We overclock the Intel Core i9-12900K processor up to 5500 MHz with the GIGABYTE Z690 AORUS Master motherboard and EK-Quantum water cooling.

https://www.youtube.com/watch?v=AndOG5YkXrI

In today’s article we overclock the Intel Core i9-12900K Alder Lake CPU up to 5.5 GHz with EK-Quantum custom loop water cooling and the GIGABYTE Z690 AORUS MASTER motherboard.

Alright, there is lots to go through so let’s get started.

Intel Core i9-12900K: Introduction

The Intel Core i9-12900K is part of 12th generation Intel Core processor line-up.

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.

Well … 125W+.

The overclockable K-SKU processors again come in three flavors: Core i9, Core i7, and Core i5. Each of the three flavors has a -K and -KF variant. The only difference between the two variants is that the -KF comes without integrated graphics.

The Core i9-12900K processor has 8 P-cores and 8 E-cores with a total of 24 threads. The base frequency is 3.2 GHz for the P-cores and 2.4 GHz for the E-cores. The maximum single core boost frequency is 5.1 GHz for the P-cores and 3.9 GHz for the E-cores. The maximum all core boost frequency is 4.9 GHz for the P-cores and 3.7 GHz for the E-cores. The favored P-cores can boost 100 MHz higher to 5.2 GHz. The processor base power is 125W and the maximum turbo power is 241W.

A major change from previous architectures is that Alder Lake is seemingly 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 is sort of in line with how the processor frequency has a base frequency and a maximum turbo frequency.

Another major difference between Alder Lake and really any other Intel Core processor is that, at least for the K-sku CPUs, PL1 is by default equal to PL2. That effectively means that Intel has enabled near unlimited peak turbo by default!

In this video we will cover three different overclocking strategies:

1. First, we increase the performance headroom by unlocking the Turbo Boost 2.0 limits and enabling XMP 3.0
2. Second, we test out 3 OC profiles available in the Z690 AORUS Master BIOS
3. Lastly, we do manual overclocking to get the most performance out of our system

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

Intel Core i9-12900K: Platform Overview

Along with the Intel Core i9-12900K processor and GIGABYTE Z690 AORUS MASTER motherboard, in this guide, we will be using a pair of 16GB AORUS RGB DDR5-6200 memory sticks, an RTX 2080TI graphics card, a 512GB Aorus RGB M.2 NVMe SSD, a Seasonic Prime 850W Platinum power supply, the ElmorLabs Easy Fan Controller, and EK-Quantum water cooling. All this is mounted on top of our favorite Open Benchtable. 

The cost of the components should be around $4,279.

• Intel Core i9-12900K processor: $600
• EK-Quantum Magnitude: $230
• EK-Quantum P360 water cooling kit: $550
• GIGABYTE Z690 Aorus Master motherboard: $469
• NVIDIA 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
• Open Benchtable: $200

I covered the ElmorLabs EFC in a separate article on this blog.

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. This is used for all overclocking strategies.

Intel Core i9-12900K: Benchmark Software

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

• SuperPI 4M https://www.techpowerup.com/download/super-pi/
• Geekbench 5 https://www.geekbench.com/
• Cinebench R23 https://www.maxon.net/en/cinebench/
• CPU-Z https://www.cpuid.com/softwares/cpu-z.html
• V-Ray 5 https://www.chaosgroup.com/vray/benchmark
• AI-Benchmark https://ai-benchmark.com/
• 3DMark CPU Profile https://www.3dmark.com/
• 3DMark Night Raid https://www.3dmark.com/
• CS:GO FPS Bench https://steamcommunity.com/sharedfiles/filedetails/?id=500334237
• Final Fantasy XV http://benchmark.finalfantasyxv.com/na/
• Prime 95 https://www.mersenne.org/download/

Prime95 doesn’t quite work out of the box with Alder Lake due the complicated configuration with P-cores having hyperthreading and the E-cores not. You can still set up Prime95 to run on all cores by manually setting the local.txt configuration.

For the Core i9-12900K, set:

• NumCPUs=24 (that’s the total amount of logical processors)
• NumHyperthreads=1 (that means every logical processor will run 1 thread)

Intel Core i9-12900K: Stock Performance

The first thing we must do before we start any overclocking is check the system performance at default settings.

