We overclock the Intel Core i9-13900KS up to 6300 MHz with the ASUS ROG Maximus Z790 Hero motherboard and EK-Quantum water cooling.

https://www.youtube.com/watch?v=dpWw_5KfpIw

In my other Raptor Lake overclocking guides, I focused a lot on detailed voltage tuning, but I want to keep it more uncomplicated in this guide. So, I will avoid too in-depth tuning while still trying to explain the dynamics behind voltages and frequencies.

Intel Core i9-13900KS: Introduction

The Intel Core i9-13900KS is the flagship processor of Intel’s 13^th-generation Core lineup. It is touted as the very first 6 GHz desktop CPU.

Intel Raptor Lake builds on top of the performance hybrid architecture introduced with 12^th gen Alder Lake. So, it also features Performance P-Cores and Efficient E-cores. Like Alder Lake, it is built on the Intel 7 process technology, formerly known as 10nm Enhanced SuperFin (ESF). While it may sound like Raptor Like is not much different from its predecessor, the spec sheet reads quite impressive.

Compared to its Core i9-12900KS predecessor, launched one year ago, the 13900KS has a 500 MHz higher turbo boost frequency and eight additional threads.

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

• First, we unleash the Turbo Boost 2.0 limits and enable XMP 3.0
• Second, we overclock using ASUS AI Overclocking technology
• Third, we get into the simple manual overclocking
• Lastly, finetune our manual overclock using Intel’s OCTVB technology

However, before we jump into overclocking, let us quickly review the hardware and benchmarks used in this video.

Intel Core i9-13900KS: Platform Overview

The system we’re overclocking today consists of the following hardware.

Item	SKU	Price (USD)
CPU	Intel Core i9-13900KS	699
Motherboard	ASUS ROG Maximus Z790 Hero	760
CPU Cooling	EK-Quantum Velocity2 EK-Quantum Surface S360 EK-Pro QDC Kit P360	137 98 777
Fan Controller	ElmorLabs EFC-X9	60
Memory	G.SKILL Trident Z DDR5-7200 CL34 32GB	310
Power Supply	Antec HCP 1000W Platinum	110
Graphics Card	ASUS ROG Strix RTX 2080 TI	880
Storage	AORUS RGB 512 GB M.2-2280 NVME	120
Chassis	Open Benchtable V2	200

ElmorLabs EFC-X9

The ElmorLabs EFC-X9 is an evolution of the original EFC, which I’ve used in many previous SkatterBencher guides. The X9 supports up to nine independent PWM fans and can supply 4.5A per header or a maximum of 20A total. Additionally, there are two external temperature sensor connectors, plus an onboard ambient temperature and humidity sensor. There’s also dedicated EFC software to control the device.

As usual, I map the fan curve to the water temperature and have the fans scale from 0 to 100% fan duty cycle from 25C to 40C water temperature.

The main takeaway from this configuration is that it gives us a good indicator of whether the cooling solution is saturated. Suppose the CPU is at TjMax, and the water temperature exceeds 40 degrees Celsius. In that case, it means the fans are at maximum speed, and thus the cooling solution is saturated. Improving the cooling solution by adding radiators or changing to more powerful fans would be the right action.

Suppose the CPU is at TjMax and the water temperature is below 40 degrees Celsius. In that case, it means the cooling solution is not saturated. Therefore, to improve the CPU temperature, you may enhance the thermal transfer of the CPU heat into the loop by changing the thermal paste, delidding, or changing the water block.

Intel Core i9-13900KS: 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
• Shadow of the Tomb Raider https://store.steampowered.com/app/750920/Shadow_of_the_Tomb_Raider_Definitive_Edition/
• Final Fantasy XV http://benchmark.finalfantasyxv.com/na/
• Prime 95 https://www.mersenne.org/download/

Intel Core i9-13900KS: Stock Performance

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

Please note that out of the box, the Maximus Z790 Hero fully unleashes the Turbo Boost 2.0 power limits. So, to check the performance at default settings, you must enter the BIOS and

• Go to the Extreme Tweaker menu
• Set ASUS MultiCore Enhancement to Disabled – Enforce All Limits

Then save and exit the BIOS.

The default Turbo Boost 2.0 parameters for the Core i9-13900KS are as follows:

• PL1: 253W
• PL2: 253W
• Tau: 33sec
• ICCMax: 488A

Here is the benchmark performance at stock:

• SuperPI 4M: 25.948 seconds
• Geekbench 5 (single): 2,244 points
• Geekbench 5 (multi): 23,379 points
• Cinebench R23 Single: 2,340 points
• Cinebench R23 Multi: 39,711 points
• CPU-Z V17.01.64 Single: 967.1 points
• CPU-Z V17.01.64 Multi: 17,185.4 points
• V-Ray 5: 27,036 vsamples
• AI Benchmark: 5,448 points
• 3DMark Night Raid: 90,285 points
• CS:GO FPS Bench: 624.88 fps
• Tom Raider: 207 fps
• Final Fantasy XV: 207.69 fps

Here are the 3DMark CPU Profile scores at stock

• CPU Profile 1 Thread: 1,284
• CPU Profile 2 Threads: 2,492
• CPU Profile 4 Threads: 4,802
• CPU Profile 8 Threads: 9,101
• CPU Profile 16 Threads: 11,726
• CPU Profile Max Threads: 16,927

When running Prime 95 Small FFTs with AVX2 enabled, the average CPU P-core clock is 5107 MHz, and the average CPU E-core clock is 3926 MHz with 1.079 volts. The average CPU temperature is 84 degrees Celsius. The ambient and water temperature is 25.7 and 29.9 degrees Celsius. The average CPU package power is 252.9 watts.

When running Prime 95 Small FFTs with AVX disabled, the average CPU P-core clock is 5357 MHz, and the average CPU E-core clock is 4250 MHz with 1.191 volts. The average CPU temperature is 84 degrees Celsius. The ambient and water temperature is 25.3 and 28.7 degrees Celsius. The average CPU package power is 252.9 watts.

Now, let us try our first overclocking strategy.

