SkatterBencher #37: Intel Core i5-12400 Overclocked to 5300 MHz

We overclock the Intel Core i5-12400 processor up to 5300 MHz with the ASUS ROG Maximus Z690 Extreme motherboard and EK custom loop water cooling.

As I had already hinted at in my launch article, non-K overclocking on Alder Lake is back. However, the enthusiastic media and hyped-up overclocking community ignored an essential part of the story. There’s a lot more to non-K overclocking than meets the eye, and with this article, I set out to provide context for those who’d like to be informed beyond the hype.

The Core i5-12400 is part of the non-overclockable Alder Lake desktop processor product line. It is the first non-K Intel part I’ve overclocked in a long time.

Intel Core i5-12400: Introduction

The Intel Core i5-12400 is part of the 12th generation Intel Core processor lineup.

Intel Alder Lake has an all-new core design with performance hybrid architecture featuring Performance Cores and Efficient cores. It’s built on the Intel 7 process technology, formerly known as 10nm Enhanced SuperFin (ESF). It’s a scalable SoC architecture which means Alder Lake will cover all client segments from 9W for ultra-thin notebooks to 125W for gaming and workstation desktops.

While we extensively covered the K-SKU processors on this channel already, we’ve said almost nothing about the non-K processors. The non-K processors include essentially all the processors that don’t carry the K-suffix and are, by definition, not overclockable.

The non-overclockable processors are the most significant chunk of the Alder Lake lineup. Of the 28 desktop parts, 22 parts are not overclockable.

alder lake non-k desktop cpus

While the Core i7 and Core i9 Alder Lake non-K parts have E-cores, from Core i5 downwards, the processors only have the high-performance P-cores. It is important to note that they also have a completely different and much smaller H0 die. This will be important later on.

alder lake c0 and h0 dies

The Core i5-12400 processor has 6 P-cores and a total of 12 threads. The base frequency is 2.5 GHz for the P-cores. The Turbo Boost 2.0 frequency for single-core is 4.4 GHz and for all-core is 4 GHz. The processor base power is 65W, and the maximum turbo power is 117W.

In this article, we will cover four different overclocking strategies:

  • First, we enable AVX512, MCE, and XMP
  • Second, we overclock using BCLK using the conventional method
  • Third, we use non-K BCLK overclocking and a fixed CPU ratio
  • Lastly, we use non-K BCLK overclocking and a dynamic turbo CPU ratio
skatterbencher #37 overclocking strategies

However, before we jump into overclocking, let us quickly go over the hardware and benchmarks we use in this article.

Intel Core i5-12400: Platform Overview

Along with the Intel Core i5-12400 processor and ASUS ROG Maximus Z690 Extreme motherboard, in this guide, we will be using a pair of 16GB DDR5-6200 Hynix memory sticks, an RTX 2080TI graphics card, a 512GB M.2 NVMe SSD, a Seasonic Prime 850W Platinum power supply, the ElmorLabs Easy Fan Controller, the ElmorLabs Power Measurement Device, the ElmorLabs EVC2, and EK-Quantum water cooling. All this is mounted on top of our favorite Open Benchtable V2. 

The cost of the components should be around $4,927.

  • Intel Core i5-12400 processor: $200
  • EK-Quantum Magnitude 115X: $242
  • EK-Quantum Kit P360 water cooling kit: $578
  • ASUS ROG Maximus Z690 Extreme motherboard: $1,400
  • ASUS ROG RTX 2080 TI graphics card: $1,500
  • AORUS RGB 16GB DDR4-4400 memory: $400
  • AORUS RGB 512 GB M.2-2280 NVME: $110
  • Seasonic Prime 850W Platinum power supply: $200
  • ElmorLabs Easy Fan Controller: $20
  • ElmorLabs EVC 2: $32
  • ElmorLabs Power Measurement Device: $45
  • Open Benchtable V2: $200

I know the motherboard is not quite a price or segment match for the Core i5-12400 processor. The article’s primary focus is to explain how Alder Lake non-K overclocking works, not whether it makes sense from a value for money perspective. However, I will get into that later in this article.

ElmorLabs EFC, EVC2, & PMD

I explained how I use ElmorLabs products in SkatterBencher #34. By connecting the EFC and PMD to the EVC2 device, I monitor the ambient temperature (EFC), water temperature (EFC), fan duty cycle (EFC), and CPU input power (PMD). I include the measurements in my Prime95 stability test results.

I also use the ElmorLabs EFC to map the radiator fan curve to the water temperature. Without going into too many details: I have attached an external temperature sensor from the water in the loop to the EFC. Then, I use the low/high setting to map the fan curve from 25 to 40 degrees water temperature. I use this configuration for all overclocking strategies.

Intel Core i5-12400: Benchmark Software

We use Windows 11 and the following benchmark applications to measure performance and ensure system stability.

Intel Core i5-12400: Stock Performance

Before starting any overclocking, we must first check the system performance at default settings.

Please note that the ASUS ROG Maximus Z690 Extreme enables AVX512 and MultiCore Enhancement out of the box. So, to check the performance at default settings, you must go to the BIOS and

  • Enter the Extreme Tweaker menu
  • Set ASUS MultiCore Enhancement to Disabled – Enforce All Limits
  • Enter the AVX Related Controls submenu
  • Set AVX512 to Disabled

Then save and exit the BIOS.

Here is the benchmark performance at stock:

  • SuperPI 4M: 39.810 seconds
  • Geekbench 5 (single): 1,624 points
  • Geekbench 5 (multi): 8.989 points
  • Cinebench R23 Single: 1,578 points
  • Cinebench R23 Multi: 11,223 points
  • CPU-Z V17.01.64 Single: 651.1 points
  • CPU-Z V17.01.64 Multi: 4,912.4 points
  • V-Ray 5: 8,205 vsamples
  • AI Benchmark: 2,551 points
  • 3DMark Night Raid: 51,118 points
  • CS:GO FPS Bench: 545.27 fps
  • Final Fantasy XV: 179.33 fps
Core i5-12400 stock benchmark performance

Here are the 3DMark CPU Profile scores at stock

  • CPU Profile 1 Thread: 870
  • CPU Profile 2 Threads: 1,615
  • CPU Profile 4 Threads: 2,737
  • CPU Profile 8 Threads: 4,537
  • CPU Profile 16 Threads: 5,886
  • CPU Profile Max Threads: 5,846
Core i5-12400 stock 3dmark cpu profile performance

When running Prime 95 Small FFTs with AVX enabled, the average CPU P-core clock is 3140 MHz with 0.909 volts. The average CPU temperature is 45 degrees Celsius. The ambient and water temperature is 23.2 and 29.0 degrees Celsius. The average CPU package power is 65 watts.

Core i5-12400 stock prime95 small ffts avx enabled

When running Prime 95 Small FFTs with AVX disabled, the average CPU P-core clock is 3500 MHz with 0.956 volts. The average CPU temperature is 47 degrees Celsius. The ambient and water temperature is 23.2 and 29.3 degrees Celsius. The average CPU package power is 65 watts.

Core i5-12400 stock prime95 small ffts avx disabled

Now, let us try our first overclocking strategy.

However, before we get going, make sure to locate any of the following three buttons: Safe Boot button, ReTry button, and CMOS Clear button

The Safe Boot button temporarily applies safe settings to the BIOS while retaining the overclocked settings, allowing you to modify the settings which caused a boot failure.

The ReTry button forces the system to reboot if it locks up during the boot process, and the Reset button is rendered useless. It will not change anything to your BIOS settings.

You can find both Safe Boot and ReTry buttons at the bottom right of your motherboard.

Pressing the Clear CMOS button will reset all your BIOS settings to default which is helpful if you want to start your BIOS configuration from scratch. However, it does not delete any of the BIOS profiles previously saved. The Clear CMOS button is located on the rear I/O panel.

OC Strategy #1: AVX512 + MCE + XMP 3.0

In our first overclocking strategy, we take advantage of enabling AVX512, ASUS MultiCore Enhancement, and Extreme Memory Profile.

AVX512

One of the most exciting topics of the Alder Lake product launch was the support for AVX-512 instructions. While Intel had communicated that Alder Lake does not officially support AVX-512, they had mentioned P-core support for the instruction set in technical documentation.

As became evident when Alder Lake launched, while official support was indeed missing, you can still enable AVX-512 through BIOS. But there’s a caveat. As only the P-cores support the instruction set, you must disable the E-cores to enable AVX-512. Of course, that would seriously hamper the multi-core performance of the 12900K.

alder lake enable avx-512

Fortunately, our Core i5-12400 does not have E-cores. So, enabling AVX-512 is very much the definition of free performance.