Please note that out of the box, the GIGABYTE Z690 AORUS MASTER fully unleashes the Turbo Boost 2.0 power limits. So, to check the performance at default settings you must

• Enter the Advanced Mode and stay in the Tweaker menu
• Enter the Advanced CPU Settings submenu
  • Set Turbo Power Limits to Enabled
  • Set Package Power Limit1 – TDP (Watts) to 241
  • Set Package Power Limit2 (Watts) to 241
  • Set Package Power Limit2 Time to 56

Then save and exit the BIOS.

Here is the benchmark performance at stock:

• SuperPI 4M: 34.269 seconds
• Geekbench 5 (single): 2,010 points
• Geekbench 5 (multi): 18,110 points
• Cinebench R23 Single: 2,018 points
• Cinebench R23 Multi: 26,844 points
• CPU-Z V17.01.64 Single: 818.5 points
• CPU-Z V17.01.64 Multi: 11,470.6 points
• V-Ray 5: 18,495 vsamples
• AI Benchmark: 4,324 points
• 3DMark Night Raid: 78,846 points
• CS:GO FPS Bench: 597.26 fps
• Final Fantasy XV: 189.14 fps

Here are the 3DMark CPU Profile scores at stock

• CPU Profile 1 Thread: 1,105
• CPU Profile 2 Threads: 2,127
• CPU Profile 4 Threads: 4,187
• CPU Profile 8 Threads: 7,910
• CPU Profile 16 Threads: 10,217
• CPU Profile Max Threads: 11.619

When running Prime 95 Small FFTs with AVX enabled, the average CPU P-core clock is 4700 MHz and average CPU E-core clock is 3700 MHz with 1.128 volts. The average CPU temperature is 84 degrees Celsius and the average CPU package power is 241 watts.

When running Prime 95 Small FFTs with AVX disabled, the average CPU P-core clock is 4892 MHz and average CPU E-core clock is 3700 MHz with 1.186 volts. The average CPU temperature is 83 degrees Celsius and the average CPU package power is 240 watts.

Now, let us try our first overclocking strategy.

However, before we get going, make sure to locate the CMOS Clear button

Pressing the CMOS Clear button will reset all your BIOS settings to default. This is useful in case your system does not boot up after overclocking or 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: Unleashed Turbo + XMP 3.0

In our first overclocking strategy we simply take advantage of unleashing the Turbo Boost 2.0 power limits and Intel XMP 3.0.

Turbo Boost 2.0

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 a proprietary EWMA formula. This stands for Exponentially Weighted Moving Average. There are 3 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. Very importantly, PL1 should not be set higher than the thermal solution cooling limits.
• Power Limit 2, or PL2, is the maximum power the processor is allowed to use for limited amount of time.
• Tau is a weighing constant used in the algorithm 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.

Obviously, Turbo Boost 2.0 technology is available on Alder Lake as it’s the main driver of performance over the base frequency.

A major difference between Alder Lake and really any other Intel Core processor is that, at least for the K-SKU CPUs, PL1 is by default equal to PL2. This is very different from before where PL1 would equal the TDP and PL2 would range from 200 to 250W. This change effectively means that Intel has enabled near unlimited peak turbo by default!

For the 12900K the maximum power limit is set at 241W.

The maximum performance is therefore entirely limited by the capabilities of your cooling solution. If your cooling solution is insufficient, then the processor will reduce the operating frequency at the maximum allowed temperature or TjMax. For Alder Lake CPUs that’s at 100 degrees Celsius.

Note that the GIGABYTE Z690 Aorus Master has the Turbo Boost 2.0 fully unleashed by default, so you don’t have to make any BIOS changes to benefit from the additional performance.

Intel Extreme Memory Profile 3.0

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

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

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

There’s a lot more to the new XMP 3.0 standard which is outside the scope of this overclocking guide. By the time this video goes out I should already have a more detailed overview up on my channel.

GIGABYTE Z690 XMP 3.0 BIOS Tools & Features

GIGABYTE has put in quite some effort to support all the new features that come with the Intel XMP 3.0 specification. Let’s have a quick overview.

First, in the Easy Mode you can select the different XMP profiles with the click of a button. So, no need to access the advanced options to benefit from XMP.