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

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 IO panel.

OC Strategy #1: Unleashed Turbo Boost 2.0 + XMP 3.0

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

Intel Turbo Boost 2.0

Intel Turbo Boost 2.0 Technology allows the processor cores to run faster than the base operating frequency. Turbo Boost is available when the processor works below its 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. 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 the average power will not exceed. Historically, this has always been set equal to Intel’s advertised TDP. 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, in seconds, is the time window for calculating the average power consumption. The CPU will reduce the CPU frequency if the average power consumed is higher than PL1.

Turbo Boost 2.0 technology is available on Raptor Lake as it’s the primary driver of performance over the base frequency.

Similar to Alder Lake, but a significant change from any previous Intel Core processors is that, at least for the K-SKU CPUs, PL1 is by default equal to PL2. This change effectively means that Intel has enabled near-unlimited peak turbo by default!

For the 13900KS, the maximum power limit is set at 253W.

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

An easy ASUS MultiCore Enhancement option on ASUS motherboards allows you to unleash the Turbo Boost power limits. Set the option to Enabled – Remove All Limits and enjoy maximum performance.

Adjusting the power limits is strictly not considered overclocking, as we don’t change any of the CPU’s thermal, electrical, or frequency parameters. Intel provides the Turbo Boost parameters as guidance to motherboard vendors and system integrators to ensure their designs enable the base performance of the CPU. Better motherboard designs, thermal solutions, and system configurations can facilitate peak performance for longer.

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 and 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 primarily based on the XMP 2.0 standard for DDR4 but has additional functionality.

There’s a lot more to the new XMP 3.0 standard, which is outside the scope of this overclocking guide. If you’re interested in more details about XMP 3.0, check out my Alder Lake launch blog.

BIOS Settings & Benchmark Results

Upon entering the BIOS

• Go to the Extreme Tweaker menu
• Set Ai Overclock Tuner to XMP II
• Set ASUS MultiCore Enhancement to Enabled – Remove All Limits

Then save and exit the BIOS.

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

• SuperPI 4M: +0.79%
• Geekbench 5 (single): +3.34%
• Geekbench 5 (multi): +11.88%
• Cinebench R23 Single: +0.34%
• Cinebench R23 Multi: +2.91%
• CPU-Z V17.01.64 Single: +0.06%
• CPU-Z V17.01.64 Multi: +0.07%
• V-Ray 5: +3.73%
• AI Benchmark: +11.47%
• 3DMark Night Raid: +1.08%
• CS:GO FPS Bench: +1.02%
• Tomb Raider: +0.00%
• Final Fantasy XV: +0.01%

Here are the 3DMark CPU Profile scores

• CPU Profile 1 Thread: +0.08%
• CPU Profile 2 Threads: +0.68%
• CPU Profile 4 Threads: +0.15%
• CPU Profile 8 Threads: +0.01%
• CPU Profile 16 Threads: +0.51%
• CPU Profile Max Threads: +0.31%

As expected, since we’re not increasing the frequency of the CPU cores, the performance improvement is relatively limited. That said, improving the memory performance by using XMP 3.0 does help in memory-sensitive benchmark applications. We see the highest performance improvement of +11.88% in Geekbench 5.

When running Prime 95 Small FFTs with AVX2 enabled, the average CPU P-core clock is 5204 MHz, and the average CPU E-core clock is 4120 MHz with 1.119 volts. The average CPU temperature is 100 degrees Celsius. The ambient and water temperature is 26.4 and 31.0 degrees Celsius. The average CPU package power is 310.0 watts.

When running Prime 95 Small FFTs with AVX disabled, the average CPU P-core clock is 5456 MHz, and the average CPU E-core clock is 4299 MHz with 1.206 volts. The average CPU temperature is 100 degrees Celsius. The ambient and water temperature is 26.5 and 30.7 degrees Celsius. The average CPU package power is 315.2 watts.

OC Strategy #2: AI Overclocking + XMP 3.0

In our second overclocking strategy, we use the ASUS AI Overclocking feature integrated into the ASUS ROG BIOS.

ASUS AI Overclocking

ASUS AI overclocking uses a unique strategy for automatic overclocking. Instead of working with preset frequency and voltage profiles, the system will monitor the CPU and cooling system throughout an initial testing phase. Based on its findings, it will then predict the optimal settings. The system will automatically guide the overclocking process and adjust voltages and frequency to match your cooling system.

The better your cooling, the higher your AI overclock.

There are three steps to enabling AI overclocking. First, reset the BIOS to default settings. Then, reboot and enter the operating system. Run a couple of heavy workloads, such as Cinebench R23, Realbench, or Intel XTU, for 10 to 30 minutes. Then return to the BIOS and enter the AI OC Guide menu from the top. Make sure to read through the explanation and click Enable AI when ready.

In addition to automatic overclocking, AI Overclocking provides a lot of advanced information and suggestions in the AI Features menu. The information includes:

• P0 VID and SP values for the P-cores and E-cores
• Turbo Ratio suggested overclocking parameters
• Adaptive Voltage and AC loadline suggested parameters

The SP value is based on the combination of maximum boost frequency, temperature, and P0 VID. Generally, it indicates the quality of a particular core. A higher SP value would indicate a better-quality core with superior overclocking capabilities, though it’s not an exact science. The overclocking suggestions are based on a continued evaluation of your CPU thermal solution.

There are a couple of additional options to finetune the AI Overclock configuration, including one we’ll use in this configuration:

Optimized AVX Frequency allows you to toggle between Normal Use and Heavy Use, where Heavy Use should be selected if you’re running extreme workloads like Prime95 with AVX enabled.

