% slower than Intel HD 4000 in the Pro when it doesn't get a boost from Turbo.
So the Pro machine is still worth it if you want a faster Integrated GPU and a higher resolution screen. Anyone who cares enough about the Pro machine would take their time to research and look things up.
It's likely the Thunderbolt ports were routed through GT 650M in the rMBP, so when you force Intel HD 4000, it can't reach them to display anything.
I'm actually thinking Apple also requested Intel to increase the base clocks of the Intel HD 4000 in the rMBP 13"... judging from the fact that they pushed more VRAM to it.
Say... the ULV models have 350MHz base clocks, the regular volt models have 650MHz base clocks.
I'd say it's not unlikely to see a 800MHz base clocks in the modified cores... but then... I may just be grabbing at straws, and Apple really just increased VRAM.
So here are the magic 22nm chips were talking about, theyre very similar except for frequency and thermals:
Base 2012 MB Air 13
i5-3427U 1.8 GHZ: 2.6 GHz 2 core turbo, 2.8 GHz max single core
HD 4000: 350 MHz 1150 MHz max
Base 2012 MB Pro 13
i5-3210M 2.5: 2.9 GHz 2 core turbo, 3.1 GHz max single core
HD 4000: 650 MHz 1100 MHz max
Benchmark choice shouldnt really matter here because these CPUs are identical, just different clocks and thermal environment. According to Notebookcheck, the Air can finish the multithreaded benchmark wPrime 2.0 1024m in 664 seconds while the Pro can do it in 575 seconds. Thats 87% of the pros score while running for 11 minutes with the CPU pegged.
They also tested for throttling by looping Cinebench R11.5 multithreaded 50 times and did not see a drop in score. The MBA hits 91% of the MBPs performance even while doing this long render.
Extended multithreaded workloads don't seem to bring the Air off it's turbo clocks and it obviously does very well vs. the Pro using a single core where the thermal disadvantage matters less. Base clocks matter deceptively little; If the CPU has extra thermal headroom it will run at Turbo clocks and if it is too hot it will throttle right down bellow base clock if necessary. The CPU goes to some some super low clock speed at idle so I think base clock speed can be regarded as a number mostly for marketing purposes.
The Air also scores within 1% of the Pro in every 3DMark bench. The Air has the slightly higher max HD 4000 turbo and it would seem that Turbo clocks are being used for it to match the Pro. To stay this cool the air puts out noticeably more noise but it is amazing that it runs without significant throttling. I bet the new 13" Pro with dual fans is amazingly silent with only a 22nm dual core to cool. It will be interesting to see what Intel did for Apple with the IGP in this new 13 Pro. With all that cooling the penalty for turning up HD 4000's clock speed might be worth it to make using that resolution smoother.
Source:
http://www.notebookcheck.net/Review-Apple-MacBook-Air-13-Mid-2012-Subnotebook.80041.0.html
http://www.notebookcheck.net/Review-Apple-MacBook-Pro-13-2-5-GHz-Mid-2012-Notebook.79640.0.html