In the Advanced mode there’s more options.

• DDR5 Auto Booster will increase the memory frequency from DDR5-4800 to DDR5-5000
• DDR5 XMP Booster lists pre-configured memory profiles for a variety of DDR5 ICs. In my BIOS version I only get Micron and Hynix, but I expect more profiles for other ICs in future BIOS releases
• Extreme Memory Profile gives you access to XMP profiles on your memory module. As mentioned before, XMP 3.0 offers 3 vendor profiles and 2 user profiles. In this case, only the 3 vendor profiles show because I have not made any custom profile myself.

In the Advanced Memory Settings submenu, you will find even more options.

• SPD Info provides you with the SPD information of each of your memory modules
• SPD Setup is an experimental feature that allows you to create custom XMP profiles.

Entering the SPD Setup submenu opens up a world of memory tuning options.

By far my favorite feature is the Performance Index. This index provides you with a quick overview of which XMP profile will give you the most performance. As you can see, while XMP 2 gives us the highest frequency at DDR5-6400, XMP 1 and 3 should give us better performance at DDR5-6200. I don’t know the way the performance index is calculated or gauge its accuracy, but it’s a good starting point nonetheless.

Then we also see the 2 XMP 3.0 user profiles. These fields are entirely programmable by the user. In the selection menu below the profile columns, we can find a couple easy options to set the user profile:

• Clear will clear the custom profile
• Current will load the DRAM settings currently set in BIOS
• JEDEC will load the JEDEC DDR4-4800 settings
• XMP1, XMP2, XMP3 will load the settings from the respective XMP profile
• Micron 1, 2, 3, 4 and Hynix 1, 2, 3, 4 will load the pre-configured memory profiles from the DDR5 XMP boost menu we saw earlier.

Of course, we can also manually change the values. Clicking Set will flash the SPD with our profile. We will now be able to select it upon reboot. Well, let me get back to that.

By selecting the User profiles, they will also show up in the Performance Index chart. This will help us evaluate how well our new profiles are scoring against the vendor provided XMP profiles.

As I mentioned, we will be able to select the custom user profiles upon reboot. However, we must make sure to set Memory Boot Mode to Disable Fast Boot.

Let’s reboot.

As you can see, we can find 5 profiles in the Extreme Memory Profile option. Selecting Profile 5 provides us with the DDR5-6600 profile we customized before.

Anyway, let’s go back to our overclocking strategy.

Upon entering the BIOS

• In Easy Mode, change X.P.M. Disabled to X.M.P. Profile1

Then save and exit the BIOS.

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

• SuperPI 4M: +0.10%
• Geekbench 5 (single): +0.25%
• Geekbench 5 (multi): +7.00%
• Cinebench R23 Single: +0.00%
• Cinebench R23 Multi: +2.78%
• CPU-Z V17.01.64 Single: +0.28%
• CPU-Z V17.01.64 Multi: +0.13%
• V-Ray 5: +1.61%
• AI Benchmark: +7.54%
• 3DMark Night Raid: +0.71%
• CS:GO FPS Bench: +0.39%
• Final Fantasy XV: +0.66%

Here are the 3DMark CPU Profile scores at stock

• CPU Profile 1 Thread: +0.54%
• CPU Profile 2 Threads: +2.02%
• CPU Profile 4 Threads: +0.38%
• CPU Profile 8 Threads: +0.47%
• CPU Profile 16 Threads: +2.04%
• CPU Profile Max Threads: +2.94%

As expected , we see the largest performance difference in multi-threaded applications which were previously constraint by the maximum power of 241W. We see up to 7.54% performance increase in AI Benchmark.

When running Prime 95 Small FFTs with AVX enabled, the average CPU P-core clock is 4881 MHz and average CPU E-core clock is 3700 MHz with 1.203 volts. The average CPU temperature is 99 degrees Celsius and the average CPU package power is 302.5 watts.

When running Prime 95 Small FFTs with AVX disabled, the average CPU P-core clock is 4900 MHz and average CPU E-core clock is 3700 MHz with 1.189 volts. The average CPU temperature is 84 degrees Celsius and the average CPU package power is 242.7 watts.