After enabling AI Overclock and adjusting the Optimized AVX Frequency, the following settings have changed:

• Performance Core Ratio: AI Optimized
  • 1-Active P-core Ratio: 60X -> 63X (OCTVB)
  • 2-Active P-core Ratio: 60X -> 62X (OCTVB)
  • 3-Active P-core Ratio: 56X -> 62X (OCTVB)
  • 4-Active P-core Ratio: 56X -> 60X (OCTVB)
  • 5-Active P-core Ratio: 56X -> 60X (OCTVB)
  • 6-Active P-core Ratio: 56X -> 58X (OCTVB)
  • 7-Active P-core Ratio: 56X -> 58X (OCTVB)
  • 8-Active P-core Ratio: 56X -> 58X (OCTVB)
• Efficient Core Ratio: Ai Optimized
  • Up to 5-Active E-core Ratio: 43X -> 46X
  • Up to 16-Active E-core Ratio: 43X -> 43X
• Optimized AVX Frequency: Heavy Use
• AVX offset: 3X
• Per P-core Ratio Limit:
  • Core 0, 1, and 6: 63X
  • Core 2, 3, 4, 5, 7: 58X
• Adaptive Voltage: 1.458V
• Package Temperature Threshold: 90C
• AC loadline: 0.03
• VRM Loadline: 3

XMP Tweaked

XMP Tweaked is a new option available under Ai Overclock Tuner in addition to XMP I and XMP II. All three settings load the memory kit XMP profile but do it slightly differently.

• XMP I loads only the primary timings, frequency, and voltage. The secondary timings are adjusted by the motherboard auto-rules
• XMP II loads the complete XMP profile, including the primary and secondary timings, the memory frequency, and the voltage.
• XMP Tweaked loads the complete XMP profile and makes further adjustments to various timings if possible

In this OC Strategy, I try XMP Tweaked instead of the XMP II I usually use.

BIOS Settings & Benchmark Results

Upon entering the BIOS

• Go to the Extreme Tweaker menu
• Enter the Ai Overclocking Guide
• Go through the guide, then click Enable AI
• Set Ai Overclock Tuner to XMP Tweaked
• Set Optimized AVX Frequency to Heavy AVX

Then save and exit the BIOS.

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

• SuperPI 4M: +2.47%
• Geekbench 5 (single): +6.46%
• Geekbench 5 (multi): +20.18%
• Cinebench R23 Single: +0.81%
• Cinebench R23 Multi: +3.72%
• CPU-Z V17.01.64 Single: -2.25%
• CPU-Z V17.01.64 Multi: +0.21%
• V-Ray 5: +4.35%
• AI Benchmark: +14.15%
• 3DMark Night Raid: +1.65%
• CS:GO FPS Bench: +1.49%
• Tomb Raider: +0.00%
• Final Fantasy XV: +0.68%

Here are the 3DMark CPU Profile scores

• CPU Profile 1 Thread: +0.86%
• CPU Profile 2 Threads: +2.97%
• CPU Profile 4 Threads: +2.10%
• CPU Profile 8 Threads: +0.11%
• CPU Profile 16 Threads: +0.58%
• CPU Profile Max Threads: +0.37%

With AI Overclocking, we increase the processor frequency slightly over the stock settings from 6.0 GHz to 6.3 GHz P-Cores. Therefore, we expect a slight performance uplift, which we see in the benchmark results. The performance uplift from AI Overclocking is surprisingly good. We have the best performance improvement of +23.35% in Geekbench 5.

A quick side note on the CPU-Z single thread performance. The hardware P-state registers adjust the core priority ranking when you change the CPU ratios, essentially ensuring that the highest-clocked core has the highest priority. However, the CPU-Z benchmark seems to ignore this core prioritization and instead forces the workload to run on a favored core. The AI overclock set the Per Core Ratio limit of P-Core 4 to 58X, which is lower than the default 60X. Hence the lower performance.

When running Prime 95 Small FFTs with AVX2 enabled, unfortunately, the CPU is not stable. That’s not surprising since AI Overclock is designed for typical workload scenarios and not extreme workloads like Prime95 with AVX2 enabled.

When running Prime 95 Small FFTs with AVX disabled, the average CPU P-core clock is 5540 MHz, and the average CPU E-core clock is 4300 MHz with 1.156 volts. The average CPU temperature is 95 degrees Celsius. The ambient and water temperature is 25.7 and 30.7 degrees Celsius. The average CPU package power is 291.5 watts.

OC Strategy #3: Manual Overclock + XMP 3.0

In our third overclocking strategy, we will pursue a manual overclock. The approach is based mainly on tuning the Turbo Ratio configuration and adjusting the Adaptive Voltage mode.

I will keep the configuration itself simple for this OC Strategy. But I want to provide you with all the context to understand what I’m doing. Therefore, I will cover the frequency and voltage configuration topics in detail.

Intel Turbo Ratios

Generally speaking, on Intel platforms, there are two ways to manually configure the CPU ratio: sync all cores or use turbo ratio configuration.

Sync All Cores sets a single fixed ratio applied to all cores. This is very much the historical way of Intel CPU overclocking. Turbo Ratio configuration allows us to modify the default Intel frequency specification and configure an overclock for various scenarios.

Before we go any further, there are three critical elements of understanding any Turbo Ratio configuration:

1. You can configure the maximum allowed CPU core ratio for any number of active cores
2. You can configure the maximum allowed CPU core ratio for a given CPU core
3. The turbo ratio configuration for P-core and E-cores is independent.

To explain the first point, let’s take the default configuration of the 13900KS. The 13900KS has a total of 8 P-cores. Therefore, we can configure the maximum allowed P-core ratio for when 1 P-core is active, when 2 P-cores are active, all the way up to when 8 P-cores are active. The standard configuration allows every P-Core to boost to 5.6 GHz when all cores are active.

In our overclock, we adjust the Turbo Ratio configuration to boost

• to 6.2 GHz when up to 4 P-cores are active,
• to 5.7 GHz when up to 8 P-cores are active.

To explain the second point, again, let’s consider the 13900KS default specification. While the 13900KS has 8 identical P-cores, two of those cores are called the favored cores. The maximum allowed frequency for the favored cores is 6.0 GHz, whereas the other cores are limited to 5.6 GHz.

