This high-end „Alder Lake“ chip is Intel’s CPU comeback kid
„Alder Lake,“ Intel’s family of 12th Generation processors, has arrived—and with it, a new CPU paradigm. Intel’s Core i9-12900K desktop CPU ($589) leads the pack of the company’s 12th Generation processors, and brings with it a whole host of upgrades and innovations to the desktops of now and tomorrow. These tick-ups include support for the new, high-speed DDR5 RAM standard, as well as an upgrade to PCI Express 5.0, on the first new motherboard platform to support the latest chips, the Intel Z690. Intel also worked closely with Microsoft to optimize the new CPUs for Windows 11, adding new scheduling features that intelligently load up the Core i9-12900K depending on which cores are being used where, and for what. Alder Lake and the Core i9-12900K indeed impress, but our relationship with the CPU…is complicated. For all the outright wins we saw in our benchmarks (and there were many), the added cost of upgrading to yet another new motherboard platform won’t outweigh the win percentages for many shoppers. Intel’s older-yet-still-reliable „Comet Lake“ Core i9-10900K kept itself in the race during several benchmarks, while the eight-core, rather cheaper AMD Ryzen 7 5800X ($449 list price, but currently snipe-discounted to $386 on Amazon and Newegg) proves itself a worthy contender on performance-versus-price in PC gaming. The high cost of a new Z690 motherboard (the cheapest are just under $200, per our Z690 motherboard guide) and DDR5 adoption, along with Intel’s insistence on upgrading your system to Windows 11, are all front-facing considerations for anyone who’s considering 12th Generation Core as their next big desktop upgrade. That—and a not-insignificant problem in which our test platform, and several prebuilt Alder Lake PCs, could not launch certain popular games that use specific DRM—temper Alder Lake with a bit of wait-and-see caution. Our initial Alder Lake takeaway is „Intel’s on the upswing, with some caveats.“ But read more about our findings below. The Alder Lake Basics: A Whole Lotta ‚New‘ Leveraging Intel’s so-called „7 Process,“ the company’s launch of its new 12th Generation desktop CPUs sees the new chips built on 10nm lithography, finally breaking the company out of its half-decade love/hate affair with the 14nm process and its subsequent „14nm+“-based iterations that followed for years after. (Read more in-depth about how Intel defines its „7 Process“ at ExtremeTech.) This is the first time Intel has moved one of its desktop consumer chip stacks completely beyond a form of the 14nm process in just over five years. AMD, via its manufacturing partnership with TSMC, has seen the Austin chip maker producing its wafers on 7nm lithography for nearly three years now, while Intel is just sidling up to 10nm. Is that a sign for what we should expect to see in benchmarks? Not if Intel’s new bag of tricks has anything to say about it… Intel Bets Big on big. LITTLE Though this would normally be the part of the review where we dive straight into spec comparisons, let’s take a quick sidebar first to learn a bit more about a „big. LITTLE“ chip design: What is it, what does it mean for desktop processors, and do they really need it? In a big. LITTLE approach, a chip design stacks a set of cores focused on peak performance alongside cores focused more on efficiency and power management, both on the same die. That philosophy is nothing new; smartphone processors have been employing versions of a big. LITTLE architecture for years as an efficiency measure. It’s also not technically new for Intel, either—the company first launched an x86 processor based on a big. LITTLE design back in 2020, known as „Lakefield.“ The Lakefield silicon made it into only a few scattered laptops and mobile-device experiments, like the Lenovo ThinkPad X1 Fold. But it set the pace for Intel’s big announcement this year: Intel Performance Hybrid Architecture. Intel is touting Performance Hybrid Architecture as the company’s „biggest architectural shift in a decade.“ In the technology’s Alder Lake debut, the Core i9-12900K (along with the rest of the chips announced thus far in the 12th Generation desktop stack) will each come with two sets of heterogenous cores, instead of the monolithic core design we’ve seen in generations past. The core types are „P-cores“ („P“ being short for „performance“) and „E-cores“ (with the „E“ for „efficient“). The P-cores will be built on Intel’s „Golden Cove“ microarchitecture, while the E-cores are based on „Gracemont.