OC Strategy #2: OC Profiles + XMP 3.0

In our second overclocking strategy we try the three available performance tuning profiles in the Z690 AORUS Master motherboard BIOS.

Two of the three profiles can be found under the CPU Upgrade option: Gaming Profile and Max Performance Profile. The third profile is available as its own option and is called Enhanced Multi-Core Performance. Each profile has a different approach to performance tuning.

• Enhanced Multi-Core Performance sets all cores to the maximum available turbo ratio. On the 12900K that means every core will run at 5.2 GHz. In addition, all E-cores run at their maximum Turbo Boost of 3.9 GHz.
• Gaming Profile retains the detail Turbo Boost behavior but disables the E-cores, so the CPU goes from 16 cores with 24 threads to 8 cores with 16 threads
• Max Performance Profile sets the P-cores to 5.0 GHz and the E-cores to 3.9 GHz

It is important to note that only the Enhanced MultiCore Performance profile is in fact overclocking your CPU. That’s because it runs 6 out of the 8 P-cores higher than specification. The two other profiles are not actually overclocking as they run all cores within voltage and frequency specification.

For each of the three profiles we will also be using a different XMP 3.0 profile. The profiles are:

• XMP 1: DDR5-6200 40-40-40-80 at 1.35V
• XMP 2: DDR5-6400 42-42-42-84 at 1.45V
• XMP 3: DDR5-6200 38-38-38-76 at 1.50V

OC Strategy #2A: EMCP + XMP 1

For Strategy 2A we use Enhanced Multi-Core Performance and the first XMP profile.

Upon entering the BIOS

• In Easy Mode, change X.P.M. Disabled to X.M.P. Profile1
• Enter Advanced Mode and stay in the Tweaker menu
• Set Enhanced Multi-Core Performance 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: +0.19%
• Geekbench 5 (single): +3.13%
• Geekbench 5 (multi): +11.58%
• Cinebench R23 Single: +1.34%
• Cinebench R23 Multi: +6.84%
• CPU-Z V17.01.64 Single: +2.74%
• CPU-Z V17.01.64 Multi: +5.63%
• V-Ray 5: +5.55%
• AI Benchmark: +16.65%
• 3DMark Night Raid: +3.92%
• CS:GO FPS Bench: +1.52%
• Final Fantasy XV: +1.79%

Here are the 3DMark CPU Profile scores at stock

• CPU Profile 1 Thread: +1.00%
• CPU Profile 2 Threads: +4.80%
• CPU Profile 4 Threads: +5.47%
• CPU Profile 8 Threads: +5.68%
• CPU Profile 16 Threads: +8.50%
• CPU Profile Max Threads: +7.84%

We see performance gains across the board but most noticeably in the multithreaded benchmark applications. That’s of course because the default all-core turbo frequency is 4.9 GHz and this profile increases this to 5.2 GHz.

When running Prime 95 Small FFTs with AVX enabled, the average CPU P-core clock is 4856 MHz and average CPU E-core clock is 3900 MHz with 1.201 volts. The average CPU temperature is 99 degrees Celsius and the average CPU package power is 303.7 watts.

When running Prime 95 Small FFTs with AVX disabled, the average CPU P-core clock is 5099 MHz and average CPU E-core clock is 3900 MHz with 1.278 volts. The average CPU temperature is 100 degrees Celsius and the average CPU package power is 304.3 watts.

OC Strategy #2B: Gaming + XMP 2

For Strategy 2B we use the Gaming Profile and the second XMP profile.

Upon entering the BIOS

• In Easy Mode, change X.P.M. Disabled to X.M.P. Profile2
• Enter Advanced Mode and stay in the Tweaker menu
• Set CPU Upgrade to Gaming Profile

Then save and exit the BIOS.

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

• SuperPI 4M: +0.01%
• Geekbench 5 (single): +1.99%
• Geekbench 5 (multi): -17.26%
• Cinebench R23 Single: +0.15%
• Cinebench R23 Multi: -24.27%
• CPU-Z V17.01.64 Single: -11.39%
• CPU-Z V17.01.64 Multi: -28.83%
• V-Ray 5: -20.26%
• AI Benchmark: +4.02%
• 3DMark Night Raid: -18.43%
• CS:GO FPS Bench: +5.89%
• Final Fantasy XV: +1.85%

Here are the 3DMark CPU Profile scores at stock

• CPU Profile 1 Thread: -1.09%
• CPU Profile 2 Threads: -7.33%
• CPU Profile 4 Threads: -22.55%
• CPU Profile 8 Threads: -30.19%
• CPU Profile 16 Threads: -6.34%
• CPU Profile Max Threads: -17.95%

The performance is a mixed bag with the gaming profile. As expected, we see the largest performance losses in multi-threaded benchmark applications. But not all our results are that straight-forward.