However, as several media have reported, Intel has disabled the AVX-512 capability through updated microcode. On newer public BIOSes with the 0x18 microcode, you’ll no longer find AVX-512 support. That said, enthusiasts have figured out a way to inject the older microcode into newer BIOSes for those who want AVX-512 support.

As we’re still on an older BIOS with the 0x15 microcode, we still benefit from the AVX-512 opportunity.

ASUS MultiCore Enhancement

ASUS MultiCore Enhancement is a single BIOS option that removes all limits constraining the Turbo Boost 2.0 algorithm. Effectively, it allows the CPU to run at maximum turbo boost frequencies indefinitely.

Intel Turbo Boost 2.0 Technology allows the processor cores to run faster than the base operating frequency if the processor is operating below rated power, temperature, and current specification limits. The ultimate advantage is opportunistic performance improvements in both multi-threaded and single-threaded workloads.

The turbo boost algorithm works according to a proprietary EWMA formula. EWMA stands for Exponentially Weighted Moving Average. There are three parameters to consider: PL1, PL2, and Tau.

  • Power Limit 1, or PL1, is the threshold that the average power will not exceed. Historically, this has always been set equal to Intel’s advertised TDP. Importantly, PL1 should not be set higher than the thermal solution cooling limits.
  • Power Limit 2, or PL2, is the maximum power the processor can use for a limited amount of time.
  • Tau is a weighing constant used to calculate the moving average power consumption. Tau, in seconds, is the time window for calculating the average power consumption. If the average power consumed is higher than PL1, the CPU will reduce the CPU frequency.
intel turbo boost 2.0

A significant change from previous architectures is that Alder Lake is moving away from the TDP concept and instead has two power-related specifications:

  • Processor Base Power, formerly the TDP and PL1
  • Maximum Turbo Power, formerly PL2

This new format aligns with how the processor frequency has a base frequency and a maximum turbo frequency. For the 12400, the base power is 65W, and the maximum turbo power is 117W.

intel alder lake updated power definitions

Intel Extreme Memory Profile 3.0

Intel Extreme Memory Profile, or XMP, is an Intel technology that lets you automatically overclock the system memory to improve system performance. It extends the standard JEDEC specification that allows a memory vendor to program different settings onto the memory stick.

Intel Extreme Memory Profile 3.0 is the new XMP standard for DDR5 memory. It is based mainly on the XMP 2.0 standard for DDR4 but has additional functionality.

intel extreme memory profile xmp 3.0 ddr5

The XMP 3.0 standard is designed with six sections. One global section describes the generic data used across the profiles. The other five sections are designed for five profiles, respectively.

  • Profile 1 is for the performance profile (this is the standard XMP as we know it)
  • Profile 2 is for the extreme settings (this could be a higher frequency at higher voltage)
  • Profile 3 is for the fastest settings (this could be tighter timings at higher voltage)
  • Profiles 4 and 5 are rewritable and are for user custom settings

There’s a lot more to the new XMP 3.0 standard, but that falls outside the scope of this overclocking guide. If you’re interested in more details about XMP 3.0, feel free to check out my Alder Lake launch article.

Upon entering the BIOS

  • Enter the Extreme Tweaker menu
  • Set Ai Overclock Tuner to XMP I
  • Set ASUS MultiCore Enhancement to Enabled – Remove All Limits
  • Enter the AVX Related Controls submenu
  • Set AVX512 to Enabled

Then save and exit the BIOS.

We re-ran the benchmarks and checked the performance increase compared to the default operation.

  • SuperPI 4M: +1.08%
  • Geekbench 5 (single): +0.80%
  • Geekbench 5 (multi): +11.20%
  • Cinebench R23 Single: +2.34%
  • Cinebench R23 Multi: +10.74%
  • CPU-Z V17.01.64 Single: +4.73%
  • CPU-Z V17.01.64 Multi: +0.18%
  • V-Ray 5: +5.66%
  • AI Benchmark: +20.85%
  • 3DMark Night Raid: +0.96%
  • CS:GO FPS Bench: +2.86%
  • Final Fantasy XV: +3.12%
Core i5-12400 avx512 mce xmp benchmark performance

Here are the 3DMark CPU Profile scores at stock

  • CPU Profile 1 Thread: +7.93%
  • CPU Profile 2 Threads: +8.30%
  • CPU Profile 4 Threads: +0.91%
  • CPU Profile 8 Threads: +0.37%
  • CPU Profile 16 Threads: +0.07%
  • CPU Profile Max Threads: +1.98%
Core i5-12400 avx512 mce xmp 3dmark cpu profile performance

As expected, we see the most extensive performance difference in multi-threaded applications utilizing AVX instructions as these were previously constraint by the maximum long-term power of 65W. We see up to a 20.85% performance increase in AI Benchmark.

When running Prime 95 Small FFTs with AVX enabled, the average CPU P-core clock is 4000 MHz with 1.057 volts. The average CPU temperature is 64 degrees Celsius. The ambient and water temperature is 23.5 and 31.8 degrees Celsius. The average CPU package power is 110 watts.

Core i5-12400 avx512 mce xmp prime95 small ffts avx enabled

When running Prime 95 Small FFTs with AVX disabled, the average CPU P-core clock is 4000 MHz with 1.046 volts. The average CPU temperature is 57 degrees Celsius. The ambient and water temperature is 23.5 and 30.8 degrees Celsius. The average CPU package power is 95.9 watts.

Core i5-12400 avx512 mce xmp prime95 small ffts avx disabled

Now, we move on to actual overclocking.

I assume everyone watching this channel is familiar with Intel’s processor and chipset segmentation. In short, only specific CPUs and chipsets are enabled for overclocking. Generally speaking, the CPU SKUs include those with the suffix K or X (and their variants). Only Z and X chipsets allow for overclocking, though since the 500 series chipsets, memory overclocking is also available on the H and B chipsets.

The main difference between the CPUs enabled for overclocking, or “unlocked,” and those that are not, is the ability to change the ratio multipliers of various parts inside the CPU, including the CPU cores, Ring, integrated graphics, and so on. Furthermore, on “locked” or non-overclockable parts, the base clock frequency is restricted to 103 MHz. That effectively limits the CPUs to run up to 3% higher than the default specification.

The history of Intel’s locked and unlocked processors goes back a long time. Let’s take a trip down memory lane.

Brief and Incomplete History of Intel (Un)Locking CPUs

To give a processor its operating frequency, you need two things: a reference clock and a multiplier. For most modern CPUs, there’s a crystal on the motherboard used by either the CPU or chipset to produce an internal reference clock of 100 MHz. This 100 MHz is then multiplied using a PLL to achieve the final operating frequency. On Alder Lake, a 38.4MHz crystal is used to generate the 100MHz BCLK frequency, which then is multiplied by 53X to achieve the maximum Turbo Boost 3.0 frequency on the Core i9-12900K.

To lock a processor to a specific frequency, you, therefore, need to lock three parts: the crystal, the reference clock, and the multiplier.

how to lock a cpu

Intel’s Locking – The Beginning

Intel’s story of locking and unlocking CPUs for performance enthusiasts and overclockers begins in mid-1998.

Leading up to the crackdown on overclocking, Intel was upset at people increasing the processor frequency. There were two main reasons why Intel didn’t like this.

First, at the time, the operating frequency was a key selling point of the CPU as it was the main differentiator of compute performance. For every additional 33 MHz, Intel could charge a significant premium. Therefore, there was a direct financial impact when customers buy a lower-rated part and run it at the frequency of a more expensive part

Of course, today’s situation is very different as there is a wide variety of performance differentiating features such as core count, integrated graphics, and memory speed support.

Two, people set up businesses selling “re-marked” processors. Remarked processors are lower spec parts sold as higher spec parts. One could argue this is the same as pre-binned CPUs, though it’s not quite the same. A remarked CPU would be advertised and sold as a different product out of the box. In contrast, a pre-binned CPU is still advertised and sold as the original product but with a specific overclocking capability. Remarking was usually done by changing the exterior writing on the black SEC cartridge or, in some cases, by replacing the case entirely with a bogus one.

pentium II remarked processor

There was a significant financial incentive to remark processors. A 350nm Pentium II 233 Klamath processor was listed at US$636, while its bigger brother, the Pentium II 300, was listed at a whopping US$1981.

pentium II price list

However, things changed when the 250nm Deschutes Pentium II came to market. While early and engineering samples had an unlocked multiplier, most parts were multiplier-locked. For overclockers, that left only the reference clock to overclock the CPU.

