Introduction
Continuing Intel’s tradition of previewing their new products prior to the actual embargo, we are now allowed to share with you information regarding Intel’s new upcoming CPUs. Intel has announced that 12th-generation Intel Core CPUs aka Alder Lake will debut on desktop and we will have 6 unlocked SKUs at launch. Intel brands the 12th-gen Core CPUs as the “best gaming processors in the world” while also claiming improved experience and performance that will please content creators as well as lifting up overclocking expectations.
In this article we’ll cover details regarding the new 12th-gen Core CPUs, what to expect and more information regarding the hybrid architecture that will debut with Alder Lake. Reviews will be up November 4th and in those reviews we will be directing readers to this article.
That said, let’s jump right in.
Intel: The story so far…
Intel has a lot riding on this release and newly-appointed CEO Pat Gelsinger has been busy making the rounds to turn things around for Intel since he stepped into his new role earlier this year. Pat Gelsinger has been very vocal about his plans and most of his actions as of taking duty has been under the microscope from the tech media including Intel entering the contract foundry business where they’ll be providing foundry services for customers like Amazon. Pat has also been talking to people from across the globe regarding expanding Intel’s foreign foundries and he’s been eyeing Europe to further solidify not only Intel but the West’s independence over Taiwan’s TSMC dominion over the global chip supply.
For us consumers though, good will can only be established with good products and Alder Lake will be the first new product under Pat Gelsinger’s term. That being said, Intel’s first step into gaining back their leadership performance in the mainstream desktop space rides on the success of Alder Lake.
AMD has decidedly beat Intel with their Zen 3 architecture both in single-threaded and multi-threaded applications. Ryzen has been a growing threat to Intel since the first Zen release back in 2017, decidedly putting Intel in bad light with their insistence on keeping core counts low whilst AMD offers more and more cores. This may be partly due to their manufacturing woes to transition to 10nm, ending their tick-tock cycle. With 2020 proving to be a tough year for all but a boon for digitalization in all aspects of life, Intel poured their efforts into providing what they have for now and prolonging 14nm by just one more generation.
This year, Intel has made many bold announcements aside from their IDM 2.0 strategy: Intel will starting tapping into 3rd-party foundries to make their chips while they optimize their own. Intel has also announced their entry into the discrete graphics with their Arc GPU brand. Intel has also rebranded their node naming, further diluting into the marketing mess that is node names with their 10nm Enhanced SuperFin process now known as the Intel 7 process. It has been noted that Intel has always had denser transistor counts against smaller nodes from other foundries.
Adding More Cores: Hybrid Architecture
Adding to the node naming confusion is now Intel’s new approach to adding more cores: the hybrid architecture. Intel still believes in a monolith chip design. One big piece of silicon housing everything in that die. This is in large contrast to AMD’s chiplet design which separates cores into clusters of their owns dies. This has so far worked for AMD to great effect but Intel’s persistence sees them devising smart ways to a solution; not just to increase core count, but increase efficiency as well.
Intel’s 12th-gen Core CPUs will feature 2 kinds of cores: the familiar core design we normally see and what is referred to as the P-core (performance) based on Golden Cove and then we have high-efficiency E-cores (efficiency) based on Gracemont. If this would be a traditional gen-on-gen release, the P-cores are expected to yield 28% performance increase over Skylake. That would usually be the end of that but this hybrid architecture approach adds in the E-cores which can handle very light workloads with very minimal power. This allows Alder Lake to park P-cores when they’re not needing, only bringing them in for gaming, content creation and then combining both E-cores and P-cores when really taxing workloads are being performed.
Intel 12th-gen Core Processor Line-Up
Intel will be releasing six unlocked CPUs for launch of their 12th-gen CPUs, offering both K and KF variants of the 12th-gen Core i9, Core i7 and Core i5. These CPUs will be launched alongside Z690 chipset motherboards aimed for enthusiasts, many of which have already been announced by partners like ASUS, MSI, ASRock and GIGABYTE plus more. A second wave of releases is expected for 2022 and will include the rest of the 12th-gen CPU lineup.
The Core i9-12900K and i9-12900KF will be the flagship models and will have 8 P-cores with HyperThread + 8 E-cores. This effectively makes them 24-thread CPUs. Take note of the core convention as Intel will be using the E-cores as well to segregate their SKUs. You’ll see the convention of P-cores + E-cores used so you may see us show the core counts as 8+8, with the current line-up’s models all having HyperThreading. It has been shown that upcoming SKUs may not include E-cores altogether but for now, the 6 launch models have this commonality.
Going back, the Core i7-12700K and Core i7-12700KF will be 8C/16T+4C parts while the Core i5-12600K+i5-12600KF both a 6C/12T+4 part. Again, the KF models will lack the Intel UltraHD 770 IGP and will slightly be cheaper than their full K-type siblings.
Further features worth noting is Turbo Boost for these CPUs: the Core i9 and i7 will have Turbo Boost Max 3.0 while the Core i5 won’t. In the chart above you’ll also see that we don’t have Thermal Velocity Boost anymore. There is also a new column for Processor Base Power and Maximum Turbo Power. We’ll discuss this in a later segment. Lastly, Intel has confirmed that the 12th-gen Core CPUs do not support AVX512.