• In CPU-Z single we lose more than 10% performance despite running the same P-core frequency as default settings
• AI Benchmark performance in fact increased by about 4% despite it being a multi-threaded benchmark application. It is fairly memory sensitive so I think the XMP profile helped here versus stock
• CS:GO performance increased by almost 6% or 30 frames per second

I didn’t look too deeply into why the benchmark results are like this. I’m sure there will be plenty of media out there that will have done a thorough analysis of Alder Lake performance with and without E-cores enabled.

When running Prime 95 Small FFTs with AVX enabled, the average CPU P-core clock is 4900 MHz with 1.128 volts. The average CPU temperature is 83 degrees Celsius and the average CPU package power is 210.9 watts.

When running Prime 95 Small FFTs with AVX disabled, the average CPU P-core clock is 4900 MHz with 1.1.153 volts. The average CPU temperature is 75 degrees Celsius and the average CPU package power is 176.2 watts.

OC Strategy #2C: Max Performance + XMP 3

For Strategy 2C we use the Max Performance Profile and the third XMP profile.

Upon entering the BIOS

• In Easy Mode, change X.P.M. Disabled to X.M.P. Profile3
• Enter Advanced Mode and stay in the Tweaker menu
• Set CPU Upgrade to Max Performance Profile

Then save and exit the BIOS.

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

• SuperPI 4M: -2.38%
• Geekbench 5 (single): -0.75%
• Geekbench 5 (multi): +10.84%
• Cinebench R23 Single: -3.47%
• Cinebench R23 Multi: +5.92%
• CPU-Z V17.01.64 Single: -0.33%
• CPU-Z V17.01.64 Multi: +2.86%
• V-Ray 5: +4.73%
• AI Benchmark: +9.94%
• 3DMark Night Raid: +1.96%
• CS:GO FPS Bench: -0.35%
• Final Fantasy XV: +0.07%

Here are the 3DMark CPU Profile scores at stock

• CPU Profile 1 Thread: -2.08%
• CPU Profile 2 Threads: -0.80%
• CPU Profile 4 Threads: +2.15%
• CPU Profile 8 Threads: +1.88%
• CPU Profile 16 Threads: +5.27%
• CPU Profile Max Threads: +5.28%

The performance result is again a mixed bag. We lose slightly in single threaded benchmark applications as all cores are running at 5.0 GHz. At default settings, one core can boost to 5.2 GHz and another core will run at 5.1 GHz. In multi-threaded benchmark applications we see a performance improvement over stock thanks to the 100 MHz extra frequency on the P-cores as we increased the all-core frequency from 4.9 GHz to 5.0 GHz.

When running Prime 95 Small FFTs with AVX enabled, the average CPU P-core clock is 4884 MHz and average CPU E-core clock is 3900 MHz with 1.204 volts. The average CPU temperature is 99 degrees Celsius and the average CPU package power is 305.5 watts.

When running Prime 95 Small FFTs with AVX disabled, the average CPU P-core clock is 5000 MHz and average CPU E-core clock is 3900 MHz with 1.225 volts. The average CPU temperature is 89 degrees Celsius and the average CPU package power is 266.6 watts.

OC Strategy #3: Manual Overclock

In our third and final overclocking strategy we will pursue a manual overclock.

We will be using a range of overclocking features on the Alder Lake CPUs including By Core Usage CPU Ratios, AVX2 Negative Ratio Offset, and Advanced Voltage Offset.

Active Turbo Ratio

Generally speaking, on Intel platforms there are two ways to manually configure the CPU ratio: Sync all cores and Active Turbo Ratio. Sync All Cores sets 1 ratio that is applied to all cores. This is very much the traditional way of overclocking.

