SkatterBencher #33: Intel UHD Graphics 770 Overclocked to 2378 MHz

skatterbencher 33 uhd graphics 770

We overclock the Intel UHD Graphics 770 up to 2378 MHz with the ROG Maximus Z690 Extreme motherboard and EK-Quantum water cooling.

I’ve been looking forward to this one ever since I found out Intel’s maximum ratios were not enough to get the most out of the integrated graphics. So, for the first time in a long while, it’s necessary to use BCLK overclocking on an Intel platform to achieve the best performance.

This was a fun one, hope you enjoy!

https://www.youtube.com/watch?v=Sihfywr0gy0

Intel UHD Graphics 770: Introduction

The Intel UHD Graphics 770 is the integrated graphics portion of the Intel Alder Lake Core i9-12900K processor. It features the Intel Xe-LP microarchitecture which was first introduced on 11th generation Tiger Lake and Rocket Lake.

During the Architecture Day 2020, Intel introduced the Xe-LP graphics architecture for integrated and entry applications. The Xe-LP architecture is designed specifically to achieve best-in-class performance efficiency on 4 vectors: graphics, AI, media, and display. It features many improvements over the Gen11 architecture. Intel claims Xe-LP provides users with a 2X generational improvement both for gaming and encoding.

Note that while the integrated graphics are based on the Xe-LP architecture, it still carries the UHD Graphics branding. That’s because Iris Xe branding is reserved only for systems populated with 128-bit (dual channel) memory and, I think, 80 or more Execution Units. Systems that don’t meet this specification must revert to UHD Graphics branding.

We already overclocked the Rocket Lake Xe-LP UHD Graphics 750 integrated graphics in SkatterBencher #28 where we managed to increase the frequency from 1.3GHz to 1.75GHz. While the graphics microarchitecture remained pretty much the same, there’s a significant difference in process technology. Rocket Lake is produced on the Intel 14nm++ process and Alder Lake is produced using the Intel 7 process technology formerly known as 10nm Enhanced SuperFin (ESF). Intel 7 is one step more advanced than Tiger Lake’s 10nm SuperFin process technology.

If we compare the stock frequencies, we can see that

  • Xe-LP on Intel 14nm++ clocks up to 1300 MHz on the Core i9-11900K
  • Xe-LP on Intel 10nm SuperFin clocks up to 1450 MHz on the Core i9-11900KB and Core i9-11980HK
  • Xe-LP on Intel 7 clocks up to 1550 MHz on the Core i9-12900K
Intel Xe-LP on 14nm++ 10nm SF and Intel 7 process node frequencies

So, we can expect to see higher overclocking frequencies compared to Rocket Lake.

In today’s video, we tackle overclocking the UHD Graphics 770. We will cover 4 overclocking strategies.

  • First, we enable ASUS Multi-Core Enhancement and Intel XMP
  • Second, we increase the graphics frequency to 2.1 GHz using the Graphics Ratio
  • Next, we use base clock frequency overclocking to further increase the graphics frequency
  • Lastly, we push the integrated graphics as far as we can with more aggressive voltages

However, before we jump into the overclocking let us quickly go over the hardware and benchmarks we use in this video. You’ll see there’s quite a couple of new things we’re doing with this system.

First things first, though. How does the integrated graphics actually work? And how does it get its frequency and voltage?

Intel UHD Graphics 770: Design, Clocking, and Voltage

As the term integrated graphics already reveals, the UHD Graphics 770 Xe-LP is integrated into the Alder Lake CPU die. While it takes a sizeable chunk of the total die the majority of the chip is used to place the 8 Golden Cove CPU P-cores and 8 Gracemont CPU E-cores, its cache, and the Ring bus.

alder lake building blocks

Slice, Unslice, & Display Block

The design of the Xe-LP integrated graphics can be separated into three parts: the “Slice”, the “Unslice”, and the display block.

The Slice is a cluster of sub-slices, each of which contains the elements of the actual graphics compute engine like the execution units. This is the part that powers your games and any compute tasks.

The Unslice holds the elements with fixed-function geometry capabilities and fixed-function media capabilities. So, if you’re encoding or decoding a video, or simply watch Netflix, this part ensures proper video encoding. It also contains the connection to the ring bus via which the IGP can send and receive data from the system memory.

The display block contains support for the display outputs like HDMI or DisplayPort.

intel xe-lp architecture

The slice part of the UHD Graphics 770 inside our desktop Alder Lake Core i9-12900K contains 2 sub-slices with each 16 execution units for a total of 32 execution units. That’s far less than the 6 sub-slices and 96 execution units present in the mobile Alder Lake parts. Each execution unit also has 7 threads, totaling 224 threads.

Intel Alder Lake UHD Graphics specifications

Clocking

When it comes to clocking, the frequency is based on the base clock frequency or reference clock. By default, the BCLK is 100MHz, but it can be increased when overclocking. For the integrated UHD Graphics 770, the base clock frequency is first halved, then multiplied by the Graphics Ratio to obtain the final operating frequency.

In my Alder Lake launch article, I said that like on Rocket Lake Alder Lake does not support independent Slice and Unslice frequencies. However, that’s not entirely correct.

Unlike Rocket Lake, on Alder Lake the Slice and Unslice are 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, just like on Rocket Lake, we only have control over the Slice frequency using the Graphics Ratio.

