When you’re in the market for a processor, there is a list of things you should be considering. Traditionally, pretty much the only thing most consumers look at is its total Gigahertz power. Many of those people probably don’t even know what it means (it’s the number of clock cycles—effectively, calculations—a processor completes in one second, in billions; referred to as a system’s clock speed), but it’s an easy thing to compare. If you’re buying a laptop and can choose the processor you want, you can assume, generally, that the one rated at 2.5 GHz is probably faster than the one rated at 2.3 GHz.
The past few years have brought an added wrinkle: Boosting speed. Most processing units, graphical and computational, now have a base clock speed and a boost speed. Intel® calls this Turbo Boost; AMD calls it Turbo Core.
So what does all this mean? More importantly: What does it mean for you? First, let’s talk about the purpose of a “base” clock speed. The faster your processor runs, the more power it requires and the more heat it generates. Take, for example, the Intel® Core™ i7-5820K. It’s a 6-core CPU with a base clock speed of 3.3 GHz and a Turbo Boost speed of 3.6 GHz. For the most part, you want your processor to be running at that slower speed. Basic tasks don’t need a 3.6 GHz processor to run. Indeed, most of them don’t need 3.3 GHz. So, in moments when you don’t need the higher speed, why would you want to raise your power bill and generate extra heat?
For a long time, over-clocking was for enthusiasts only. Over-clocking is a process that takes a capable processor and changes its clock multiplier. Every CPU has a low-level clock that is multiplied in order to reach the number we all know. A CPU with a 300 MHz low-level clock and an 11x multiplier has an effective clock speed of 3.3 GHz. With the right processor, you can change that multiple, thus over- (or under-)clocking your processor. But while the hardest of the hardcore would use liquid nitrogen coolers to break over-clocking world records, most people would be stuck with the number on the box.
These Turbo modes are essentially over-clocks for the masses, but you don’t choose the speed; the system does. When your computer realizes that it needs more clock cycles (say, when you’re trying to render a video), then it will crosscheck the need for speed with its temperature. If it’s cool enough, that means there’s thermal overhead for it to over-clock, at which point it will bring itself up to the boost speed. How long it lasts depends both on how long the system feels it needs to boost, and also whether or not it continues to be reasonably cool.
But it’s worth noting that that top clock speed is for one processor. If you’re running a program that only uses a single processor, then you’ll get that full boost. But if you’re using all available cores (six, in the case of the 5820K), they don’t all boost to that maximum speed. One core would hit 3.6, but all six might only go up to 3.4 GHz when Turbo Boost is activated. (This also depends on your motherboard, and high end/enthusiast motherboards will allow these numbers to go higher than low-end ones.)
With phones, things are a little bit different. More often than not, the manufacturer won’t even tell you what a chip’s base clock speed is, because it’s a number that more or less won’t tell you anything about the chip itself. Under normal conditions, your desktop processor will run at its base speed. A phone, on the other hand, will pretty much never run at that speed. That’s because the base speed on ARM chips, which power nearly every mobile device on the market, is just a few hundred megahertz. But this allows them to run in an idle state with minimal power draw/heat generation.
This base clock speed will never be in effect during actual use. When your phone turns on, the processor screams into action and runs right up around that promised speed. How long it stays there, however, is frequently down to the manufacturing of the phone itself, because as the processors overheat, they throttle themselves. This is true of most processors, but depending on how aggressively it’s done, you may have two phones with identical chips that run at effectively different clock speeds.
Such was the case in 2013 when Google’s Nexus 5 was found to throttle itself heavily due to structural heating problems caused by the phone’s design. A phone with a plastic chassis is more likely to overheat than one with a metal chassis (premium components aren’t just about looking pretty), and phones that don’t dissipate heat particularly well simply won’t run at the same speeds as better-designed phones.
In general, take the boost speeds as a guideline rather than a rule. On a desktop, you never have to worry about your computer running below its base speed (unless you want it to), but on a mobile device with heat and battery-power constraints, it’s more complicated. You aren’t likely to see a phone running at 600 MHz, but that 1.7 GHz processor may, in actuality, be a bit more like a 1.3 GHz processor with the occasional 400 MHz boost. Unfortunately, it’s nearly impossible to know without independent verification where any given system might fall, and the benchmarks that many reviewers run can’t necessarily take this variability into account.