“ This will help to explain the core/thread disparity that you’ll see in the specs below, as only the P-cores will be Hyper-Threaded this time around (that is, supporting two processing threads per P-core), while E-cores will support just one thread per core. These P-cores and E-cores swap off various responsibilities under typical system usage, depending on the task at hand. P-cores, for example, are better for hitting peak performance during demanding tasks such as gaming, while the E-cores are better suited to picking up requests from background tasks that aren’t as sensitive to lower latency speeds. Also, if there’s a task that needs a lot of throughput at once (think multi-core rendering, and the like), the load can be balanced between both P-cores and E-cores, however the operating system’s scheduler sees fit. That scheduler, then, is a key player, especially when the CPU’s resources are much in demand. (And that demand is the point of buying any high-end CPU!) So, speaking of schedulers…meet Windows Thread Director, the ostensible Spielberg of the whole scheme. Windows Thread Director While Intel’s approach to big. LITTLE may be old hat for smartphones and mobile devices, there’s a lot of „new“ on tap here for the laptop and desktop scene. To compensate for that newness, Intel has worked closely with Microsoft this time around, joint-developing a new scheduler for Windows 11, Intel Thread Director. Think of the scheduler as a traffic cop for Windows 11 (or any modern OS, for that matter): It tells bits of programs where they should each run on a processor, based on a variety of factors. That includes thermal/cooling capacity, available power draw, performance peaks, and task/thread priority. This process is relatively straightforward on traditional desktop-processor designs, and it works the same, in principle, on Windows 10 as it has in previous versions. With the introduction of hybrid architecture to the desktop market, however, Intel had to get creative. On a chip where the cores aren’t homogenous, without a little extra help, Windows won’t know which cores it can send programs to most optimally. Enter Thread Director. Thread Director is informed by a new microcontroller on the CPU itself, which will feed Windows 11 more detailed hardware telemetry about the current status of the chip and its cores, versus older releases in Intel’s desktop line. Information that was previously left a mystery to Windows—think aspects like thermals, power settings, and which threads can take more instructions—is now communicated to the scheduler in microseconds, leaving almost no impact and (in theory) adding considerable performance gains, depending on the workflow and the various kinds of overhead that can be leveraged. All of this is necessary, because as you’ll see in our benchmarking results below, Windows 10 can be worse at scheduling a 12th Gen desktop CPU than Windows 11 is. This is mostly due to Windows 10 not knowing what it’s looking at when it sees an E-core, thinking that it’s just a low-performance core that doesn’t have the same power spectrum on tap as the P-cores. As such, it can schedule the E-cores „incorrectly,“ as it were (though that’s an over-simplification), due to its limited information. Meanwhile, the added info being sent from Thread Director to the Windows 11 scheduler is all incorporated into the mix, which should, again in theory, add up to increased performance in that specific OS. Intel even provided several custom „workflow“ benchmarks to its reviewer pool this time, designed to simulate those scenarios where the impact of its hybrid-architecture approach (and Thread Director’s effect) would be most apparent. We ran the Photoshop and Premiere Pro versions of these benchmarks, which showed some impressive results that shouldn’t be shoved aside as mere marketing hoopla. (More on that in a minute.) New Day, New Power Definitions Last up, before we jump into the spec table, Intel has once again reclassified our known definitions of processor power draw. The company says it’s now moving beyond the familiar Thermal Design Power rating (often expressed as „TDP“ in watts), to its newest nomenclature, „Base Power“ and „Maximum Turbo Power.“ In practice, things won’t change all that much for consumers. Both 12th Gen and 11th Gen („Rocket Lake“) chips at the Core i9 tier will retain their 125-watt Base Power rating on the back of the package. Rather, Base Power and Maximum Turbo Power are being used as a more accurate measurement of what kind of spectrum users can expect during both base and boost frequency peak usage, with the Core i9-12900K being rated at „125-watt Base Power,241-watt Turbo Power.“ Specs and Comparisons: Intel Core i9-12900K With that bomber-load of backgrounder out of the way, let’s jump into a look at the full Alder Lake stack, with the Core i9-12900K at the top… First up, the obvious: Intel has chopped down the list of 12th Gen processors available at launch considerably from the 14 options rolled out for early 2021’s 11th Gen/Rocket Lake debut, down to just six „K“ and „KF“ versions. (The KF chips lack integrated graphics.) These chips are all premium ones that are unlocked for overclocking, so it’s tweakers and performance enthusiasts only for now, as far as Intel is concerned. (The premium Z690 motherboards going on sale for the initial Alder Lake launch reflect that focus.
