Apple’s M-series chips have dominated the efficiency conversation for years, and have legitimized Arm as a computing platform in a way no other manufacturer has. Ever since the first M1 showed just how far Arm-based silicon could go in a laptop, Intel’s CPUs have been playing catch-up. With Lunar Lake, Intel finally seems to be closing that gap… though not entirely.
I ran a series of real-world power draw tests comparing Apple’s M4 Pro in my 16-inch MacBook Pro against Intel’s Core Ultra 8 256V, a Lunar Lake chip inside the HP OmniBook 16. It’s not a perfect apples-to-apples comparison, of course. These are two different operating systems and two very different cooling designs, but the results were too interesting to ignore nonetheless.
Before getting into the details, there are some caveats to set out, because the data tells a nuanced story about both platforms.
About this article: The Apple MacBook Pro was purchased by me. The HP OmniBook X Flip 16 was sent to us by HP. Neither company had any input into this article’s content.
Neither chip ran at full pelt
And that’s by design
In my testing, both SoCs appeared to throttle long before reaching their theoretical limits. The HP OmniBook 16’s fans spun up quickly, and package power quickly decreased over time, hinting that it was quickly crossing thermal thresholds that the system wanted to avoid at all costs.
The M4 Pro, on the other hand, was in macOS’s default Low Power Mode, which intentionally reduces its power budget. Apple’s implementation doesn’t just underclock the CPU; it dynamically adjusts GPU clocks, memory controllers, and background task scheduling to minimize drain.
That makes this comparison particularly interesting. Despite Apple’s M4 Pro being deliberately power-limited, it consistently delivers exceptional sustained performance per watt as numerous benchmarks have solidifed over the last year. And, crucially, it completes tasks faster, meaning that it reaches “race to idle” more efficiently. Even though raw package power figures sometimes show the M4 pulling slightly more under full load, its energy efficiency remains superior because it gets the work done faster and spends less time there. Keep in mind that package power, while the easiest way to measure CPU power consumption, includes other aspects of the SoC, such as the power consumed by the memory as it’s an instrinsic part of the SoC.
Intel’s Lunar Lake platform represents a genuine leap forward from Meteor Lake and older designs from the American chipmaker. It consolidates the CPU, GPU, and NPU into a unified tile built on TSMC’s 3 nm process for the compute tile and TSMC’s 6nm process for the platform controller, which is a major architectural overhaul designed to rival Apple’s unified SoC approach. Intel has been set on “busting the myth” of Arm’s comparative efficiency for quite a while now, and Lunar Lake was a step in the right direction to achieving that.
With that said, the raw efficiency gap remains. We already know that both Qualcomm and Apple best Intel in raw performance when comparing to Intel’s Core Ultra 7 258V, and the 256V we’re testing here is quite similar: it has slower clock speeds and less L3 cache, but that’s about it. The moat that Intel needed to cross in order to gain comparable efficiency seemed incredibly ambitious, yet a year after Lunar Lake’s release and numerous chipset and driver updates, there are real things for Intel to gloat about.
Things look good for Intel, but not Arm-levels of good
Yet there’s very real progress here
Getting into our testing environment, power information on Windows was captured using HWInfo, and power information on macOS was captured using powermetrics. Idle power was calculcated by shutting down all background processes, and CPU usage was pinned at 100% using “Endurance” on macOS and OCCT on Windows. All tests were run for ten minutes. These aren’t lab-grade conditions, but they’re consistent enough to provide meaningful comparisons of real-world behavior.
Again, this isn’t a true apples-to-apples match. Different operating systems, schedulers, and power management approaches make perfect parity impossible. On Windows, “package power” covers CPU, GPU, and NPU activity, which roughly aligns with the combined SoC measurement on macOS. It’s not exact, but close enough for general analysis. In our case, the CPU power that HWInfo can calculate on Windows does not encompass all components required to make a comparison, and arrives at a significantly lower value than those measured here, hence why package power is used. With that said, for the purposes of a general overview, these are comparable.
First, let’s look at our idle capabilities when comparing both chipsets.
At idle, Intel clearly holds the advantage, which is certainly a reversal of what you might expect given Arm’s reputation. The M4’s SoC never truly goes silent, as background power management and DRAM access remain active.
Intel’s chip, by contrast, drops to sub-1 W levels when fully idle, a credit to Lunar Lake’s aggressive power-gating, redesigned idle states, and features like its Low Power Island. Still, average figures only tell part of the story, as Apple’s spikes are brief and workload-dependent, while Intel’s lower idle power doesn’t automatically translate to better efficiency under use.
Now let’s look at what happens when both chips are pushed to their limits.
Here’s where context matters. At first glance, both chips appear roughly neck-and-neck, as after all, the M4 Pro draws around 38 W under full sustained load, while Intel’s 256V averages about 21.45 W. Looks good, right? That’s far closer than the prior generation, where Intel’s designs could spike well past 60 to 80 W. But Apple’s advantage lies not in its lower wattage, but in how efficiently it translates that power into work done.
In practice, the M4 Pro completes heavy CPU tasks much faster than Intel’s Lunar Lake, meaning its total energy consumed over the duration of a workload is often significantly lower. Apple’s “race to idle” approach (that is, boosting performance briefly, then dropping back to near-idle) remains key to its battery life lead. Even if peak power draw looks similar, the M4 spends less time drawing it. Coupled with the fact that Apple’s M4 Pro absolutely smokes the Core Ultra 7 256V in benchmarks, it no longer looks as good for Intel.
In fact, the most telling aspect of the drop-off in power consumption is why the power dropped. It’s not because Intel has a more efficient chip; in fact, as temperatures rose above 60C, it underclocked and undervolted more and more. By the end of the test, it was running at a mere 1.5 GHz. The M4 Pro was still operating at maximum frequencies (4.5 GHz for P cores and 2.8 GHz for E cores) across the board at the ten minute mark.
How close can Intel really get?
We don’t really know, yet
I wouldn’t say that Intel has “busted the myth” as such; rather, the company has shown that x86 can be optimized. Lunar Lake certainly closed the efficiency gap, and the architectural unification, improved low-power states, and reduced platform overheads make it the most competitive Intel has ever been in the space so far. However, Apple still wins in all of the places that it counts, like real-world energy use and workload responsiveness.
Sure, the M4 Pro may draw slightly more power under load and at idle, but it finishes tasks faster, transitions smoother, and wastes less energy in the process. In general usage, the power spikes of the Intel Core Ultra 7 256V are significantly higher than anything I’ve seen from Apple Silicon, and you’ll find Lunar Lake’s chips spiking to high power levels rather frequently. It’s a leaps and bounds improvement over predecessors, don’t get me wrong, but it’s not beating Arm, nor is it busting any myths. It’s just starting to catch up.
Panther Lake is going to be on its way to PCs and laptops next year, manufactured on Intel’s shiny new 18A process. It’s meant to be more energy efficient, and with that, we’ll see when consumers get their hands on real devices if it’s truly the moment that Intel matches Arm’s efficiency.