If we combine point 1 and point 2, we can identify the following scenarios:

• If 1 P-core is active and it is a favored P-core, then that core will run at 6.0 GHz
• If 1 P-core is active and it is not a favored P-core, then that core will run at 5.6 GHz
• If 2 P-cores are active and both are a favorite P-core, then both cores will run at 6.0 GHz
• If 2 P-cores are active and one is a favorite P-core while the other isn’t, then one core will run at 6.0 GHz, and the other will run at 5.6 GHz.
• If 2 P-cores are active and neither is a favorite P-core, then both cores will run at 5.6 GHz
• If 3 or more P-cores are active, then any P-core will run at 5.6 GHz.

Often, we use the Per Core Ratio limit to prevent the weaker cores from boosting to the highest frequency. For example, in this OC strategy, I restrict Core1 to 61X and Core7 to 60X while the other cores can boost to 62X.

To explain the third point, again, let’s refer to the 13900KS. The CPU has a total of eight P-cores and sixteen E-cores. While the P-cores can boost up to 6.0 GHz, the E-cores can only boost up to 4.3 GHz.

The P-core rules for maximum allowed frequency can also be applied to the E-cores. However, with one major caveat: the E-core CPU ratio can only be controlled in groups of 4 E-cores. So, for the 13900KS, since it has sixteen E-cores in total, we can configure the maximum allowed core ratio for a total of four groups of four E-cores. However, we can still configure the maximum allowed frequency for 1 active E-core up to 16 active E-cores.

We increase the E-core frequency from 4.3 GHz to 4.5 GHz in this OC Strategy.

Before we move on to the voltage, let’s first quickly talk about TjMax and thermal throttling.

TjMax & Thermal Throttling

Tjunction max, or TjMax, is the maximum thermal junction temperature allowed for a processor. If the operating temperature exceeds TjMax, internal thermal control mechanisms will reduce the operating frequency until the temperature is below TjMax.

The TjMax for Raptor Lake CPUs is 100 degrees Celsius; however, it can be manually increased to 115 degrees though it’s not recommended for long-term use.

The maximum CPU operating temperature is a sensitive topic in the PC enthusiast space. In the next couple of minutes, I want to dispel two myths regarding TjMax:

1. Running at TjMax is extremely dangerous
2. Thermal throttling causes severe performance degradation

Myth #1

The first myth is that it’s unsafe to run the processor at TjMax. I often get questions or comments on my Prime95 stability test results because the CPU runs at the maximum temperature of 100 degrees Celsius. I get asked, “Is this safe?” or told, “It’s very risky to run that temperature!”

The irony of those comments is that the operating temperature is the most conservative part of my overclocking guides. It’s the one parameter I don’t run out of spec! For this 13900KS, Intel says I can run all cores at 5.6 GHz with 1.38V up to 100 degrees Celsius. In my OC strategies, I increase the frequency and voltage, but maintain the maximum 100 degrees Celsius specification.

The bottom line is that Intel puts its money where its mouth is: Intel makes a legal commitment by its three-year warranty that your processor will perform according to its specification, which includes an operating temperature of up to 100 degrees Celsius.

Myth #2

The second myth is that thermal throttling will significantly degrade your performance and thus must be avoided at all costs. While that may have been true a decade ago, the truth is that modern CPUs have extremely advanced performance-maximizing technologies. These technologies maximize performance in any given situation, including when the operating temperature is near TjMax.

The Raptor Lake TjMax thermal throttling mechanism is quite simple: when the operating temperature reaches TjMax, the CPU reduces the CPU Turbo Ratio until the temperature is under TjMax.

In my overclocking strategies, I try leveraging these technologies to improve performance. In all-core workloads, my objective is to maximize the frequency before and at the thermal throttling point. That’s why you’ll see me include details on the clock frequencies in Prime95 workloads and focus on improving that, along with benchmark performance.

On Raptor Lake processors, the user can configure the TjMax between 62 and 115 degrees Celsius.

To illustrate the performance throttling, I tested the 13900KS with P-cores at 5.8 GHz and E-cores at 4.5 GHz in Cinebench R23. As you can see from the data, when TjMax is set to 62 degrees Celsius, you lose about 15% due to thermal throttling compared to when the TjMax is at its default of 100 degrees Celsius.

To illustrate the frequency throttling behavior, I set TjMax to 62 degrees Celsius and ran Cinebench R23. As you can see from the data, the CPU almost instantaneously responds to a temperature exceeding TjMax by reducing the P-core frequency from 5.8 to 4.7 GHz and the E-core frequency from 4.5 to 3.7 GHz.

As a side note, ASUS also has its own TjMax-style thermal throttling technology. You can use it by enabling the Package Temperature Threshold option. This technology will monitor the ASUS calibrated socket temperature and lower the Turbo Boost 2.0 PL1 limit if it exceeds the target temperature. Limiting the CPU power consumption will also reduce the operating temperature. While ASUS technology works pretty well, it is less accurate and responsive than Intel’s TjMax mechanism.

To illustrate the frequency throttling behavior, I set TjMax to 115 degrees Celsius and the Package Temperature Threshold to 62 degrees Celsius and ran Cinebench R23. As you can see from the data, the CPU responds much slower to the change in operating temperature.

At the start of the benchmark, the CPU temperature spikes up high (even above 100 degrees Celsius) before the ASUS technology aggressively adjusts the power limit to reduce the temperature. The temperature first drops significantly below the 62 degrees Celsius target, then gradually increases as the ASUS algorithm adjusts the power limit. The actual CPU Package Temperature with this setting is around 75 degrees Celsius.

Returning to the myth that TjMax thermal throttling significantly degrades performance, while I’ve tried to dispel the myth, it contains an element of truth. As is evident from the data, the performance will be reduced when the operating temperature is above TjMax. So, it’s essential to ensure you pair your powerful CPU with a cooling solution that will prevent too-aggressive performance throttling.

For this OC Strategy, I leave the TjMax at the default and disable the ASUS Package Temperature Threshold. Effectively, the maximum allowed operating temperature for the CPU is thus 100 degrees Celsius.