Of course, back in the day when we only had 1 or 2 cores, the quality difference between the cores was relatively small. So, there was not that much benefit to max out each core independently.

Nowadays, even mainstream CPUs have up to 16 cores. That means the quality difference between the cores can vary greatly. In addition, the heat produced by 16 cores at full load will be much higher than the heat produced by just a couple of cores at light load. So, if you tune your system for the worst-case scenario, you will miss out on a lot of performance in the most common scenarios.

Active Turbo Ratio configuration allows us to configure the overclock for different scenarios ranging from 1 active core to all active cores. This enables us to run some cores significantly faster than others when the conditions are right. On Alder Lake, we can configure the by core usage ratio for P-cores and E-cores separately. So, we can set the maximum ratio for 1-active P-core to 8-active P-cores and for 1-active E-core to 8-active E-cores.

Note that Active Turbo Ratio configuration is not the same as configuring each core specifically. When using Active Turbo Ratio, we determine an overclock according to the quantity of active cores. For example, if a workload is using 4 cores, then the CPU will determine by itself which cores should execute this workload and will apply our configured frequency to those cores.

AVX2 Negative Ratio Offset

Intel first introduced the AVX negative ratio offset on Broadwell-E processors. Successive generations adopted this feature and eventually expanded it with AVX2 and AVX-512 negative offsets.

AVX negative ratio offsets are very useful to achieve the maximum performance for both SSE and AVX workloads. Generally speaking, an offset of 2 or 3 is recommended but it is highly dependent on not only your cooling solution but also the motherboard you’re using. That is because AVX workloads are very demanding and therefore require great cooling and power delivery.

While the function has carried over to Alder Lake in principle, there are some big differences in the implementation of AVX ratio offset between Alder Lake and Rocket Lake.

• First, on Alder Lake, the AVX2 negative ratio offset is only applied to the P-cores. The E-core frequency is unaffected.

• Second, by default the maximum ratio during an AVX workload is the Turbo Boost 2.0 ratio. If you want an offset of 0, so the AVX workload won’t trigger a frequency reduction, you’ll need to manually set 0. Note that on some BIOSes programming 0 means programming “default”, in which case you’ll be limited to maximum 51X for the AVX ratio on the 12900K.

• Just like on Rocket Lake, the AVX negative offset is referenced against each individual core. Prior to Rocket Lake, the AVX offset would be referenced against the all-core maximum ratio. As the AVX offset will be applied to each core separately, you’ll have to be a little careful when your overclock has different ratios for different cores.

• Third, Intel has made some changes to how it flags an AVX workload. The effect is that some light AVX workloads will no longer trigger the AVX negative offset. We can demonstrate this new behavior using Y-cruncher.

We ran the Y-cruncher Component Stress Tester to run a variety of AVX workloads on the 12900K. We configure the 12900K as follows: 5GHz for the P-cores, 3.9GHz for the E-cores, and an AVX offset of 5 ratios. Then we use HWiNFO to monitor the effective clock during the benchmark

During the Y-cruncher workloads, we see that the during BKT the P-core effective clock frequency is 5GHz. Since this is a non-AVX workload, it’s as expected. Then there are two AVX workloads, BPP and SFT, where the P-core effective clock drops to 4.5 GHz. This is expected as the 5 GHz P-core frequency is reduced by 500 MHz due to the AVX offset.

During the FFT workload, also AVX, the frequency is back to 5 GHz. So, the AVX offset was not triggered. Then the next 4 AVX workloads, N64, HNT, VST, and C17, bounce back and forth the AVX offset and non-offset.

Advanced Voltage Offset

Advanced Voltage Offset was first introduced with the Comet Lake architecture in 2020 and is of course also included with Alder Lake. It is more commonly known as the V/F Offset Mode or V/F Points. It is an extension of the Adaptive Voltage Mode which we discussed many times before.

Advanced Voltage Offset exposes some of the points on the V/f curve to the end-user and allows for manual adjustment of these points. The amount of V/f points is not architectural and can change between SKUs. In other words, specific CPU models can have more or less pre-defined V/F points.

The only requirement for the V/f curve is monotonicity. Following a monotonic function, as a rule the voltage for a given CPU ratio must be equal to or higher than the next lower ratio. So, the voltage for 48X must be equal to or higher than 47X.

