The core idea: GHz is only half the equation
A gigahertz is one billion clock cycles per second. A rough working model is:
performance ≈ clock frequency × work completed per cycle
The second half of that formula is usually called IPC, or instructions per cycle. Two chips at the same frequency can finish very different amounts of work, because their designs differ and so do the programs and data they run. A higher clock doesn't guarantee a faster chip in a given task. A lower clock doesn't doom one either.
MakeUseOf's worked example says a 4 GHz CPU at two instructions per cycle handles 8 billion instructions a second, and one at four per cycle handles 16 billion. That arithmetic is fine as a classroom illustration. It is not a way to predict real-world speed. Instructions differ in complexity, programs stall on dependencies and memory, and results shift with workload and system conditions.
IPC is also not a number on the box. It varies by workload, so reviewers estimate it by running tests at fixed clocks, typically on a single thread. Treat any "IPC score" as a benchmark outcome, not a vendor spec.
Fact-check: the example numbers need correcting
The i7-4790K versus i7-12700K comparison. The article says both share a 4.0 GHz base clock. That is not right for the 12700K. According to Intel's spec pages:
| Spec | Core i7-4790K (Q2'14) | Core i7-12700K (Q4'21) |
|---|---|---|
| Cores / threads | 4 / 8 | 12 (8 P + 4 E) / 20 |
| Base frequency | 4.00 GHz | 3.60 GHz (P-core), 2.70 GHz (E-core) |
| Max turbo | 4.40 GHz | 5.00 GHz |
| Cache | 8 MB | 25 MB Smart Cache plus 12 MB L2 |
| Power | 88 W TDP | 125 W base, 190 W max turbo |
The 12700K is a hybrid design with two core types, so it has no single base clock. The article's conclusion still stands, though. The 2014 chip's base clock is higher than the 2021 chip's P-core base clock, and the newer chip is far faster in most workloads. That is a stronger version of the point than the article made.
The Ryzen 9 9950X3D. The article says it "runs at 4.7GHz." Reviewers and spec sheets agree on something else. TechPowerUp lists 4.30 GHz base with up to 5.70 GHz boost, and GamersNexus describes a 16-core, 32-thread part with a 5.7 GHz max advertised boost, 4.3 GHz base clock and 128 MB of L3 cache. TechPowerUp also cites a 170 W TDP and 230 W package power tracking (PPT).
One cache detail is easy to get wrong. The 144 MB figure is the sum of L2 and L3, not L3 alone. AMD defines "total cache" as the sum of all L2 and L3 caches on the package.
The Pentium 4 history. Intel's tested-processor list does show Pentium 4 models at 3.80 GHz, including the 670 and 570J. That supports the article's claim that the Pentium 4 era pushed clocks very high. The broader claims need softer wording:
- That the Pentium 4 "turned clock speed into the headline number" is the author's interpretation, not an established fact.
- That IPC alone explains the Pentium 4 trailing the Athlon 64 is also unsourced here.
- The shorthand is common, and it's reasonable to say NetBurst chased frequency at the expense of work per cycle. Treat it as widely held commentary rather than a measured finding.
Architecture and cache: why a chip can do more per tick
Architecture covers how the cores are built. Wider cores, better branch prediction and more cores all change real performance, but none is a simple "more is faster" switch. Core count helps parallel work such as rendering or compiling. It does little for a lightly threaded app.
Cache is small, fast memory kept close to the cores so they wait on slower RAM less often. L1 is the fastest and smallest, L2 is larger and slower, and L3 is larger again. All of it is faster than main memory. Cache size is not a universal score, because the design of the cache and the workload both matter.
AMD's 3D V-Cache stacks extra L3 and is aimed at gaming. Cache-heavy games can benefit, particularly in frame-time consistency. That does not support the article's flat claim that X3D chips are better than Intel for gaming. Results vary by game, resolution, GPU and test setup, so check current head-to-head benchmarks instead of trusting a brand-level rule.
Power and cooling: boost is a ceiling, not a promise
The advertised boost clock is the highest speed the chip may reach under the right conditions. What it actually sustains depends on workload, temperature and power limits.
Intel's 12th-gen datasheet describes PL1, PL2, PL3, PL4 and a Tau time constant as controls for matching turbo behavior to the platform's power delivery and cooling. It advises against setting PL1 above what the cooling solution can handle, and it describes Tau as the averaging time for PL1's power calculation. That is a more precise picture than "PL2 is a short burst."
Cooler buying is where the article oversimplifies. It says TDP equals PL1 and that you should size a cooler to PL2. Naming and behavior differ across families and motherboards. On the 12700K, Intel lists 125 W as base power and 190 W as maximum turbo power, and motherboard makers often set their own limits. For a cooler, check these:
- The chip's official base and maximum power figures.
- Your motherboard's default power-limit settings.
- The cooler's rated capacity.
- Reviews that test the chip at its actual power draw.
Thermal throttling follows from the same limits. If a cooler can't remove the heat, clocks drop. A modest chip with a good cooler can hold its speeds, while a flagship with a weak one cannot.
How to shop for a CPU instead of reading GHz
- Start with your workload. For gaming, look at game benchmarks that include 1% lows as well as average frame rates. For productivity, look at tests in the apps you use, such as video export, compiling or virtualization.
- Compare in your price bracket. Cross-brand and cross-generation GHz comparisons mislead. Review benchmarks do not.
- Check the test conditions. Power limits, cooler, memory and motherboard settings all change results.
- Weigh the whole spec sheet. Core and thread layout, cache, platform requirements (socket, memory type) and power needs all count.
- Plan the cooling. Match the cooler to the chip's real power behavior, not just a single TDP line.
The takeaway
The article's best point is also its most modest one. GHz still matters within the same architecture, but it can't rank CPUs across generations, designs or workloads. "Clock speed is overrated" is a fair headline if it means "don't use it alone." It would be wrong if it meant clocks are irrelevant. Just don't copy the article's 12700K and 9950X3D numbers into a shopping list. Both are wrong, and the real specs make the argument better.