Alder Lake Die Layout
As mentioned earlier in this article, Intel is retaining their monolithic chip design, utilizing a single die to contain all of the components of the CPU. Intel wanted Alder Lake to be scalable between client segments and desktop designs and easily be ported to mobile (laptops) and ultra mobile (tablets).
Intel built Alder Lake on their Intel 7 process, formerly known as 10nm Enhanced SuperFin. This node does see an increase in transistor count over last-gen but the basic layout of the CPU remains the same. The major difference here is the addition of E-cores which take up some space on the die necessitating an increase in the overall footprint of the CPU.
Intel will also include the newer UHD 770 Graphics based on Intel Xe LP with Alder Lake. This is the same architecture used on the IGPs of 11th-gen Rocket Lake CPUs and while it is not a viable gaming solution, given the skyhigh pricing of GPUs, those who may want to weather out their GPU purchase until the next payday could live off UHD 770 for work and entertainment purposes.
Alder Lake Supports DDR5 and DDR4, PCIe 5.0 Enabled
Just one generation of PCIe Gen4 and Intel is already jumping to PCIe Gen5. With 12th-gen CPUs, Intel is allocating 16 lanes of PCIe gen5 from the CPU. PCIe gen5 provides double the bandwidth of PCIe gen4, about 128GBps of bandwidth in full x16. This will usually be reserved for the first PCIe slot used mainly for graphics but motherboard partners can design boards that can split these lanes for x8/x8 use or divert them to M.2 PCIe slots. The CPU also has 4 PCIe gen4 lanes for an M.2 slot.
Another new change Alder Lake is bringing is support for DDR5. Alder Lake supports base DDR5-4800 of up to 128GB max. DDR5 in itself features a lot of changes, most notable of which is the move to have the PMIC onboard so DDR5 modules manage their own power. Linus Media Group has published a concise explanation of DDR5 which best to visually explain the way DDR5 works:
Alder Lake will be a transitional platform though and board makers will have the choice of using either DDR5 or DDR4 on their boards so users can choose which path they’d like to pursue. Intel has stated in a media Q&A that it is not possible to turn both DDR5 and DDR4 at the same time but it could be possible that some board makers may come out with slots for both that can accommodate both, just not at the same time.
P-Core and E-Core Details
Detailing the microarchitecture for Alder Lake means detailing both core architecture. The Performance cores are based on Golden Cove while the Efficiency cores are based on Gracemont. The best analogy I can think of for this arrangement is that of a hybrid-powered sports car. You use electric when city-driving but pump-out the gas when you need the torque and speed. This is how Intel is presenting their hybrid architecture philosophy. Diving deeper, we have Gracemont-based cores which are presented to have 40% more performance but at 40% power draw as well. This means that despite them being claimed to be low-power cores, Intel’s 12th-gen CPUs could exceed a Haswell system or older clock-for-clock at less power. Intel details Gracemonent extensively in their Architecture Day 2021 presentation.
The Golden Cove P-cores promises to make-up for what Intel has been lacking for the past few generations and is claimed to have an IPC increase of up to 28% than Skylake and 19% over last-gen’s backported Cypress Cove.
Making it Work Together: Intel Thread Director
In Intel’s media session, a visual demo of Thread Director was presented. This demo shows what cores are active when a certain workload is applied. Thread Director is the important catalyst that makes Intel’s hybrid architecture work by talking to the OS and letting it know which to cores to use.
Intel states that they’ve analyzed usage scenarios and has worked closely with Microsoft to engineer Windows 11 to work harmoniously with a hybrid architecture. Intel does state that Linux and Chrome OS Thread Director are still in-development but states that Alder Lake is aware enough to utilize the most optimum core utilization. Windows 10 is also stated to have a great experience. Intel is adamant about not setting hard affinity on the cores.
The Z690 Chipset
I’ve mentioned earlier that Intel will release the Z690 chipset alongside the debut lineup of 12th-gen Core CPUs. As all the launch models are unlocked SKUs meant for enthusiasts, Intel is also releasing only the Z690 chipset.
I have to say that this generation will see a massive leap in pricing for motherboards and that’s already barring shortage. Regardless, board makers will see it meets users expectations for the price as aside from the rich feature set of enthusiast-class boards, the Z690 chipset itself will be bringing some expanded IO most notably a richer USB3.2 set with USB3.2 Gen2x2 now onboard plus WIFI-6E.
Thoughts and First Impressions
Our reviews will primarily focus on gaming performance but I’ve made extensive effort in making sure that content creation and daily use is covered as well. For now, Intel is seemingly bringing a lot with this launch with PCIe Gen5 and DDR5 making their mainstream desktop debut as well as Intel’s formal move away from 14nm and the Skylake curse.
It is exciting to see how Intel’s approach at chip design will stack-up against AMD. Intel’s decision to stick with a monolithic CPU seems to be Intel still staying true to their design yet Intel isn’t chasing cores but their way of skirting around the core count race by running with a hybrid-core design is quite a wise move. Should Intel prove manage to beat or at least tie the 16-core AMD flagship 5950X in the majority of benchmarks, it’d be quite the achievement which should push AMD more to innovate further, finally bringing back the arms race that should drive mainsteam CPUs further.
Reviews for Intel’s 12th-gen Core CPU aka Alder Lake will be up by November 4.