The “soon overclocking will end” meme has been around for a long time, as proven by the opening lines of a 1998 Tomshardware article. The main reason for the feeling of desperation among PC enthusiasts was that Intel would not only multiplier-lock but also bus-lock its processors.

Despite the great fears in 1998, constraining the reference clock overclocking capabilities did not happen until much later.

Reference Clock Overclocking

That said, the Pentium III processors were all multiplier-locked. This enabled a great point of differentiation for motherboard vendors as the ability to increase the reference clock was now a key selling point for PC enthusiasts. The reference clock became one of the main battlegrounds for competitive overclocking and culminated with the LGA775 Wolfdale processors and the P45/X48 motherboards.

By 2010, the highest reference clock frequency had increased from less than 200MHz of the Pentium 3 Slot 1 to over 700 MHz. Incredibly, the world record of 750 MHz, set on October 30th, 2009, by Indonesian overclocker Benny Lodewijk, stood for more than a decade.

former world record reference clock frequency

Only last year, a Chinese overclocker Wytiwx beat the record. The highest ever reference clock frequency now stands at 766.1 MHz.

world record reference clock frequency

The front-side bus, or FSB, is a communication interface connecting the northbridge with the CPU. For the younger viewers among you, back in the old days, a motherboard would have two chips: the northbridge and the southbridge.

The northbridge would host not only the memory controller hub (with access to the system memory) but, in some cases, also links to high-speed graphics buses like AGP or PCI express.

The southbridge contained all connections to IO like low-speed PCI, SATA, USB, etc.

The northbridge would eventually integrate onto the CPU package with the release of the Nehalem architecture in 2008. Today, the southbridge is still on the motherboard — for example, the Z690 chipset.

The northbridge and CPU would connect with a high-speed front-side bus. As this bus transfers data between memory or graphics and the CPU core, it’s easy to understand that a higher frequency would undoubtedly provide much higher performance.

intel system architecture with front-side bus

For overclockers, a high FSB was one of the primary ways to improve system performance alongside higher CPU frequency and higher memory frequency. This is relevant to our story about locking and unlocking processors, and we’ll get back to it later.

But first, let’s return to the year 2003.

The Year 2003

2003 is a critical year for Intel for several reasons.

  • Intel’s NetBurst microarchitecture, which set out to achieve clock frequencies of up to 10GHz by 2011, turned out to be a wrong design choice as it ran into physical limitations (heat, power). As a result, Intel had to abandon the development of Tejas and Jayhawk products
  • The industry experienced a paradigm shift away from the frequency race and onto the multi-core race to chase performance improvements
  • AMD launched the Athlon 64, the first consumer 64-bit processor and the first processor with an integrated memory controller
  • Intel launched their very first Extreme Edition processor to counter the Athlon 64, which media generally saw as a marketing tactic to divert attention away from the significant strides forwards of their competitor

In short: Intel was getting beaten by their close rival in innovation and performance; had to abandon their mission to achieve 10 GHz and cancel ongoing CPU development; and was forced in a corner to fight back with an Extreme Edition, which was largely unappreciated by customers or media. In other words: 2003 was not a great year.

Intel’s first Extreme Edition processor, the Pentium 4 Extreme Edition 3.2 GHz, launched on November 3rd, 2003, about a week before AMD unveiled their first Athlon 64 processor with an integrated memory controller. The Extreme Edition is a 169 million transistor Pentium 4 running at 3.2 GHz with Hyper-Threading support. It also features a 2MB on-die L3 cache in addition to the standard 512KB on-die L2 cache.

first intel extreme edition processor

The L3 cache is the main point of differentiation the Gallatin core, which came down from the Xeon processor line, has over the mainstream Northwood core. The reason for adding a large L3 cache is to give the Pentium 4 as many of the benefits of an on-die memory controller without actually integrating one. A large L3 cache helps hide the overall memory latency by keeping more frequently used data in the L3 cache.

AnandTech informed Intel that a processor would be truly worthy of the Extreme Edition name if its multiplier were unlocked for overclocking. However, they suggested only offering lower multipliers to avoid the early Pentium II days’ remarking plague.

It wasn’t until the Pentium Extreme Edition 840, released on May 1st, 2005, that Intel would fully unlock the CPU multiplier. While there were engineering sample CPUs available without the CPU ratio lock, it was the first time in almost five years that enthusiasts could overclock without changing the reference clock frequency.

Intel would provide their customers with locked “regular” processors and unlocked “extreme edition” processors going forward. The Extreme Edition brand has lived on in Intel’s product portfolio until today. While a significant number of SKUs carry the X-suffix in their product name, only an exclusive few carry the Extreme Edition brand. The last Extreme Edition release is the Intel Core i9-10980XE Extreme Edition Cascade Lake processor for X299 motherboards launched on November 25th, 2019.

intel core i9-10980XE product spec

Unlocked? So, what!

Unlocking the CPU ratio multiplier offered Intel three key benefits.

  1. First, it allows Intel to claim feature parity with AMD’s FX product line, which offered an unlocked multiplier on the first Athlon 64 FX-51 processor launched in September 2003.
  2. Second, it helps justify charging a significant premium for the Extreme Edition processors as unlocked multipliers offer added value over locked processors
  3. Third, it enables Intel to fight at the top of the performance leaderboards, whether in media reviews or overclocking benchmark records.

It would’ve made a fantastic business model if the Extreme Edition were the perennial leaderboard top spot holder. However, Intel had a problem: reference clock overclocking.

As I pointed out earlier, there was heavy competition among motherboard vendors to offer the highest range of reference clock overclocking ever since Intel started locking the CPU multiplier in mid-1998. By the time Intel started rolling out its unlocked Extreme Edition processors in 2003, motherboards with the contemporary 865P chipset could run the reference clock frequency well above 300 MHz and thus offer more than 50% overclocking headroom.

intel slot 1 maximum reference clock

When Conroe came around in 2006, it was clear: the overclocking headroom of the reference clock and front-side bus frequency was more than enough to overcome the locked multiplier limitation. As a result, enthusiasts were buying the cheaper processors and would very easily overclock them to match or exceed the performance of the Extreme Edition processor.

intel core 2 overclocking

Even when it comes to the liquid nitrogen overclocking leaderboards, the Extreme Edition wasn’t necessarily the way to go. The fastest Conroe or Kentsfield processor is still a 5.9 GHz Core 2 Duo E6850, and the fastest Wolfdale or Yorkfield processor is a 6.9 GHz Core 2 Duo E8600. By the end of the LGA775 era, the reference clock overclocking capabilities exceeded 700 MHz, far above the default frequency of 333 MHz, and offering more than enough headroom for any processor.

Nehalem’s Turbo Mode and Power Control Unit

When Nehalem and the Core i7 processors came around in November 2008, things slowly started shifting away from the headroom offered by reference clock overclocking. Or, we should call it the BCLK or base clock frequency now as per Intel terminology. The base clock was significantly reduced to 133 MHz, and the front-side bus technology made way for Intel’s Quick Path Interconnect technology. Most importantly, Nehalem introduced two important technologies: Turbo Mode and the Power Control Unit.

Turbo Mode is a technology that allows the CPU cores to run faster than the base frequency if there’s sufficient power, thermal, and current headroom. The first version implemented in Nehalem was quite rudimentary in function, but later iterations would provide Intel with a robust product differentiator.

intel nehalem turbo mode

The Power Control Unit is a one million-plus transistor on-die microcontroller managing the CPU cores. It has its firmware and uses inputs like temperature, current, power, and OS requests to manage the CPU core frequency and power. Long story short: the PCU controls the CPU behavior.

intel nehalem pcu power control unit

Together, these two technologies would play a pivotal role in the coming crackdown on overclocking. I’ll get to that part of the story in a minute but first, let’s wrap up the reference, ehr, base clock story.

The Nehalem platform for desktop was split into two distinct segments: high-end desktop (Bloomfield) and mainstream desktop (Lynnfield). While Bloomfield processors generally had a base clock hard wall at about 220 MHz which could be overcome by increasing the PCIe frequency, Lynnfield suffered less from a BCLK overclocking limitation. While the maximum frequency dropped significantly from the record highs of over 750MHz to below 300MHz, the base clock frequency range was still more than sufficient.

intel nehalem bloomfield maximum reference clock

There is perhaps, no better example than HWBOT’s Wprime 1024M leaderboard ranking, which has a mix of Core i7 Extreme Edition, Xeon, and the cheapest Bloomfield topping the charts.