VccIA Voltage Rail

Voltage tuning, specifically undervolting, is the primary avenue to extract more performance from your Core i9 Raptor Lake CPU. That’s mainly because the CPU will try to boost to the maximum allowed temperature of 100 degrees Celsius, at which point it throttles the frequency. By lowering the operating voltage for a given frequency, you reduce the temperature for that frequency, thus enabling higher frequency before performance throttling.

On Raptor Lake, the VccIA voltage rail drives the voltage for the CPU cores, P-core and E-core, and the Ring. That means a single voltage is used for all these parts of the CPU. How that voltage is configured is straightforward yet complex.

There are three key aspects to understanding how voltage is configured on Intel platforms: the CPU, the motherboard design, and the voltage regulator.

Let’s start with the CPU side of the story.

CPU V/F Curves

An Intel CPU relies on many factory-fused voltage-frequency curves, or V/F curves, to regulate its dynamic compute performance behavior. A V/F curve describes the relationship between an operating frequency and the voltage required for that frequency. A lot of parts inside your CPU have a V/f curve, including those relevant to the VccIA voltage rail:

• Each of the 8 P-cores
• Each of the 4 E-core groups of 4 cores
• The ring

In the case of the Core i9-13900KS, the VccIA voltage rail is affected by no less than 13 different voltage-frequency curves. For this guide, I extracted the voltage-frequency curves of this specific 13900KS.

Based on these V/F curves, to get a specific voltage provided via the VccIA voltage rail, the CPU issues an SVID request to the voltage controller. The VID requested is the highest among all the requested voltages according to every V/F curve affecting the voltage rail.

Let’s take an example:

The highest voltage requested according to the relevant V/F curves is 1.34V by P-Core 0. This will be the VID request to the voltage controller.

Here’s another example:

In this case, the highest voltage requested according to the relevant V/F curves is 1.28V by E-Core Group 0. This will be the SVID request to the voltage controller.

The goal of the SVID voltage request from the CPU to the voltage regulator is that the effective voltage at the CPU die is equal to the requested voltage. However, as overclockers and enthusiasts know very well, that’s not always the case. That’s because there are a lot of electrical components between the voltage regulator and the CPU die.

To avoid the voltage delivered to the CPU die being lower than the requested voltage, we have two main tools: (1) AC-DC loadline and (2) VRM loadline.

AC DC Loadline

The AC DC loadline is designed for motherboard engineers to bring into the voltage-frequency curve equation the electrical impedance of the motherboard design. Electrical impedance is the opposition to alternating current and is affected by the VRM components, the PCB layout, and quality.

Electrical impedance can significantly affect the effective voltage at the CPU die. Therefore, there may be a significant difference between the requested and effective voltage. We can define the AC DC loadline parameters to account for this difference.

Adjusting the AC loadline offsets the requested voltage, defined by the factory-fused voltage-frequency curve, to account for any electrical impedance.

For example, suppose we know that a 1.4V voltage output by the voltage controller, as requested by the CPU, results in an effective voltage of 1.35V at the CPU die due to electrical impedance. In that case, we can configure the AC loadline such that the CPU requests 1.45V instead.

Adjusting the DC loadline informs the CPU about the expected effective voltage at the CPU die.

For example, suppose the voltage-frequency point is 1.4V, but we’ve configured the AC loadline such that it’s bumped up to 1.45V. If we don’t tell the CPU that a 1.45V VID request results in a 1.4V effective core voltage, then the CPU assumes the effective voltage is 1.45V. That skews any power management metrics that rely on the VID information.

By adjusting the DC loadline to account for the expected difference in voltage between the requested VID and the effective voltage, we ensure the CPU power management unit has the correct information to do its power calculations. If we don’t do this, we end up with power consumption reporting vastly different from reality which may impact the CPU’s Turbo Boost behavior.

VRM loadline

The VRM loadline is essential for two reasons.

1. First, it determines the Vdroop, which is the decrease in voltage during a sustained heavy workload.
2. Second, it determines the undershoot and overshoot, which is a voltage spike during a transient load.

Vdroop is the decrease in voltage when the CPU goes from idle to load. You want your CPU to be stable in all scenarios, so knowing the lowest voltage the CPU runs at is very important. After all, if the voltage is too low, the overclock won’t be stable.

Undershoot and its counterpart, overshoot, is a brief voltage spike that occurs when the CPU switches from idle to load or from load to idle. These spikes cannot be measured easily and usually require an expensive oscilloscope to detect. I highly recommend the ElmorLabs article titled VRM Load-Line Visualized to see a great picture of undershoot and overshoot in action.

While undershoot and overshoot are temporary spikes, an undershoot that’s too low can cause instability.

The VRM loadline setting is relevant to our overclock in two ways.

1. First, it helps control the voltage in heavy workloads. Enthusiasts often rely on VRM loadline tuning to decrease the voltage in a heavy workload, which yields lower power consumption and operating temperatures.
2. Second, since the VRM loadline can significantly affect the effective CPU core voltage, we must account for it when setting the DC loadline.

Motherboard Auto-Rules

Since we are keeping the voltage configuration as simple as possible for this OC strategy, we will rely on the motherboard auto-rules. Auto-rules are pre-programmed automatic changes to BIOS settings.

This ASUS motherboard has auto-rules to adjust the AC DC and VRM loadline when the user sets manual Turbo Ratios. In my specific case, the auto-rules make the following adjustments:

• AC LL: 0.50 -> 0.17
• DC LL: 1.10 -> 0.98
• VRM LL: 3 -> Level 4

As discussed, the VRM loadline will affect the Vdroop, which is the voltage drop under load. To illustrate this, I compare the effective voltage in idle and load to the voltage-frequency curve. As you can see from the chart, the effective voltage under load is significantly lower than we would expect from the voltage-frequency curve.

In practical terms, this means we are undervolting the CPU already. Unfortunately, the auto-rules undervolt the CPU too much, making it unstable at frequencies higher than 57X. To work around this, I adjust the VRM loadline to Level 6. This reduces the Vdroop and will lessen the decrease of voltage under load. This manual adjustment also changes the AC and DC loadline configuration.