The main purpose of the Advanced Voltage Offset is to provide end users with the ability to under-volt their CPUs at specific parts of the V/f curve. In addition to under-volting, this feature also allows over-volting. This is particularly useful when manual overclocking and when you are trying to increase the maximum frequency.

The Advanced Voltage Offset function is commonly used in two ways.

1. First, you configure a positive voltage offset for the highest V/f point. This helps achieve higher single core overclocks.
2. Second, you configure a negative voltage offset for the second highest V/f point. This helps achieve lower voltage for an all-core overclock, which results in lower temperature in all-core workloads, and thus potentially additional overclocking headroom.

On Alder Lake, this approach is a bit trickier because of available V/f points. On the 12900K, the highest V/f point is 53X and the second highest V/f point is 48X. That means if we overclock our highest CPU core to 55X, then the voltage for ratios 49X to 54X will be interpolated between these two V/f points. This lack of granularity will be headache for anyone trying to balance a high single core and high all-core overclock.

Next weird thing with Alder Lake is that the amount of available V/f points has increased from 8 to 11. However, points 7 to 10 are all copies of V/f point 11. This is by design – it’s not a bug, but we don’t know what it’s for. Possibly for future products

Upon entering the BIOS

• Enter the Advanced Mode and stay in the Tweaker menu
• Set Enhanced Multi-Core Performance to Disabled
• Enter the Advanced CPU Settings submenu
  • Set Active Turbo Ratios to Manual
    • Set Turbo Ratio (1 P-Core Active) to 55
    • Set Turbo Ratio (2 P-Core Active) to 55
    • Set Turbo Ratio (3 P-Core Active) to 54
    • Set Turbo Ratio (4 P-Core Active) to 53
    • Set Turbo Ratio (5 P-Core Active) to 52
    • Set Turbo Ratio (6 P-Core Active) to 52
    • Set Turbo Ratio (7 P-Core Active) to 52
    • Set Turbo Ratio (8 P-Core Active) to 52
    • Set Turbo Ratio (1 E-Core Active) to 40
    • Set Turbo Ratio (2 E-Core Active) to 40
    • Set Turbo Ratio (3 E-Core Active) to 40
    • Set Turbo Ratio (4 E-Core Active) to 40
    • Set Turbo Ratio (5 E-Core Active) to 40
    • Set Turbo Ratio (6 E-Core Active) to 40
    • Set Turbo Ratio (7 E-Core Active) to 40
    • Set Turbo Ratio (8 E-Core Active) to 40
• Leave the Advanced CPU Settings submenu
• Set Extreme Memory Profile(X.M.P) to Profile3
• Set Vcore Voltage Mode to Adaptive Vcore
  • Set VF Offset Mode to Selection
  • Set VF Point 6 Offset to -0.025V
  • Set VF Point 11 Offset to +0.075V

Then save and exit the BIOS.

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

• SuperPI 4M: +6.17%
• Geekbench 5 (single): +7.41%
• Geekbench 5 (multi): +14.56%
• Cinebench R23 Single: +6.49%
• Cinebench R23 Multi: +8.77%
• CPU-Z V17.01.64 Single: +8.00%
• CPU-Z V17.01.64 Multi: +7.70%
• V-Ray 5: +8.60%
• AI Benchmark: +18.29%
• 3DMark Night Raid: +4.13%
• CS:GO FPS Bench: +1.16%
• Final Fantasy XV: +1.98%

Here are the 3DMark CPU Profile scores at stock

• CPU Profile 1 Thread: +4.43%
• CPU Profile 2 Threads: +7.57%
• CPU Profile 4 Threads: +6.35%
• CPU Profile 8 Threads: +5.44%
• CPU Profile 16 Threads: +10.89%
• CPU Profile Max Threads: +9.54%

We achieve the best performance in all our benchmark applications. The highest performance gain is in AI Benchmark where we increase by 18%

When running Prime 95 Small FFTs with AVX enabled, the average CPU P-core clock is 4918 MHz and average CPU E-core clock is 3996 MHz with 1.197 volts. The average CPU temperature is 99 degrees Celsius and the average CPU package power is 303.2 watts.