The BCLK overclocking range improved with Nehalem’s successor Westmere which had Gulftown for high-end desktop and Clarkdale for mainstream desktop. While the Gulftown 980X and 990X were the way to go for multi-threaded benchmarks, a Core i7 970 could still be competitive given your motherboard would provide sufficient base clock headroom.

However, when it comes to raw overclocking capability and maximum frequency – which was very important in most contemporary competitive benchmarks – the cheapest Core i5 and Core i3 Clarkdale processors would be right up there above 7 GHz with the most expensive Core i7s thanks to the BCLK overclocking headroom.

But then came Sandy Bridge, and everything came to a screeching halt. No more BCLK overclocking.

Sandy Bridge Overclocking Controversies

Page 5 of AnandTech’s Sandy Bridge Preview article is titled “Overclocking Controversy” and accurately captures the sentiment of the PC enthusiast community. Anand points out,

Ever since before the Pentium III; Intel had aspirations of shipping fully locked CPUs. But while multipliers were locked, Intel left FSB overclocking open as that would be an end-user or system integrator decision and not something that could be done when selling an individual CPU.

Three things had changed since Intel’s decision to lock the CPU multiplier

  1. First, through the Extreme Editions, Intel had begun experimenting with multiplier-unlocked processors and had been able to evaluate the impact on its business from various angles: the marketing benefit, the financial benefit, and impact on return rates, for example.
  2. Second, starting from Penryn and later adopted in Nehalem, with Turbo Boost, Intel now has a way to exploit the overclocking headroom out of the box.
  3. Third, with the integration of the Power Control Unit, Intel had a much tighter grip on everything related to overclocking. While Nehalem still relied on an external clock generator provided by the motherboard, Intel integrated the clock generator for Sandy Bridge in the PCH.

Remember this last point, as it will play an important role when we finally get to Alder Lake non-K overclocking.

As it would turn out, the BCLK overclocking capabilities of Sandy Bridge processors were extremely limited as it was tied to the DMI and PCIe clocks. The best BCLK overclock is still only 111.86 MHz, not even 12% higher than the base frequency of 100 MHz.

sandy bridge overclocking architecture

So, just like Tomshardware had predicted in December 1998: overclocking was dead. Right?

Not quite. Intel had made several important decisions that would keep overclocking alive and well.

Since 2009, the company has experimented with much more affordable multiplier unlocked processors. These weren’t Extreme Editions, but K processors. The first K processor was the Pentium E6500K, a multiplier unlocked Wolfdale processor launched exclusively in China.

Following what I assume had been a successful experiment, in May 2010, Intel introduced the Core i7-875K Lynnfield and Core i5-655K Clarkdale processors. While the Core i5 model was priced $40 higher than its locked counterpart, surprisingly, the Core i7 model was priced significantly lower than its locked counterpart.

core i7-875k & i5-655k overclocking

With the launch of Sandy Bridge, Intel introduced the K business model to the broader market: a slightly higher priced but fully unlocked model alongside a locked counterpart.

So, overclocking was saved, right? Well, sort of.

Sandy Bridge’s overclocking controversy didn’t end with the limited base clock overclocking. There were several more overclocking-related issues.

First, Intel forced the customer to choose between overclocking or using integrated graphics. The P67 chipset would support CPU overclocking but wouldn’t allow for integrated graphics (and its useful QuickSync acceleration). The H67 chipset would enable full use of the integrated graphics but disabled the overclocking capability. So, if you bought a Core i7-2600K, you had to pick: OC or IGP.

sandy bridge overclocking support

Fortunately, Intel swiftly addressed this controversy by introducing the first Z-chipset: the Z68. Z68 would allow you to use the integrated graphics as well as enjoy the overclocking capabilities of your CPU

Second, CPU ratio overclocking did not go very smoothly at all. Most CPUs would initially be limited to the 53X ratio. Only after introducing an option called Internal PLL overvoltage override could you increase the CPU ratio beyond 53X.

Funny side note: this particular issue also existed on Sandy Bridge-E processors but curiously worked slightly differently. I made a video 11 years ago detailing how to use this setting to get a much higher frequency on ambient cooling.

Third, the CPU ratio range was minimal as it would only go up to 57X. That was too limited for extreme overclockers as plenty of Sandy Bridge CPUs would overclock well past 5.9GHz.

Lastly, memory overclocking was also incredibly limited as Sandy Bridge offered memory ratios up to only DDR3-2133. As a result, the highest memory frequency ever achieved on Sandy Bridge processors is around DDR3-2368. That’s significantly lower than the DDR3-3200+ its mainstream predecessor Lynnfield was capable of.

sandy bridge max memory clock

All things considered, while the K-SKU saved overclocking on the mainstream platform, it was indeed severely restricted.

Intel had made an additional decision to unlock every processor on its high-end desktop platform fully. Unlocking also meant the creation of BCLK multipliers which allowed you to run at 100, 133, 166, or 250 MHz base clock frequency.

intel sandy bridge-e overclocking architecture

But for mainstream non-K processors, the overclocking fun was over. From Sandy Bridge to Alder Lake, intel has imposed restricted overclocking. For Sandy Bridge, Ivy Bridge, and Haswell, the maximum base clock frequency for non-K locked processors is limited to around 111 MHz.

Whereas for unlocked K processors, the maximum BCLK frequency is much higher.

sandy bridge ivy bridge haswell bclk overclocking records

Skylake Oopsie

Before I get back to the Core i5-12400, I have to mention Skylake as there are striking parallels with the non-K overclocking situation on Alder Lake.

Skylake included a new feature called BCLK Governor, an integrated circuit that calculates BCLK real-time and will issue a machine check error if BCLK is higher than the allowed limit. Importantly, BCLK calculation doesn’t rely on using the Integrated Clock Controller (ICC) from the PCH as it works with external clocks as well.

intel bclk governor

External clocks? Yes, another key new feature on Skylake was the re-enabling of full range fine-grain BCLK overclocking. For this purpose, motherboard vendors could add an external clock generator to the motherboard, bypass the internal clock generator, and use it as the base clock frequency. The result was quite spectacular as the BCLK could overclock to over 550 MHz, the highest since the Core 2 days.

asus maximus viii series overclocking records

With this kind of BCLK overclocking capability, Intel needed their BCLK Governor is to measure the actual BCLK frequency and prevent the CPU from operating if it exceeds the hard-coded limitation.

On December 2nd, 2015, almost four months after the launch of Intel’s 6th generation processors and brand-new CPU architecture, Skylake, the overclocking world woke up to a Core i3-6320 processor with an overclocked BCLK frequency of 120 MHz on a Supermicro H170 motherboard.

As we’d learn in the days and weeks to come, Supermicro had found a way to work around the Power Control Unit which is responsible for managing the CPU’s power consumption, performance, as well as housing the BCLK Governor. With the PCU out of the way, there was no way to check the BCLK frequency and throw an error. So, non-K overclocking was back.

Unfortunately, disabling the PCU came with a couple of significant drawbacks. For example:

  • C-states are disabled; therefore, the CPU will constantly run at its highest frequency and voltage
  • Turbo-boost is disabled
  • Integrated graphics are disabled
  • AVX2 instruction performance is poor, approximately 4-5 times slower due to the upper 128-bit half of the execution units and data buses not being taken out of their power-saving states
  • CPU core temperature readings are incorrect
skylake non-k overclocking issues

Despite the drawbacks, some motherboard vendors went all-in on promoting non-K overclocking. Most notably, ASRock set up an entire marketing campaign for its SKY OC feature.

As expected, Intel pressured motherboard partners to remove non-K overclocking support, and most motherboard partners withdrew official support. That said, extreme overclockers and enthusiasts could still find and use the old BIOSes with non-K OC support.

The BCLK Governor measures the real-time BCLK frequency and checks it against a hard-coded limitation. Since Skylake, the BCLK limit is 103 MHz.

So, while the highest non-K BCLK frequency for Skylake was 212.99 MHz, ever since Skylake, the non-K maximum BCLK frequency is about 103 MHz.

skylake kaby lake coffee lake comet lake rocket lake non-k maximum bclk

And here’s where we pick up our overclocking guide.

OC Strategy #2: Conventional Non-K BCLK Overclocking

In our second overclocking strategy, we overclock the Core i5-12400 processor using the conventional method with base clock frequency. We set the BCLK frequency to 102.9 MHz to see the performance with the traditional overclocking restrictions in place.