• AC LL: 0.50 -> 0.17 -> 0.01
• DC LL: 1.10 -> 0.98 -> 0.49
• VRM LL: Level 3 -> Level 4 -> Level 6

Again, I plot the voltage in idle and load scenarios. You can see that the voltage under load has significantly increased from the auto-rule configuration but is still considerably lower than the factory-fused voltage frequency curve. With this configuration, 57X was stable in medium-load all-core workloads though the voltage is too high to stay below TjMax when running Prime95.

One last thing I want to highlight: did you notice that the idle voltage for ratios 60X, 61X, and 62X is 1.465V instead of the expected 1.435V? That’s not an HWiNFO reporting error or related to the loadline configuration, but rather a curiosity with the Ring voltage.

Before I can explain the curiosity, first we have to talk about the Adaptive Voltage Mode.

Intel Adaptive Voltage Mode

There are two main ways of configuring the voltage for the CPU cores: override mode and adaptive mode.

• Override mode specifies a single static voltage across all ratios. It is mainly used for extreme overclocking purposes where stability at high frequencies is the only consideration.
• Adaptive mode is the standard mode of operation. In Adaptive Mode, the CPU relies on the V/F curves to set the appropriate voltage for the VccIA voltage rail.

Both override and adaptive mode settings can be configured via the CPU registers. So, in effect, we control the CPU VID request to the voltage controller. This is Intel’s intended way of overclocking.

Of course, most voltage controllers also allow independent configuration. For example, they enable us to configure a voltage offset to the requested voltage. It is often unclear from the motherboard BIOSes which method of setting the CPU core voltage we’re using when we type in the desired voltage. For the purpose of this guide, however, let’s ignore the capabilities of the voltage controllers and focus on Intel’s intended way of overclocking.

We can specify a voltage offset for override and adaptive modes. Of course, this doesn’t make much sense for override mode – if you set 1.35V with a +50mV offset, you could just set 1.40V – but it can be helpful in adaptive mode. The entire V/F curve can be offset by up to 500mV in both directions.

As I mentioned, Intel offers great granularity for tuning the many V/F curves inside the CPU. Let’s forget about the E-cores and Ring to keep things simple and assume a case where we set a global adaptive voltage for the CPU P-cores. Now let’s dig into what happens when we set a global adaptive voltage.

First, disregarding any user-set global or V/f point offsets, the adaptive voltage set in the BIOS is mapped against what’s called the “OC ratio.” The “OC Ratio” is the highest ratio configured for the CPU. When you leave everything at default, the OC ratio is determined by the default maximum turbo ratio. In the case of the 13900KS, that ratio would be 60X.

When you manually overclock, the OC ratio is the highest ratio you configure across all the various settings and options.

Second, specific rules govern what adaptive voltage can be set.

A) the voltage set for a given ratio n must be higher than or equal to the voltage set for ratio n-1.

Suppose our 13900KS runs 60X at 1.43V. In that case, setting the adaptive voltage, mapped to 60X, lower than 1.43V, is pointless. 60X will always run at 1.43V or higher. Usually, BIOSes will allow you to configure lower values. However, the CPU’s internal mechanisms will override your configuration if it doesn’t follow the rules.

B) the adaptive voltage configured for any ratio below the maximum default turbo ratio will be ignored.

Take the same example of the 13900KS, specified to run 60x at 1.43V. If you try to configure all cores to 57X and set 1.20V, the CPU will ignore this because it has its own factory-fused target voltage for all ratios up to 60X and will use this voltage. You can only change the voltage of the OC Ratio, which, as mentioned before, on the 13900KS, is 60X and up.

C) for ratios between the OC Ratio and the next highest factory-fused V/f point, the voltage is interpolated between the set adaptive voltage and the factory-fused voltage.

Returning to our example of our 13900KS specified to run 60X at 1.43V, let’s say we manually configure the OC ratio to be 65X at 1.5V. The target voltage for ratios 61X, 62X, 63X, and 64X will now be interpolated between 1.43V and 1.50V.

So, in conclusion.

The adaptive voltage set in BIOS is mapped against the “OC Ratio.” Unless explicitly programmed, the OC Ratio is the highest ratio configured for the CPU across all settings, including by core usage, per core ratio limit, and OCTVB. The voltage for ratios lower than the OC ratio is set either by its factory-fused V/f point or, if there’s no V/f point, interpolated between the next and previous V/f point.

A feature Intel introduced with Alder Lake that’s been flying under the radar is Per Core Voltage. To make a long story short, we can independently program an adaptive voltage for each core. If we program the adaptive voltage for a specific core, it is mapped against that core’s OC ratio, which is the Per Core Ratio Limit. The same principle also applies to the E-core groups and the Ring, by the way.

For this OC Strategy, I use an Adaptive Voltage of 1.435V.

Now let’s return to that Ring voltage curiosity I mentioned before. Since we use an adaptive voltage of 1.435V for 62X, we’d assume this to be the effective voltage in idle at 62X. But looking at the voltage-frequency curves, it seems the voltage is around 1.46V.

To make a long story short: I don’t quite know what’s going on. Here’s what seems to be going on: when we set the Adaptive Voltage for the CPU cores, it also sets this adaptive voltage for the Ring. When we check HWiNFO, the Ring VID is 1.435V, precisely as we set for the cores in the BIOS. Then, it seems there’s a rule where the Core VID must be a minimum of 30 mV higher than the Ring VID. That shows up as 1.465V VID in HWiNFO.

Why and how … I don’t know. I also confirmed this behavior on the 13900K, but I’ve yet to figure out this Raptor Lake mystery. If I find the reason, I’ll let you guys know.

AVX Negative Ratio Offset

We set an AVX negative ratio offset of 5, reducing the CPU frequency in some AVX workloads.

While AVX negative ratio offset has been on Intel CPUs for ages, since Alder Lake, there have been a couple of changes in the behavior. I discussed that in previous Raptor Lake guides. The long story short is that the AVX offset doesn’t apply to all AVX workloads, such as Cinebench R23.

