Intel architecture explained
With a brand-new architecture consisting of two different types of core, the 12th generation Intel chips are very different to what came before
For its 12th-generation CPUs, based on the Alder Lake architecture, Intel has gone back to the drawing board and come up with something very different. Most CPUs in the range now have two types of core: performance (P-cores) and efficiency (E-cores).
The idea behind the design is similar to that of ARM’s Big.Little architecture, which combines low-power cores for smaller background tasks, and large performance cores when power is needed. That’s exactly what we have with Alder Lake, with P-cores taking care of high-end tasks and the E-cores there for efficient background processing.
All cores, both P and E, are built using the Intel 7 process, which was known as the 10nm Enhanced SuperFin. These processors finally see the end of the 14nm process, and we’re now onto the second generation of Intel 10nm fabrication (it’s the first time we’ve seen this on Intel’s desktop processors, whereas AMD has been at 7nm since 2019).
P-cores
The P-cores are your traditional processing cores. If this was any other year and Intel hadn’t introduced E-cores, then the 12th-generation would have consisted of these alone. For this generation, we have the Golden Cove architecture, which Intel claims gives around 19% more instructions per cycle compared to the older 11th generation products.
Every P-core has HyperThreading, which doubles the number of threads available. As has been standard for a long time, the P-cores have two clock speeds quoted. The first is the base clock speed, which is the minimum that the processor will run at. This differs massively across the range, with the unlocked CPUs generally having higher base speeds; locked processors will clock down further.
Then there’s the Turbo Clock speed, which is the maximum clock speed that the P-cores will jump to when pushed. These speeds rely on the available thermal overhead: the better the cooling in your computer, the higher the clock speed and the faster your computer will run.
Just to confuse things, Intel has also introduced Turbo Boost Max Technology 3. This is for lightly threaded performance and is designed to make the best-performing cores run faster. During manufacturing, it’s normal for some cores to be capable of running at higher clock rates. Turbo Boost Max 3 is there to identify these cores automatically, pushing them higher when there’s enough thermal headroom (such as in lightly threaded tasks).
E-cores
E-cores are the efficient ones. On the CPU, several E-cores can fit into the same space as a P-core. These cores use the Gracemont architecture, and Intel says that each E-core offers 40% more performance over a Skylake core when running at the same power. Skylake launched back in 2015, so this doesn’t sound like a very impressive gain; however, the E-cores aren’t there to deliver the best performance for high-end tasks; they’re there to take care of background work efficiently.
As Intel says, E-cores are optimised for multicore performance-per-watt tasks. In other words, when you’ve got less demanding background tasks, the E-cores can step in and run them efficiently, using less power and generating less heat than a CPU with P-cores only.
E-cores don’t support HyperThreading, so it’s strictly one thread per core. As a result, many 12th generation Intel processors have an odd-looking number of cores and threads. The Core i9-12900K, for example, has eight P-cores delivering 16 threads and eight E-cores. That’s a total of 24 threads spread between the hybrid architecture.
While the E-cores are an exciting development, be aware that they don’t appear in every processor in the range: cheaper models ditch E-cores and just have P-cores.
Intel claims the Golden Cove architecture gives around 19% more instructions per cycle compared to older 11th gen products
Thread Director
Having a new architecture is great, but it does add some complexity into the mix, as your operating system has to know which cores to send tasks to. For this, Intel has its new Thread Director architecture.
It’s a hardware feature that assigns threads to E-core or P-cores based on the optimisation that is required. Although it uses built-in hardware,
Thread Director requires operating system support to work best, and it’s currently only directly supported in Windows 11.
Alder Lake CPUs will run in Windows 10, and there is some optimisation available, but not the full gamut. As a result, it’s possible that Windows 10 can assign threads to the wrong processing cores, which will mean that performance can vary significantly. There are some workarounds, such as using powercfg to force a task to run in the foreground or the Process Lasso add-on. It’s not ideal, so if you don’t want to upgrade to Windows 11 and prefer to stick with Windows 10, you may want to choose a 12th gen processor that doesn’t have E-cores or go with AMD instead.
New chipsets
All of the 12th generation chips are larger than the previous generation, which means a new socket: LGA-1700. The new socket isn’t compatible with older coolers, either, although many manufacturers do offer adapters for them. Noctua, for example, will send out a free adapter for its older coolers, although it now sells products that are LGA-1700-compatible.
Of course, it wouldn’t be a new Intel processor without the need for a new chipset as well. Alder Lake chips launched with the high-end Z690 motherboards, but since then, cheaper B660, H670, and H610 motherboards have been launched.
All common sizes (ATX, micro-ATX and mini-ITX) are supported.
As well as the new processor architecture, the chipsets support some of the CPU’s new features.
DDR5 memory
One of the biggest changes with Alder Lake chips is that they support DDR5 memory as well as DDR4 memory. The ultimate advantages of DDR5 are its higher capacities per stick and faster speeds. However, DDR5 is relatively expensive and the current speeds aren’t that much faster than the best DDR4 memory. You do have to make the memory decision at the time you buy, however, as a motherboard supports either DDR4 or DDR5.
PCI-E standards
PCI-E 5 (sometimes called Gen 5) is also supported, with up to 16 lanes for storage and graphics. That’s something that will come in the future. For now, there are four lanes of PCI-E 4 (Gen 4) for storage via M.2 slots, with some motherboards offering up to four slots.
Product variations
The differences between the available models are both range (i9, i7, i5 and i3) and then the features, denoted by letters at the end of a model name, such as “K” or “F”. Given the new hybrid architecture, the processor range is slightly more complicated than you might expect, and what’s available differs through the range. Our table on p76 helps you see the true differences between processors, but we’ll explain the range as best we can.
Core i9 has Intel’s high-end, performance-led processors. Everything in the range has P-cores (eight) and E-cores (eight), offering the overall best performance of the range. These processors tend to run hotter than the lower-end chips.
The Core i7-range is the next step down, offering enthusiast-level performance. Again, all CPUs have eight P-cores, with the number matching those on i9 chips, but half the number of E-cores (four), which has an impact on some tasks; however, for any task that requires P-cores, such as gaming, performance between the two chips is very similar.
Things start to get more complicated when we get down to the Core i5 range. Here, we see a drop to six P-cores across the range. However, only the Core i5-12600K has four E-cores, while the rest of the range ditches them entirely.
That has quite an impact on performance for some tasks.
Down at the entry-level Core i3, we have four P-cores only, and there are no E-cores. This makes the i3 more of a standard upgrade on the previous generation, with none of the main advantages of the new architecture.
The variations in the range come from the letters at the end of the name. Any processor with a “K” in it is an unlocked variation. These chips can have their multiplier adjusted manually to boost the overall speed limit; higher-frequency chips require better cooling.
Then there are “F” chips in the lineup. These ditch the integrated graphics and come at a slightly lower price, although they have the same CPU performance as the regular version. If you’re always going to use a graphics card, you might as well save a bit of cash. It’s possible to get chips that use both letters: the Core i7-12700KF is an unlocked CPU with no onboard graphics.
All chips have an L2 cache and smart cache (effectively L3), which are shared between all cores. The total number of each is dependent on the total number of cores.
All of the 12th generation chips are larger than the previous generation, which means a new socket: LGA-1700