Upon entering the BIOS

  • Enter the Extreme Tweaker menu
  • Set Ai Overclock Tuner to XMP I
  • Set BCLK Frequency to 102.9
  • Set ASUS MultiCore Enhancement to Enabled – Remove All Limits
  • Enter the AVX Related Controls submenu
  • Set AVX512 to Enabled

Then save and exit the BIOS.

We re-ran the benchmarks and checked the performance increase compared to the default operation.

  • SuperPI 4M: +2.51%
  • Geekbench 5 (single): +8.19%
  • Geekbench 5 (multi): +12.43%
  • Cinebench R23 Single: +9.70%
  • Cinebench R23 Multi: +12.02%
  • CPU-Z V17.01.64 Single: +6.70%
  • CPU-Z V17.01.64 Multi: +2.04%
  • V-Ray 5: +7.29%
  • AI Benchmark: +25.25%
  • 3DMark Night Raid: +6.04%
  • CS:GO FPS Bench: +3.48%
  • Final Fantasy XV: +4.05%
Core i5-12400 conventional overclock benchmark performance

Here are the 3DMark CPU Profile scores at stock

  • CPU Profile 1 Thread: +10.00%
  • CPU Profile 2 Threads: +10.71%
  • CPU Profile 4 Threads: +1.24%
  • CPU Profile 8 Threads: +0.84%
  • CPU Profile 16 Threads: +2.09%
  • CPU Profile Max Threads: +4.45%
Core i5-12400 conventional overclock 3dmark cpu profile performance

Up until this point in the article, we’ve used performance tuning options that should be available on any Intel Z690 motherboard. The performance improvement is, therefore, quite impressive. Even with a small bump in base clock frequency and enabling AVX512, MCE, and XMP, we’re seeing the single-threaded performance increase up to 10% in 3DMark CPU Profile and multi-threaded performance increase up to 25.25% in AI Benchmark.

When running Prime 95 Small FFTs with AVX enabled, the average CPU P-core clock is 4013 MHz with 1.039 volts. The average CPU temperature is 61 degrees Celsius. The ambient and water temperature is 22.7 and 30.9 degrees Celsius. The average CPU package power is 103.2 watts.

Core i5-12400 conventional overclock prime95 small ffts avx enabled

When running Prime 95 Small FFTs with AVX disabled, the average CPU P-core clock is 4013 MHz with 1.021 volts. The average CPU temperature is 54 degrees Celsius. The ambient and water temperature is 22.5 and 30.0 degrees Celsius. The average CPU package power is 90 watts.

Core i5-12400 conventional overclock prime95 small ffts avx disabled

OC Strategy #3: Non-K Overclock + Fixed CPU Ratio & Voltage

In our third overclocking strategy, we will finally actually overclock the non-K Alder Lake processor.

To understand how non-K overclocking works, I need to explain a little more about how Intel enables overclocking on its CPUs. Let’s pick up the story at Nehalem.

Intel’s Power Control Unit

As I already mentioned, the integrated Power Control Unit was a key innovation of the Nehalem processor as it provided Intel control over the CPUs power management. Via its integrated microcontroller, Intel could set the power and performance rules of its processor, govern it, and if necessary, even change its behavior at a later stage through BIOS updates.

intel nehalem pcu power control unit

With Sandy Bridge, Intel also added an integrated clock generator and essentially cut off any external control over the base clock frequency. Overclocking control was not almost entirely under Intel’s control.

While tight control was not a big problem for most people inside Intel, an internal group of overclocking enthusiasts realized actions were necessary to ensure overclocking support was not only maintained but also further improved.

In a 2013 interview, Mike Moen from Intel explained the origins and the purpose of this overclocking workgroup.

While Mike has since left Intel, his successor Dan Ragland now heads the Intel Overclocking Lab. He and his team continue to push the overclocking agenda within the organization. For those who want to get to know Dan, both Intel (1, 2) and Techtechpotato have excellent interviews on their respective YouTube channels.

Where Nehalem introduced features that would end up severely restricting overclocking on Intel platforms, Haswell introduced perhaps the single most consequential feature for overclocking most enthusiasts have never heard about: the OC Mailbox.

Intel’s OC Mailbox

Haswell is a pivotal architecture for Intel overclocking as it introduces the OC Mailbox. The OC Mailbox is a Model Specific Register, or MSR, interface which communicates directly with the CPU Power Control Unit. It is designed to program overclocking and overvolting parameters to the CPU.

intel oc mailbox haswell

Over the years, the OC Mailbox has replaced many legacy overclocking programming functions and significantly expanded its scope.

The original OC Mailbox commands were limited to

  • Reading the overclocking capabilities of the CPU core, Ring, and Uncore (0x01). The capabilities include the fused max ratio limit, whether overclocking is supported, whether voltage override is supported, and whether voltage offset is supported.
  • Reading the overclocking per core ratio limit capability (0x2)
  • Reading and writing voltage and frequency overrides (0x10/0x11)
  • Reading and writing SVID override for internal voltage regulator (0x12/0x13)
  • Reading and writing miscellaneous overclocking commands like FIVR controls (0x14/0x15)

On Alder Lake, the OC Mailbox has expanded to include support for more domains (CPU P-core, E-core, Ring, Graphics, System Agent) and many more functions (from 5 in Haswell to 20+ in Alder Lake).

From a business perspective, Intel uses the OC Mailbox in two main ways:

  1. One, to increase value for enthusiast K-SKU and Z-chipset customers by enhancing and expanding overclocking and performance tuning functionality
  2. Two, to gatekeep and restrict overclocking for non-K and non-Z products

Therefore, in theory, the OC Mailbox should be locked unless you have a -K or -X processor and a Z- or X- chipset. This has a significant impact on the non-K overclocking experience. It means:

  • No voltage offset or override
  • No turbo ratio control beyond the default maximum limit
  • No per core ratio limit
  • No VF Points
  • No BCLK Aware Adaptive Voltage
  • No Thermal Velocity Boost functions or OCTVB
  • No AVX offsets
alder lake non-k no oc mailbox

To the untrained eye, it may seem that there’s nothing special about the unlocked parts. But once you know the tremendous value created through the efforts of expanding the OC Mailbox, you start to appreciate the work Intel’s OC Lab team is doing to enable overclocking.

Furthermore, you can understand why it makes more than sense for Intel to charge a premium for these advanced features.

intel K price premium

Do note that access rules to the OC Mailbox can be enabled and disabled by Intel how they wish. So, they could allow parts of the OC Mailbox to work only on specific SKUs or disable access altogether.

An important note is that when enabling non-K overclocking for Alder Lake CPUs, the OC Mailbox status and access do not change. That means non-K Alder Lake processors have no access to the tools provided via the OC Mailbox.

So, let’s finally get to the point of this article. With all these restrictions in place, how on earth is it possible that Alder Lake non-K overclocking works?

Alder Lake CPU Microcode 0x9

In a previous segment, I’ve mentioned the BCLK Governor feature. The BCLK governor was introduced on Skylake processors and is specifically designed to restrict out-of-range BCLK frequencies. Out of range, in this case, means up to 103 MHz for the locked non-overclockable parts and unlimited for the unlocked overclockable parts.

The BCLK governor is a standard feature of the CPU microcode and monitors both the internal and external clock generator. Due to an apparent oversight, the BCLK governor does not function correctly on a specific older pre-release version of the CPU microcode 0x9.

The BCLK governor works correctly for the integrated clock generator but not for the external clock generator. Using this CPU microcode in combination with an external clock generator means you can increase the base clock frequency on a locked non-overclockable part beyond the artificial limitation of 103 MHz.

We can only speculate what caused this oversight. One theory is that this happened due to a fundamental change in the clocking structure. Since integrating the clock generator on Sandy Bridge, it’s always been the PCH that generates the 100MHz BCLK frequency. That was the case until Rocket Lake. However, first on Tiger Lake and now also on Alder Lake, BCLK is added on the CPU die. The 38.4 MHz crystal still connects to the PCH, but instead of driving a 100MHz clock to the CPU, it now just forwards that 38.4 MHz clock to the CPU BCLK PLL. The CPU, in turn, generates its 100MHz reference clock. This fundamental change in clocking structure may explain why the oversight occurred.

alder lake overclocking architecture

Later microcode versions don’t have this oversight, and non-K BCLK overclocking is restricted.