BIOS Settings & Benchmark Results

Upon entering the BIOS

• Go to the Extreme Tweaker menu
• Set Ai Overclock Tuner to XMP II
• Set ASUS MultiCore Enhancement to Enabled – Remove All Limits
• Set DRAM Frequency to DDR5-7600MHz
• Set Performance Core Ratio to By Core Usage
  • Set 1-Core to 4-Core Ratio Limit to 62
  • Set 5-Core to 8-Core Ratio Limit to 57
• Enter the Specific Performance Core submenu
  • Set Performance Core0, Core2, Core3, Core4, Core5, and Core6 Specific Ratio Limit to 62
  • Set Performance Core1 Specific Ratio Limit to 61
  • Set Performance Core7 Specific Ratio Limit to 60
• Leave the Specific Performance Core submenu
• Set Efficient Core Ratio to Sync All Cores
  • Set ALL-Core Ratio limit to 45
• Enter the AVX Related Controls submenu
  • Set AVX2 Ratio Offset to per-core Ratio Limit to User Specify
  • Set AVX2 Ratio Offset to 5
• Leave the AVX Related Controls submenu
• Enter the DIGI+ VRM submenu
  • Set CPU Load-line Calibration to Level 6
• Leave the DIGI+ VRM submenu
• Set Global Core SVID Voltage to Adaptive Mode
  • Set Offset Mode Sign to +
  • Set Additional Turbo Mode CPU Core Voltage to 1.435
• Set High DRAM Voltage Mode to Enabled
• Set DRAM VDD Voltage to 1.45
• Set DRAM VDDQ Voltage to 1.45

Then save and exit the BIOS.

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

• SuperPI 4M: +4.16%
• Geekbench 5 (single): +5.04%
• Geekbench 5 (multi): +21.20%
• Cinebench R23 Single: +3.93%
• Cinebench R23 Multi: +7.05%
• CPU-Z V17.01.64 Single: +4.08%
• CPU-Z V17.01.64 Multi: +3.26%
• V-Ray 5: +7.25%
• AI Benchmark: +18.87%
• 3DMark Night Raid: +2.40%
• CS:GO FPS Bench: +1.48%
• Tomb Raider: +0.48%
• Final Fantasy XV: +0.86%

Here are the 3DMark CPU Profile scores

• CPU Profile 1 Thread: +1.64%
• CPU Profile 2 Threads: +4.25%
• CPU Profile 4 Threads: +3.77%
• CPU Profile 8 Threads: +1.90%
• CPU Profile 16 Threads: +7.17%
• CPU Profile Max Threads: +4.38%

Even though we only slightly increase the single and all-core P-core boost frequency, thanks to the aggressive undervolting, we’re seeing a decent performance uplift across the board. We see a maximum multi-thread performance improvement of 21.20% in Geekbench 5.

When running Prime 95 Small FFTs with AVX2 enabled, the average CPU P-core clock is 5274 MHz, and the average CPU E-core clock is 4175 MHz with 1.111 volts. The average CPU temperature is 100 degrees Celsius. The ambient and water temperature is 25.9 and 31.9 degrees Celsius. The average CPU package power is 308.8 watts.

When running Prime 95 Small FFTs with AVX disabled, the average CPU P-core clock is 5533 MHz, and the average CPU E-core clock is 4494 MHz with 1.190 volts. The average CPU temperature is 100 degrees Celsius. The ambient and water temperature is 25.7 and 30.7 degrees Celsius. The average CPU package power is 313.6 watts.

OC Strategy #4: OCTVB + XMP 3.0

In our fourth and final overclocking strategy, we resort to advanced manual overclocking with Intel’s Overclocking Thermal Velocity Boost toolkit. Our main objective is to squeeze a little more frequency out of the chip and provide higher performance in light workloads.

OCTVB – OverClocking Thermal Velocity Boost

In 2018 Intel introduced Thermal Velocity Boost along with the Core i9-8950HK Coffee Lake mobile flagship processor, and it’s since been an indispensable feature on Intel Core processors.

Thermal Velocity Boost does two things.

1. First, it decreases the operating voltage if the CPU temperature is below the TjMax.
2. Two, it opportunistically increases the clock frequency above the Turbo Boost 2.0 and 3.0 frequencies based on how much the processor operates below its maximum temperature.

With the introduction of the Intel Cryo Cooling Technology in 2020, Intel opened up the TVB configuration to motherboard vendors. The feature is named OverClocking Thermal Velocity Boost, or OCTVB for short.

The easiest way to think of OCTVB is limiting, or clipping, the maximum allowed CPU ratio based on the CPU operating temperature. The hotter the CPU, the more you clip the CPU ratio. OCTVB is based on the by core usage Turbo Ratio configuration. For each number of active cores, you can define two temperature points, each with a unique number of “down-bins’. A down-bin is essentially the number of ratios you want to drop.

Let’s take the configuration of this OC Strategy.

When 1 P-core is active, the base ratio is 63X, so the frequency will be 6.3 GHz. However, when the temperature is 68 degrees Celsius, the ratio is clipped by 1X. That means the maximum ratio is now 62X.

When all 8 P-cores are active, the base ratio is 61X, so the frequency will be 6.1 GHz. However, when the temperature is 75 degrees Celsius, the ratio is clipped by 2X. That means the maximum ratio is now 59X. When the temperature hits 85 degrees Celsius, the ratio is clipped once more by 1X. So, the resulting maximum ratio is now 58X.

Note that we have TjMax configured at the default of 100 degrees Celsius. So beyond 100 degrees Celsius, the CPU will automatically reduce the frequency to stay within the thermal limit.

Testing an OCTVB configuration is notoriously tricky because you can’t simply stress-test as you’d typically do. So, most of OCTVB validation is just running your benchmark test suite to see if there are any instabilities. Still, as I’ve demonstrated in previous SkatterBencher guides, we can leverage the Thermal Velocity Boost Voltage Optimizations feature to get a rough idea of how many extra bins we can get with lower temperatures.

