cpukernel-extensionsperformance
Intel mobile CPUs now have a really low base frequency and a high Turbo Boost freq, e.g. i5 2.4~4.1 GHz.
Turbo at 3.8+ makes it go all the way up to 100°C. Using 'Turbo Boost Switcher' to disable TB makes the freq too low at 2.4.
Is there any way to limit CPU freq at some intermediate value like 3.2?
Turbo Boost is a feature that, when fewer than the total number of cores are being used, the processor can turn off the unused cores and increase the clock speed on the rest of the cores. This increases performance (the cores that are being used get faster) and can reduce power usage.
For instance, if you have the 2011 MacBook Air with the 1.6 GHz Core i5 (as detailed in this AnandTech article), the "Turbo Ratio" is set to 0047. The number sets the amount of boost by digit (4 cores = 0, 3 cores = 0, 2 cores = 4, 1 core = 7; this is a dual-core CPU, so 3- and 4-core mode is irrelevant). The number specifies the boost in 100 MHz increments above the listed clock speed, so if two cores are running, they can be boosted to 1.6 GHz + 400 MHz = 2.0 GHz and if one core is running, it can be boosted to 1.6 GHz + 700 MHz = 2.3 GHz.
To see what your system is set to, go to Console and open the kernel.log (or system.log in Mountain Lion*), then search for AppleIntelCPUPowerManagement and look for "Turbo Ratios". The Mac I'm on now (a 2.0 GHz Core i7 MacBook Pro), for example, has Turbo Ratios set to 6689 for a maximum clock of 2.9 GHz when using only one core.
* thanks to @gentmatt in the comments below
So, I ran some Cinebench R11.5 benchmarks to get more information about the single threaded performance of my i5-3427U1 (1.8-2.8 GHz) CPU:
1 thread → 1.10
I also contacted the author of the linked article by AnandTech - @anandshimpi2. Anand Lal Shimpi was so kind to send me a benchark chart which compared the single threaded performance of the 2011 13" Macbook Airs with the 2012 13" Macbook Airs using Cinebench R11.5.
Into this picture, I have added information about the maximum single threaded CPU clocks along with a comparison of the measured performance such as Anand Lal Shimpi has suggested.
In this info graphic I am comparing specifically:
- 2011 13" Macbook Air, i5-2557M 1.7GHz/2.4GHz/2.7GHz
- 2012 13" Macbook Air, i5-3427U 1.8GHz/2.6GHz/2.8GHz
As you can see, the single threaded clock of the 2012 model is 3.7% increased in turbo boost mode.
However, the benchmark score saw a 9.9% increase which is not simply due to the higher clock, but also due to the more efficient Ivy Bridge architecture which is about 5-15% faster3 than Sandy Bridge.
In sum, the benchmark scores are as expected whereas the Turbo Ratios log entries are not. Anand Lal Shimpi had "figured it was a bug, submitted it to Apple back in July"4.
Best Answer
As I understand the intel feature it's on or off. There is no in between setting. Also running at above the base frequency will cause the CPU temp to creep up slowly.