Intel would prefer non-K parts not to be overclocked via unsupported methods. I don’t believe that’s primarily because of the risk of financial loss – we’ll get back to that later – but mainly for security reasons. After all, if you’re using a non-K processor, you wouldn’t want an external actor to be able to make it unstable by accessing the external clock generator.

Considering this exploit is not present in the newer microcode, I don’t expect Intel to come down too harshly on this issue. Intel can encourage motherboard partners to ship newer BIOSes without this older microcode or could even enforce updating the CPU microcode via a Windows Update as it did for the unlocked Pentium G3258 Anniversary Edition and non-Z motherboards.

Motherboard vendors have a couple of options to support non-K BCLK overclocking.

  1. First, they can provide specific BIOS releases with microcode 0x9 to end-users.
  2. Second, they can release BIOSes with both microcodes and allow users to choose which microcode to load.

That’s how it works on the ASUS ROG Maximus Z690 Extreme motherboard. By enabling the “Unlock BCLK OC” option, the BIOS will load the 0x9 microcode. If this feature is disabled, it will load the latest microcode. That’s 0x15 on my 0801 BIOS.

asus z690 enable non-k bclk oc

Just as a reminder: while on microcode 0x9, the BCLK governor for the external clock generator doesn’t work correctly but enabling non-K overclocking does not provide access to any of the other functions available in the OC Mailbox.

So even though there’s the ability for non-K overclocking, we are severely restricted in the options we have at our disposal to make it work correctly for a daily system.

So, what do we have to work with? We only need two things when overclocking: more frequency and more voltage. To get a thorough understanding of Alder Lake’s configuration, let’s have a closer look at its frequency and voltage topology.

Intel Alder Lake Clocking Topology

The clocking of Alder Lake is more similar to Tiger Lake than it is to Rocket Lake as it inherits the CPU internal clock generator from Tiger Lake.

alder lake clocking topology

The standard Alder Lake platform has a 38.4MHz crystal as a reference clock to the PCH. The PCH will then generate three clocks:

  1. 38.4 MHz reference clock for the CPU internal clock generator
  2. 100MHz PCIBCLK for PCIe, DMI, and I/O
  3. 24MHz frequency for TSC, display, and SVID controller

The CPU internal clock generator then generates the 100MHz base clock frequency used for all the parts inside the CPU. That is different from Rocket Lake, where the PCH PLL would generate the 100MHz base clock frequency with no interference from the CPU.

However, just like Rocket Lake, it is possible to bypass the internal clock generator and use an external clock generator to feed the 100MHz BCLK frequency to the CPU. There doesn’t seem to be any apparent difference between using the internal or external clock generator. However, motherboard vendors will argue the external clock generator will offer greater granularity and range.

Whichever way you get the 100 MHz BCLK, this clock is multiplied with specific ratios for each of the different parts in the CPU.

Each P-core has its PLL and can run at its independent frequency. That is pretty much identical to Rocket Lake. However, things are a bit different for the E-cores as they are grouped in clusters of 4. The maximum supported ratio is up to 120X though I doubt anyone will ever get there.

While the technical documentation references a different ratio for the E-cores L2 cache, it appears that the Atom L2 cache frequency is linked to the Atom cluster.

The Ring frequency ties together the Ring, last-level cache (L3), and CBo or Cache box. The Ring ratio can go up to 85X, but again I don’t think anyone will need ratios that high.

The GT frequency or graphics frequency uses the same 100MHz BCLK but first divides it by two and then multiplies it with the GT ratio. The GT ratio can go up to 42X, and while on Rocket Lake, this was more than sufficient to achieve the highest OC frequency; on Alder Lake, that’s not quite the case, as I demonstrated in SkatterBencher #33.

Unlike Rocket Lake, on Alder Lake, the Slice and Unslice frequency is decoupled. That means they’re running at different frequencies. The default maximum boost frequency of the Slice is 1550 MHz, and the default maximum frequency of the Unslice is 1350 MHz. However, like on Rocket Lake, we can only control the Slice frequency using the Graphics Ratio.

The System Agent ratio is either 16X or 32X.

Finally, the two memory controllers and system memory are linked and driven by the same 100 MHz BCLK; however, they also support 133 MHz as base frequency. The 133 MHz is derived from the 100MHz BCLK, which is first multiplied by four then divided by 3, and the memory controller drives the clock frequency for the system memory.

The Alder Lake memory controller supports gear down mode, which, when enabled, effectively halves the memory throughput. In gear-down mode, the memory controller frequency is also halved and thus runs at half the system memory frequency. On Alder Lake, there’s now an additional gear down – Gear 4 – which makes the memory controller frequency run at a quarter of the memory speed. This can be useful for DDR5 overclocking.

The memory ratio goes up to 63X and, combined with the 133 MHz frequency, thus results in up to DDR4-8400 or DDR5-16800. That said, there’s also an option that allows near unlimited DDR frequency requests up to a theoretical DDR-34000

When we overclock using the base clock frequency, we affect the operating frequency of all the parts driven by that frequency. I think it’s evident to everyone watching that lacking access to the multipliers higher than the default maximum is a significant downside of non-K overclocking.

Intel Alder Lake Voltage Topology

While Alder Lake clocking resembles Tiger Lake more than Rocket Lake, things are not quite that similar in the voltage department.

Alder Lake transitions away from using FIVR for the Cores, Ring, and integrated graphics compared to Tiger Lake. Instead, Alder Lake uses power gates. However, unlike Rocket Lake, some parts of the Alder Lake CPU are powered by FIVR.

alder lake vs tiger lake fivr

If we only consider the Alder Lake CPU and disregard the chipset, there are seven different voltage inputs.

Let’s go over them one by one.

  • First, the easiest one: VccGT. This voltage powers the GT or integrated graphics, including the Slice, Unslice, and Display block. It supports both override and adaptive voltage mode
  • Second, VccIA. IA stands for Intel Architecture and is the collective term for all Intel CPU cores. Therefore, logically, the VccIA powers both the P-cores and, if present, E-cores. It also powers the CLR, including the Ring, Last-level Cache, and Cache Boxes. All parts will run at the same voltage as this is a single rail. VccIA supports both override and adaptive voltage mode
    • Note that the VccIA voltage goes through a power gate before reaching the P-cores and E-cores. This enables turning off voltage to specific cores for power-saving and core disabling purposes.
  • Third, VccInAux. Here’s where the voltage gets more complicated than Rocket Lake. VccIn is the input voltage for the fully integrated voltage regulators. The FIVRs power many parts inside the CPU, including the E-core L2 cache, Uncore (VccSA), etc.
  • Fourth, the VDD2 voltage rail is directly providing a voltage to the memory data pins on the processor, which are for the DDR PHY of the memory controller
  • Fifth, VccMipiLP drives some of the analog parts.
  • Sixth, the Vcc1p8cpu drives the parts that require a 1.8V supply
  • Lastly, the Vcc1p05CPU drives the parts of the CPU that need a 1.05V supply.
alder lake voltage topology

To understand the limitations of non-K overclocking, you also have to know how Intel CPUs manage these voltage rails. Let’s first focus on the VccIA voltage rail.

The VccIA rail is powered by an external VRM which connects to the PCU to power the P-cores, E-cores, and Ring. The PCU allows two modes of operation: adaptive mode and override mode.

In adaptive mode, the CPU issues VID voltage requests to the VRM according to the factory-fused voltage-frequency curves and power management. Simply put: if the CPU boosts to the highest frequency, it needs to request a higher voltage from the motherboard voltage controller. You can also add a voltage offset to whatever VID voltage the CPU requests.

In override mode, the CPU uses a fixed voltage for all scenarios. It will still issue VID requests to the VRM controller, but the VRM controller can either follow this request or send whatever voltage it wants. That is how manual voltage configuration from the BIOS typically works: you configure the VRM controller rather than the CPU voltage request.

Both VccIA and VccGT voltage rail communicate with the VRM controller using the SVID IMPV9.1 standard. The end-user can control both voltage rails using the OC Mailbox functionality.

While most of the OC Mailbox functions are not available for non-K processors, it looks like at least the VccIA voltage control is available though I didn’t check the VccGT.

Surprisingly, it’s also possible to control the VccSA system agent voltage via the OC Mailbox on my system. It is surprising because there are reports of other people trying non-K overclocking who cannot change the VccSA voltage. Furthermore, since the VccSA rail is not powered by a separate voltage regulator but instead is powered via the FIVR, it should technically be entirely unavailable for non-K overclocking.