BIOS Settings & Benchmark Results

Upon entering the BIOS

• Go to the Extreme Tweaker menu
• Set Ai Overclock Tuner to XMP II
• Set ASUS MultiCore Enhancement to Enabled – Remove All Limits
• Set DRAM Frequency to DDR5-7600MHz
• Set Performance Core Ratio to By Core Usage
  • Set 1-Core to 3-Core Ratio Limit to 63
  • Set 4-Core to 6-Core Ratio Limit to 62
  • Set 7-Core and 8-Core Ratio Limit to 61
• Enter the Specific Performance Core submenu
  • Set Performance Core0, Core2, Core3, Core5, and Core6 Specific Ratio Limit to 63
  • Set Performance Core4 Specific Ratio Limit to 62
  • Set Performance Core1 and Core7 Specific Ratio Limit to 61
  • For each Performance Core, set Specific Voltage to Adaptive Mode
  • For each Performance Core, set Offset mode Sign to Plus
  • For each Performance Core, set Additional Turbo Mode CPU Voltage to 1.475
• Leave the Specific Performance Core submenu
• Set Efficient Core Ratio to Sync All Cores
  • Set ALL-Core Ratio limit to 45
• Enter the Specific Efficient Core submenu
  • For each Efficient Core Group, set Specific Ratio limit to 45
  • For each Efficient Core Group, set Specific Voltage to Adaptive Mode
  • For each Efficient Core Group, set Offset mode Sign to Plus
  • For each Efficient Core Group, set Additional Turbo Mode CPU Voltage to 1.3
• Leave the Specific Efficient Core submenu
• Enter the AVX Related Controls submenu
  • Set AVX2 Ratio Offset to per-core Ratio Limit to User Specify
  • Set AVX2 Ratio Offset to 6
• Leave the AVX Related Controls submenu
• Enter the DIGI+ VRM submenu
  • Set CPU Load-line Calibration to Level 6
• Leave the DIGI+ VRM submenu
• Enter the Thermal Velocity Boost submenu
  • Set TVB Voltage Optimizations to Disabled
  • Set Enhanced TVB to Disabled
  • Set Overclocking TVB to Enabled
  • Set 1-Core to 8-Core Active to Enabled
  • For 1-Core and 2-Core Active, set Temperature A to 68
  • For 3-Core to 5-Core Active, set Temperature A to 70
  • For 6-Core Active, set Temperature A to 60
  • For 7-Core and 8-Core Active, set Temperature A to 75
  • For each Core Active, set Negative Ratio Offset A to User Specify
  • For 1-Core to 6-Core Active, set Ratio Offset A to 1
  • For 7-Core and 8-Core Active, set Ratio Offset A to 2
  • For 1-Core and 2-Core Active, set Temperature B to 115
  • For 3-Core to 6-Core Active, set Temperature B to 90
  • For 7-Core and 8-Core Active, set Temperature B to 85
  • For each Core Active, set Negative Ratio Offset B to User Specify
  • For 1-Core and 2-Core Active, set Ratio Offset B to 0
  • For 3-Core to 8-Core Active, set Ratio Offset B to 1
• Leave the Thermal Velocity Boost submenu
• Set Global Core SVID Voltage to Adaptive Mode
  • Set Offset Mode Sign to +
  • Set Additional Turbo Mode CPU Core Voltage to 1.475
• Set High DRAM Voltage Mode to Enabled
• Set DRAM VDD Voltage to 1.45
• Set DRAM VDDQ Voltage to 1.45

Then save and exit the BIOS.

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

• SuperPI 4M: +5.49%
• Geekbench 5 (single): +9.14%
• Geekbench 5 (multi): +21.21%
• Cinebench R23 Single: +4.27%
• Cinebench R23 Multi: +6.51%
• CPU-Z V17.01.64 Single: +4.27%
• CPU-Z V17.01.64 Multi: +4.05%
• V-Ray 5: +7.99%
• AI Benchmark: +19.79%
• 3DMark Night Raid: +2.90%
• CS:GO FPS Bench: +1.49%
• Tomb Raider: +0.48%
• Final Fantasy XV: +0.36%

Here are the 3DMark CPU Profile scores

• CPU Profile 1 Thread: +1.87%
• CPU Profile 2 Threads: +4.29%
• CPU Profile 4 Threads: +5.83%
• CPU Profile 8 Threads: +3.11%
• CPU Profile 16 Threads: +7.72%
• CPU Profile Max Threads: +4.79%

While this OC strategy allows us to squeeze more frequency out of our CPU, it won’t improve the performance in all scenarios because it’s configured through OCTVB. In fact, the performance improvements should only appear in light workloads such as the 3DMark CPU profile, which doesn’t push the CPU to the TjMax. We see the highest performance improvement of +21.21% in Geekbench 5 single.

When running Prime 95 Small FFTs with AVX2 enabled, the average CPU P-core clock is 5281 MHz, and the average CPU E-core clock is 4262 MHz with 1.112 volts. The average CPU temperature is 100 degrees Celsius. The ambient and water temperature is 25.4 and 31.0 degrees Celsius. The average CPU package power is 310.3 watts.

When running Prime 95 Small FFTs with AVX disabled, the average CPU P-core clock is 5538 MHz, and the average CPU E-core clock is 4439 MHz with 1.193 volts. The average CPU temperature is 100 degrees Celsius. The ambient and water temperature is 25.0 and 30.5 degrees Celsius. The average CPU package power is 312.9 watts.

Intel Core i9-13900KS: Conclusion

Alright, let us wrap this up.

Overclocking the Core i9-13900KS isn’t very different from overclocking any other Raptor Lake Core i9 processor. The main avenue for improving performance is by undervolting the CPU in all-core workloads, enabling XMP, and unleashing the Turbo Boost 2.0 limits.

I tried out a couple of 13900KS processors and found they were at least on par with or perhaps slightly better than my best 13900K processor. But that’s not to say there might be great 13900K CPUs better than the average 13900KS CPUs.

Getting the CPU to run over 6 GHz, even in light all-core workloads, is extraordinary. But that’s what we’re used to now with Raptor Lake.

Anyway, that’s all for today! I want to thank my Patreon supporters 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!