I am well aware of what the Anandtech review says guys. I read all of their reviews and have a lot of respect for their work, but I feel that they dropped the ball on this one. If you read the bit about throttling it simply does not make any sense.
They managed to see minor throttling using a relatively non-intense and old game (HL2), yet they could not see throttling using Cinebench. So I decided to dig further. The problem is that right now there is no way to see in real time whether throttling occurs. Anandtech made the conclusion that throttling is not occurring based on the fact that Cinebench produced the same results when they ran it 20 times in a row, or whatever the number was. Yet, a simple game demonstrated a small drop in performance. None of this made sense to me.
If you read any reviews of the same mobile CPUs (there are some on Anandtech's site even) that are in the rMBP (and MBP) you'll see that throttling normally occurs when the CPU cores reach over 90 degrees Celsius. This is how Intel's Turbo boost is designed. Let me be clear though what I mean by throttling. The CPU runs normally at 2.6 GHz (actually it runs at less when it's idling or not busy, but we're going off on a trangent). It also has Turbo Boost which allows the CPU to run some or all of the cores at a higher frequency when a large load is placed, such as a game or a rendering task etc. The CPU will TurboBoost, but if it gets too hot it will throttle back to the default 2.6 GHz, or near that region. So effectively you're not getting all of the speed that the CPU is capable of. If the cooling system in the MBPs was better and able to keep the core temps closer to 90 degrees Celsius, you would see better performance. I can hit 104 degrees Celsius on some of the cores. Sure, that is still below T junction, but regardless it's not very good for the CPU and you're not taking full advantage of Turbo boost.
Now, the problem is that I still haven't managed to monitor real time CPU frequencies on the rMBP in OS X to be able to prove that I am not just a raving lunatic. Please understand that I am not some kid that does not know what they're talking about and I am not here to shout, "ZOMG you guys, this thing is TOO HOT!@!! lolz run for the hills". Most guys will say, "but it's working fine". Sure, it's working fine, but if the cooling system was better, it would work even more fine. This is the crux of the matter. I would like Apple to man up and address it.
I can do this on a 2011 MBA in OS X and it shows some interesting things when you put a proper load on the CPU. Throttling starts at around 93 to 94 degrees Celsius. The 2011 1.8 GHz MBA (i7-2677M) can turbo boost up to 2.6 GHz when both cores are being used. It will throttle back all the way to 2.3 GHz to try and cool itself down. If the cooling system was able to keep up, you would get the full benefit of an overlock up to 2.6 GHz for longer periods of time. I am not going to extrapolate these results to the MBP. As I said, I am still looking for a way to monitor this on the 2012 MBP and right now I don't have the time, nor the inclination to start mucking about with code to resolve the kernel panics I am having with fakeSMC (which is what allows me to monitor real-time CPU frequencies in older Macs).
Going back to the Anandtech review. Cinebench in OS X is a completely flawed benchmark. Why don't you run it and see. I can run it and get a certain score and then I can max out the CPU to 100%, using "yes > dev/null" or another method, run Cinebench again and I get the same score. So trying to use it to ascertain whether throttling is occurring is like using a big stick with a length of about 2 metres to try and measure the precise size of gap that is less than 1 cm. It simply does not work. Cinebench is not stressful enough to cause throttling on a 2011 MBA. I can demonstrate this since I am able to monitor the real time frequency. Therefore, it's unlikely to be able to cause throttling on a MBP, and, even if it did, the scores it will report are completely pointless to gauge whether throttling occurred.
At the end of the day, allowing a CPU to get to just below T junction is simply not good for it, especially for extended periods. Most of us will probably not suffer from the long term ill-effects, because we change these things quite often, but, the cooling system on Apple's portables definitely needs improvement. In my opinion it is pointless to argue against this. Running below 90 degrees will be beneficial to you and to your Mac.
Back to the drawing board.