Modern laptops are in a similar place to phones. A new MacBook Air, for example, has a 1.6 GHz Intel Core i5 with a 2.7 GHz boost speed. This allows for the best tradeoff between performance and battery life but, as with phones, laptops have lesser cooling systems than desktop chips, meaning they can’t necessarily sustain those boosted speeds.
If you know your priorities, then you can know what you need. Do you want lower heat generation and power draw/higher battery life but the ability to spike if needed? Look for a processor with a more impressive boost but a lower base speed. Want the ability to overclock it even further? Look for a system with an “unlocked multiplier.” Intel marks those processors with a “K” and AMD with an “FX.” And when you’re choosing a phone, see if there are any outlets that have reported throttling issues on a particular model. Don’t just take the manufacturer’s word for it.
Also, don’t focus exclusively on a clock speed. It’s a helpful metric when comparing different versions of the same processor line, but a 4 GHz AMD CPU is not necessarily more powerful than a 3.5 GHz Intel CPU. Even comparing a modern Intel chip to an older one doesn’t tell you much. A single clock cycle is now far more efficient than it was in the past, so nearly any Intel Core chip is more powerful than any chip from the Pentium days.
And that’s why benchmark tests exist, because they’re the only way to compare performance directly across brands and product lines. There are basically two types of benchmarks: conceptual and practical. Those aren’t official designations, but they get at the underlying point. A conceptual benchmark is specifically and exclusively a benchmark. It’s a specific program that might run in your browser or as its own executable. These output scores can be used to compare processors directly, though they aren’t particularly meaningful in and of themselves. What does the 5820K’s Cinebench R15 single threaded score of 140 mean, or its multi-threaded score of 1,025? And what does it mean that the 5930K gets scores of 146 and 1,083 respectively? Are those slightly higher numbers worth the $200 premium over the 5820K? Some benchmarks test how quickly a system will run the benchmark, and though many of these try to simulate real-world use, the scores don’t necessarily mean much. How would a 10 or even a 100 ms difference in the speed it takes to complete the web benchmark Mozilla Kraken affect your actual experience? It will most likely be a little bit faster, but it’s hard to really know.
Practical benchmarks use programs to do specific tasks—for example, rendering a video or compressing a series of files. Anandtech’s WinRAR test compresses 2,876 files—totaling 1.52GB—and times it. The 5820K completes this task in 46.17 seconds. The 5930K finishes it in 44.95. The $1,000+ 5960X completes it in 34.25. Even though those first two numbers are close, they’re easy to understand. The more expensive chip is slightly faster (as it should be), and the $1,000 monster crushes both of them (as it should). Conversion of a video file might be measured in frames per second, which is also easy to understand. These numbers are more helpful in and of themselves, as they reflect actual-use cases. But, as with anything, benchmarks aren’t all you have to consider. And, of course, benchmarks aren’t all you have to consider. All three of these chips are Intel’s Haswell-E system, but the 5960X has eight cores and a 20MB cache; the other two have only six cores and 15MB. The 5820K has only 28 PCIe lanes and the others have 40. The latter, in particular, is unlikely to show up in many benchmarks, and it’s the slightly lower clock speed of the 5820K (3.3-3.6) versus the 5930K (3.5-3.7) that explains its weaker showing, not the lesser bandwidth.
It is important to keep all this in mind, as an informed consumer is an empowered consumer. It’s nice to have a solid base clock, and it’s even nicer to have a high boost speed. They are certainly crucial in determining what you want and need from your new CPU, but they should count as just one of the things you take into consideration.
|High overall clock speed||Faster||More power required|
|More heat generated|
|Low base clock (with high boost)||More efficient||Greater potential for throttled performance|
|Better battery life (portable devices)|
|Unlocked multiplier (over-clockability)||Ability to increase system performance||More expensive|
|Requires better cooling|
|Multi-core||Better multi-threaded performance||Often worse single-threaded performance|
|Hyperthreading||Effectively doubles the number of processor cores accessible by optimized programs||Most programs aren't optimized|
|Integrated graphics on chip||No need for dedicated GPU||Not all chips have them, which isn't really a con, just true|