When running Prime 95 Small FFTs with AVX disabled, the average CPU P-core clock is 5151 and average CPU E-core clock is 4000 MHz with 1.281 volts. The average CPU temperature is 99 degrees Celsius and the average CPU package power is 311.8 watts.

Intel Core i9-12900K: Overclocking Debrief

Before I jump to the conclusion, let’s do a quick debrief on overclocking the Core i9-12900K.

First of all, there’s genuinely a lot of overclocking headroom in the Alder Lake CPUs. While I didn’t quite get it running stable for my manual overclock, this CPU was able to run at 5.6 GHz on multiple P-cores. And, I’ve seen other 12900K CPUs that can even do 5.7 GHz on custom loop water cooling. That’s really exciting

The main challenge you’ll face when overclocking is the big range between the 1-core maximum frequency and the all-core maximum frequency. To set a dynamic overclock you need to use Adaptive Voltage mode and the Advanced Voltage Offset. This allows you to set the target voltage for the 1-core maximum frequency as well as keep a dynamic voltage scaling with the lower frequencies.

For the 12900K there are only 2 V/F Points relevant for manual overclocking: the highest V/F Point which maps against our highest ratio and the second highest V/F point which maps against the 48X ratio. The voltage for all the CPU ratios in between is interpolated between these two points. And that makes it very complicated.

In our manual overclock, we have a 1-core maximum overclock of 5.5GHz and an all-core overclock of 5.2 GHz. We can control the target voltage for 55X using the highest V/F point, however we have almost no control over the voltage of 52X. The lack of control is annoying as it’s really important to finetune the voltage in an all-core workload to manage the temperatures.

Speaking of the temperatures, as you can see from the Prime95 results even at default settings there’s not much thermal headroom to push the all-core frequency. When unlocking the Turbo Boost 2.0 limits, we already hit 100 degrees Celsius in an all-core AVX2 workload as the CPU package power hits 300W.

The TjMax, or maximum temperature, is the safeguard against any long-term damage as it throttles the CPU frequency to stay within the guaranteed operating temperature.

Due to the new way Intel handles AVX ratio offset, it’s still worth pursuing higher all-core frequency as some light AVX workloads will not trigger the offset.

In terms of voltage, it’s important to note that they’re much lower than what we’re used to on previous Intel platforms. Whereas on Rocket Lake we’d see over 1.5V even at default, on Alder Lake the default voltage ranges up to 1.3 to 1.35V.

In short, here’s how I’d approach P-core overclocking:

• Single P-core max voltage around or slightly higher than 1.4V. Expect 5.5GHz to 5.7GHz max frequency
• All P-core max voltage around 1.2V if your cooling solution can handle it. Expect between 5.0 and 5.2 GHz max frequency

I’d suggest to leave E-cores at default as performance impact is minimal and overclocking capability is limited to 4.1 to 4.2 GHz.

Intel Core i9-12900K: Conclusion

Alright, let us wrap this up.

I had already tested one Alder Lake 12900K CPU prior to putting this system together. So, I was primarily interested in confirming my initial overclocking results. In addition, I was also keen to find out how GIGABYTE approached this totally new architecture and memory.

Overall, I am very pleased with how my manual overclock turned out. I’m sure there’s a little more performance in it if I could get 5.6 GHz stable, but 5.5GHz on every P-core is very nice as well. I also think with Alder Lake the traditional way of overclocking – that is setting all cores to a single ratio – is officially dead. To maximize the performance by overclocking, you really need to set a dynamic overclock where many cores boost high, few cores boost even higher, and one core boosts the highest.

By default, the Z690 Aorus Master is aggressively tuned as the Turbo Boost 2.0 limits are fully unleashed. That does mean you’ll need good cooling to keep the 12900K within check in a heavy AVX all-core workload.

I was most impressed by the work GIGABYTE has put into the DDR5 aspect of Alder Lake. There are plenty of new features including the SPD Setup which open up more tuning possibilities. It is still early days for DDR5 overclocking, so it is difficult to gauge to what extend the motherboards are meeting or exceeding DDR5 overclocking expectations.

Anyway, that’s all for today!

In the coming weeks I’ll be putting up more Alder Lake overclocking content up. As per usual if you have any questions or comments, feel free to drop them in the comment section below. 

See you next time!