At the moment of writing, it’s unclear why VccSA is available on my system and not on others.

VccSA is primarily important for DDR4 overclocking to use high-frequency memory with Gear 1 enabled. Gear 1 forces the memory controller to run at the same frequency as the system memory, and increased VccSA helps improve stability. On DDR5, VccSA is of lesser importance as it can only run Gear 2, and thus the memory controller is not clocked that high.

Now that we understand the limitations of non-K overclocking let’s get to work.

In this overclocking strategy, I will use a traditional approach to overclocking by using a fixed CPU ratio and a fixed CPU voltage.

Higher frequency is something we can get by increasing the BCLK frequency. That will increase the frequency of all the connected domains inside the CPU, including the CPU P-cores, Ring, memory controller, system memory, system agent, and integrated graphics. Fortunately, while we can’t increase the ratios beyond the maximum default, we can reduce them. So, when increasing the BCLK, we can adjust the ratios for all those domains downwards to tune for stability.

We don’t have access to important tools provided via the OC mailbox, including AVX negative offset, so our worst-case scenario stability test will limit our maximum voltage, Prime 95 Small FFTs with AVX512 enabled.

When it comes to the voltage, we’re in luck too. We can use the standard methods to use either a fixed voltage provided directly from the motherboard VRM in override mode or add an offset to the standard V/F curve in adaptive mode. For this strategy, we stick with override mode.

Upon entering the BIOS

  • Enter the Extreme Tweaker menu
  • Set Ai Overclock Tuner to XMP I
  • Set BCLK Frequency to 122.5
  • Set ASUS MultiCore Enhancement to Enabled – Remove All Limits
  • Set DRAM Frequency to DDR5-6207MHz
  • Set Performance Core Ratio to Sync All Cores
    • Set 1-Core Ratio Limit to 40
  • Enter the AVX Related Controls submenu
    • Set AVX512 to Enabled
  • Leave the AVX Related Controls submenu
  • Enter the Digi+ VRM submenu
    • Set CPU Load-line Calibration to Level 8
  • Leave the Digi+ VRM submenu
  • Enter the Internal CPU Power Management submenu
    • Set Regulate Frequency by above Threshold to Disabled
  • Leave the Internal CPU Power Management submenu
  • Enter the Tweaker’s Paradise submenu
    • Set Unlock BCLK OC to Enabled
  • Leave the Tweaker’s Paradise submenu
  • Set Max. CPU Cache Ratio to 40
  • Set CPU Core/Cache Voltage to Manual Mode
    • Set CPU Core Voltage Override to 1.175

Then save and exit the BIOS.

We re-ran the benchmarks and checked the performance increase compared to the default operation.

  • SuperPI 4M: +11.41%
  • Geekbench 5 (single): +16.81%
  • Geekbench 5 (multi): +29.27%
  • Cinebench R23 Single: +21.80%
  • Cinebench R23 Multi: +33.63%
  • CPU-Z V17.01.64 Single: +19.80%
  • CPU-Z V17.01.64 Multi: +25.03%
  • V-Ray 5: +31.13%
  • AI Benchmark: +44.26%
  • 3DMark Night Raid: +24.22%
  • CS:GO FPS Bench: +4.02%
  • Final Fantasy XV: +4.94%
Core i5-12400 fixed overclock benchmark performance

Here are the 3DMark CPU Profile scores at stock

  • CPU Profile 1 Thread: +22.18%
  • CPU Profile 2 Threads: +26.25%
  • CPU Profile 4 Threads: +25.47%
  • CPU Profile 8 Threads: +25.90%
  • CPU Profile 16 Threads: +25.71%
  • CPU Profile Max Threads: +27.27%
Core i5-12400 fixed overclock 3dmark cpu profile performance

There’s not much to say about these results. Clearly, increasing the clock frequency from an all-core base of 4 GHz to 4.9GHz will positively impact the performance. We see performance improvements across the board with a maximum gain of +44.26% in AI Benchmark.

When running Prime 95 Small FFTs with AVX enabled, the average CPU P-core clock is 4900 MHz with 1.180 volts. The average CPU temperature is 93 degrees Celsius. The ambient and water temperature is 23.7 and 33.5 degrees Celsius. The average CPU package power is 156.7 watts.

Core i5-12400 fixed overclock prime95 small ffts avx enabled

When running Prime 95 Small FFTs with AVX disabled, the average CPU P-core clock is 4900 MHz with 1.180 volts. The average CPU temperature is 82 degrees Celsius. The ambient and water temperature is 23.9 and 33.0 degrees Celsius. The average CPU package power is 139.4 watts.

Core i5-12400 fixed overclock prime95 small ffts avx disabled

OC Strategy #4: Non-K Overclock + Dynamic CPU Ratio & Voltage

In the fourth and final overclocking strategy, we use a modern approach to overclock using a dynamic Turbo frequency and the dynamic Adaptive voltage mode.

By Core Usage CPU Ratio

Generally speaking, there are two ways to configure the CPU ratio on Intel platforms manually: Sync all cores and By Core Usage.

Sync All Cores sets one ratio that is applied to all cores. That is very much the traditional way of overclocking and our previous overclocking strategy approach.

By Core Usage allows us to configure the overclock for different scenarios ranging from 1 active core to all active cores. That enables us to run some cores significantly faster than others when the conditions are right.

Note that By Core Usage is not the same as configuring each core specifically. Using By Core Usage, we determine an overclock according to the actual usage. For example, if a workload uses four cores, then the CPU will determine which cores should execute this workload and apply our set frequency to those cores.

intel sync all cores and by core usage

Intel’s default Turbo Ratio 2.0 configuration works precisely like this. The Core i5-12400 default configuration allows for up to 4.4GHz when two cores are active, up to 4.2GHz when four cores are active, and up to 4 GHz when all cores are active.

The main benefit of using by core usage CPU ratios is that we can extract a lot more performance when fewer than all cores are active. For our configuration, we stick with the default turbo ratios of 44X up to 2 cores, 42X up to 4 cores, and 40X up to 6 cores.

Adaptive Voltage Mode

The VccIA input drives the voltage for both the CPU cores and Ring. That means you can only set a single voltage for these parts of the CPU.

The voltage can be set in two ways: adaptive or override.

  • Override mode specifies a single static voltage across all ratios. It is for extreme overclocking purposes where stability at very high frequencies is the only consideration.
  • Adaptive mode is the standard mode of operation. In Adaptive Mode, the V/f curve used is generated automatically by the CPU and covers the CPU ratios from the lowest supported ratio to the default maximum turbo ratio. In the case of the 12400, that’s from 8x to 44x.

V/f curve stands for voltage-frequency curve. The V/f curve determines which voltage the CPU should set for a certain frequency. Since some cores are better than others, it’s possible to see different voltages for the same frequency across the cores of your CPU.

Using adaptive voltage offset, the entire V/F curve can be offset by up to 500mV in both directions.

Also, since Comet Lake, Intel has extended the Adaptive Voltage function with an Advanced Voltage Offset. It is implemented in the ASUS bios as V/f Point Offset. V/f point offset allows the user to change the default V/f curve by offsetting the voltage at specific frequencies on the V/f curve. Unfortunately, this function is not enabled on non-K CPUs.

There are three steps to how your system sets the CPU voltage in Adaptive Mode.

  • First, the motherboard’s BIOS informs the processor about the current loadline characteristics via AC DC loadline values.
  • Then, the CPU will request a voltage from the voltage controller based on its programmed V/f curve and the motherboard loadline characteristics.
  • Finally, the voltage that reaches the CPU is the requested voltage minus any undershoot or overshoot from the VRM loadline.

The AC DC loadline characteristics allow the motherboard to inform the CPU about the VRM design. The CPU will factor in a certain voltage droop when requesting a VID based on the specific design. Voltage droop is the decrease of voltage when a core goes from idle to full load.

intel adaptive voltage

The VRM loadline setting determines how much the output voltage increases or decreases when the CPU goes from low to high load or vice versa. Simply put, a significant undershoot or overshoot can result in an unstable system. So VRM loadline helps to mitigate this problem.

On non-K processors, it’s impossible to change any point on the voltage-frequency curve. We can only affect the CPU voltage by using the override function to have a fixed voltage or by using the offset function to move the entire V/F curve up or down.

In this overclocking strategy, we use Offset mode to increase the voltage. Our offset of choice is +200mV. The main challenge with this method is that it increases the voltage across the entire V/F curve from 8X to 44X.

