Intel now sell their CPUs with an advertised Core Clock speed, and a Turbo Boost (up to) speed.
As well as this, thanks to SpeedStep (dynamic frequency scaling), a CPU will often underclock to conserve power and reduce heat. Therefore, a CPU rarely sits at its advertised core clock speed anymore; it's constantly changing in realtime between its minimum clock speed, and its Turbo Boost speed.
Take for example, my dual-core Core i7-4650U:
- Base clock: 1.7GHz
- Minimum/idle clock: 0.8GHz
- Turbo Boost (with both cores active): 2.9GHz
Here's I tried to post a graph displaying the clock speed of my CPU over the time it took to write this question (using Intel Power Gadget), but I need 10 rep to post an image, so you'll have to take my word for it: the frequency was rarely (if ever) set to 1.7GHz.
If I earn 10 rep in the meantime, I'll upload the image :)
So, my question is: what does the base clock speed of an Intel CPU actually indicate anymore? Personally, as a consumer, all I really care about is how low (for power savings) and how high (for performance) a CPU can be clocked at.
Answer
The base clock speed is a guaranteed clock speed for full typical (but not peak) utilization. As long as one is not running a power/thermal virus (i.e., not peak utilization), one should be able to fully use the hardware at the stated clock speed without violating the thermal design power.
When software is not using all the available cores, some additional thermal headroom is available which can be exploited by increasing the clock speed of the active cores (Turbo Boost).
Incidentally, minimum clock speed is not especially useful for evaluating energy efficiency since a significant amount of power consumption is static power (independent of the work done). At a certain point, race to sleep (a.k.a., hurry up and go to sleep) will be more energy efficient than reducing the clock speed because sleep states reduce the static power use.
No comments:
Post a Comment