For your reference, I’ve mapped the voltage-frequency curve of this Core i5-12400 processor with and without voltage offset. As you can see in this graph, the voltage for 8X has increased from 0.773 volts to 0.968 volts, and the voltage for 44X increased from 1.128 volts to 1.305 volts.

We move the entire curve when we adjust the BCLK frequency to 120 MHz. Whereas our original offset voltage-frequency curve uses 1.305 volts at 4 GHz, the BCLK adjusted curve maps a voltage of about 1.18 volts to 4GHz.

Just for fun, I also include the V/F curve of the 12900KF from SkatterBencher #34. As you can see, the Core i9 requires substantially lower voltage than the Core i5.

core i5-12400 v/f curve

Upon entering the BIOS

  • Enter the Extreme Tweaker menu
  • Set Ai Overclock Tuner to XMP I
  • Set BCLK Frequency to 120.5
  • Set ASUS MultiCore Enhancement to Enabled – Remove All Limits
  • Set DRAM Frequency to DDR5-6266MHz
  • Set Performance Core Ratio to By Core Usage
    • Set 1-Core Ratio Limit to 44
    • Set 2-Core Ratio Limit to 44
    • Set 3-Core Ratio Limit to 42
    • Set 4-Core Ratio Limit to 42
    • Set 5-Core Ratio Limit to 40
    • Set 6-Core Ratio Limit to 40
  • Enter the AVX Related Controls submenu
    • Set AVX512 to Enabled
  • Leave the AVX Related Controls submenu
  • Enter the Digi+ VRM submenu
    • Set CPU Load-line Calibration to Level 2
  • Leave the Digi+ VRM submenu
  • Enter the Internal CPU Power Management submenu
    • Set Regulate Frequency by above Threshold to Disabled
  • Leave the Internal CPU Power Management submenu
  • Enter the Tweaker’s Paradise submenu
    • Set Unlock BCLK OC to Enabled
  • Leave the Tweaker’s Paradise submenu
  • Set Max. CPU Cache Ratio to 40
  • Set CPU Core/Cache Voltage to Offset Mode
    • Set Offset Mode Sign to +
    • Set CPU Core Voltage Offset to 0.200

Then save and exit the BIOS.

We re-ran the benchmarks and checked the performance increase compared to the default operation.

  • SuperPI 4M: +20.05%
  • Geekbench 5 (single): +27.96%
  • Geekbench 5 (multi): +28.39%
  • Cinebench R23 Single: +31.50%
  • Cinebench R23 Multi: +33.77%
  • CPU-Z V17.01.64 Single: +27.14%
  • CPU-Z V17.01.64 Multi: +22.76%
  • V-Ray 5: +30.01%
  • AI Benchmark: +43.00%
  • 3DMark Night Raid: +21.53%
  • CS:GO FPS Bench: +6.95%
  • Final Fantasy XV: +5.24%
Core i5-12400 dynamic overclock benchmark performance

Here are the 3DMark CPU Profile scores at stock

  • CPU Profile 1 Thread: +31.72%
  • CPU Profile 2 Threads: +32.94%
  • CPU Profile 4 Threads: +24.88%
  • CPU Profile 8 Threads: +20.87%
  • CPU Profile 16 Threads: +23.53%
  • CPU Profile Max Threads: +25.37%
Core i5-12400 dynamic overclock 3dmark profile performance

Unsurprisingly, thanks to our 900MHz frequency improvement from 4.4GHz to 5.3 GHz, we achieve the best performance in single-threaded benchmark applications with up to 32% improvement in Cinebench R23. Our multi-threaded performance is also significantly better than stock as we increased the all-core frequency by 800MHz from 4 GHz to 4.8GHz. We see the highest performance increase with 43% in AI Benchmark.

When running Prime 95 Small FFTs with AVX enabled, the average CPU P-core clock is 4802 MHz with 1.216 volts. The average CPU temperature is 100 degrees Celsius. The ambient and water temperature is 23.5.7 and 33.8 degrees Celsius. The average CPU package power is 154.5 watts.

Core i5-12400 dynamic overclock prime95 small ffts avx enabled

When running Prime 95 Small FFTs with AVX disabled, the average CPU P-core clock is 4802 MHz with 1.196 volts. The average CPU temperature is 84 degrees Celsius. The ambient and water temperature is 23.5 and 33.0 degrees Celsius. The average CPU package power is 133.5 watts.

Core i5-12400 dynamic overclock prime95 small ffts avx disabled

Intel Core i5-12400: Conclusion

Alright, let us wrap this up.

First, let’s talk about the overclocking results. They are unquestionably fantastic. I could overclock this CPU up to 5.3 GHz for single-core and 4.9 GHz for all-core. That’s about 900 MHz higher than the respective default turbo boost frequencies of 4.4 GHz and 4 GHz. The performance increases across the board ranging from +5% in Final Fantasy to +44% in AI Benchmark.

The main limitation for this CPU is the H0 die which is smaller than the C0 die on the Alder Lake CPUs with E-cores. As the die is smaller, the transistor density increases and causes local hotspots, which causes the CPU to reach TjMax faster. It’s the hotspots that thermally constrain our overclock, not the total power consumption. Therefore, I suspect any 240 or 360 AIO will be able to squeeze the maximum potential out of any Alder Lake CPU lacking E-cores.

Unfortunately, the severe lack of advanced overclocking tools available on the K processors also severely limits our capacity to max out the CPU. We can still use a traditional approach with fixed frequency and voltage; and a modern approach with dynamic turbo ratios and dynamic voltage. However, each comes with its limitation as the conventional approach lacks the high single-threaded performance; and the dynamic approach uses too much voltage in all-core workloads, thus yielding too high temperatures under load.

The bigger story is not the overclocking result. As I stated in the introduction, there’s much more to non-K overclocking than meets the eye, and I felt that the hyped-up media and enthusiasts largely ignored this context and color.

Alder Lake Non-K overclocking exploits an oversight in a pre-release microcode where the BCLK governor does not monitor the base clock frequency in real-time when using an external clock generator. However, it does not change any of the other limitations imposed on locked processors. Most importantly, it does not provide us access to most of the tools inside Intel’s OC Mailbox.

Intel has good reasons to lock the overclocking features behind a paywall.

First, there’s the financial aspect. But frankly, Intel deserves to charge a price and earn money for the value it creates. Since its inception in 2013, the OC Mailbox has provided overclockers and enthusiasts with an incredible amount of value that helped unlock more performance from the processors.

The premium charged for overclocking features on Alder Lake is $18 for the Z chipset and between $66 and $100 for a K processor depending on your i5, i7, or i9 choice. That is similar to the premium charged on Rocket Lake but up significantly from Comet Lake, where the premium was $3 for a Z chipset and about $50 for any K processor. Whether this is a price you’re willing to pay for overclocking features is up to everyone individually.

Second, also obviously, when digital security and “the cyber” is in the news headlines every day, it makes sense that you’d want to prevent specific CPUs from being overclocked. I don’t think it’s reasonable to argue that every CPU should be fully unlocked. I believe it’s reasonable to argue that the range of overclockable SKUs should expand and, perhaps, be priced more affordably.

While researching this Alder Lake non-K overclocking story, I gained a lot more appreciation for both the engineers and Intel, and the motherboard vendors.

On Intel’s side, it’s clear there’s a dedicated effort to create and further enhance value for enthusiasts and overclockers. Judging by what’s come out of Intel since introducing the OC Mailbox feature with Haswell, it’s impossible not to be impressed.

On the motherboard vendor side, it’s incredible to see how they relentlessly pursue ways to find more performance despite Intel’s structured approach to designing overclocking tools. AVX-512 and non-K overclocking are two prime examples of motherboard RD teams identifying a performance opportunity and enabling it for end-users.

Anyway, that’s all for today!

This story was a huge piece to compile and had me ignore some other overclocking platforms for a while. I have a couple of exciting articles in the works, so feel free to subscribe if you’re keen on more overclocking content.

I want to thank my Patreon supporter, Coffeepenbit, for supporting my work.

As per usual, if you have any questions or comments, feel free to drop them in the comment section below. 

See you next time!

Liked it? Take a second to support Pieter on Patreon!

1 Comment

  1. SkatterBencher #38: Intel UHD Graphics 730 Overclocked to 1800 MHz - SkatterBencher

    […] covered in previous videos. I highly recommend you watch my previous videos, SkatterBencher #33 and SkatterBencher #37, if you want to know more about Alder Lake integrated graphics and Alder Lake non-K […]

Leave a Reply