On Alder Lake CPUs, the maximum Graphics Ratio Limit is 42. This is the same as Rocket Lake but lower than on the previous generation Comet Lake where the maximum ratio limit was 60. I’m not sure why the maximum ratio has been reduced.

Do note that the display output is driven by a separate fixed clock frequency and is not affected by changing the BCLK frequency or the graphics ratio.

VccGT (GT Core Voltage)

When it comes to providing an operating voltage to the graphics core, the Alder Lake processor has a dedicated voltage rail called VccGT or GT Core Voltage. It is by default around 1.05V.

The Slice and Unslice share the same voltage rail so their operating voltage will be the same.

Configuring the voltage for the graphics cores is very similar to the CPU cores. Alder Lake supports both adaptive and override voltages. When using Adaptive voltage mode, the graphics voltage will be reduced in idle whereas in Override voltage mode the voltage will remain constant.

Graphics Dynamic Frequency

The Alder Lake integrated UHD Graphics 770 supports the Graphics Dynamic Frequency feature. The Dynamic Frequency capability is designed to allow the processor to assess its thermals, current, and power to come up with a dynamic upper limit on its frequency.

On processors with Graphics Overclocking capability, the integrated graphics cores can run at higher frequencies when operating conditions allow.

When Graphics Dynamic Frequency is enabled, the graphics cores could be running at any ratio in the inclusive range between the Maximum Dynamic Frequency (RP0) and the Graphics Base Frequency (RP1). In the case of our Core i9-12900K, the graphics base frequency is 300 MHz, and the default maximum dynamic frequency is 1550 MHz. When overclocking, we simply increase this Maximum Dynamic Frequency (RP0) to a higher value.

As mentioned, the graphics frequency is in part a function of the available thermal, current, and power headroom. This is managed by the Turbo Boost 2.0 technology we are all too familiar with. Turbo Boost 2.0 will ensure the available power budget is distributed among the CPU and Graphics cores depending on the workload. So, if you have a heavy graphics-dependent workload and no CPU load, it will decrease the CPU frequency in favor of boosting the Graphics frequency higher.

Of course, knowing the power budget of the Graphics portion of the Alder Lake CPU is only 15W, we don’t expect this power budgeting to have a great impact on our overall system performance.

Intel UHD Graphics 770: Platform Overview

Along with the Intel Core i9-12900K processor and its Intel UHD Graphics 770 integrated graphics, in this guide, we will be using an ASUS ROG Maximus Z690 Extreme motherboard, a pair of AORUS DDR5-6200 Hynix memory sticks, an AORUS 512GB M.2 NVMe SSD, a Seasonic Prime 850W Platinum power supply, the ElmorLabs Easy Fan Controller, the EK-Quantum Magnitude water block, and EK-Quantum water cooling. All this is mounted on top of our favorite Open Benchtable. 

The cost of the components should be around $4,910

  • Intel Core i9-12900K processor: $600
  • EK-Quantum Magnitude: $230
  • EK-Quantum P360 water cooling kit: $550
  • ASUS ROG Maximus Z690 Extreme motherboard: $1,100
  • 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 V2: $200

Intel UHD Graphics 770: Benchmark Software

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

As you can see the benchmarks are quite different from what we usually use. I already covered how I use AI Benchmark and Handbrake in SkatterBencher #28 where I overclocked the UHD Graphics 750. I use the same methodology for this guide.

I also included 2 new benchmarks: Unigine Superposition and Spaceship.

Unigine Superposition

Established in 2005 and headquartered in Tomsk, Russia, UNIGINE is a global company focused on real-time 3D technologies. Their proprietary UNIGINE Engine has received worldwide acclaim for pushing technology frontiers further than ever imagined. You may be familiar with some of Unigine’s performance benchmarks like Sanctuary (2007), Tropics (2008), Heaven (2009), Valley (2013), and of course Superposition (2017).

Superposition is based on the UNIGINE2 engine and details the story of a lone professor performing dangerous experiments in an abandoned classroom, day in and day out. Obsessed with inventions and discoveries beyond the wildest dreams, he strives to prove his ideas.

The benchmark outputs a score which can be compared with overclockers around the world on the official Superposition leaderboards.

Spaceship

I found the Spaceship – Visual Effect Graph Demo benchmark while browsing the Steam library.

Spaceship is a short technical demo showcasing high-end visual effects of the Unity game engine. You can both play a 5-minute long first-person narrative sequence or run through an automated simulation as a benchmark. The benchmark reports simple results but is sufficient for us to track the performance improvements after overclocking.

Spaceship is an open-source project so if you are interested on how the project was made, you can also download the full Unity project to learn, mod, reuse, or explore the content. It is available as open-source directly on GitHub.

Intel UHD Graphics 770: Stock Performance

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

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

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

Then save and exit the BIOS.

Here is the benchmark performance at stock:

  • Geekbench 5 OpenCL: 9,641 points
  • Geekbench 5 Vulkan: 9,537 points
  • Furmark 1080P: 925 points
  • AI Benchmark: 1,352 points
  • 3DMark Night Raid: 13,441 marks
  • Unigine Superposition: 5,860 points
  • Spaceship: 14.1 fps
  • CS:GO FPS Bench: 52.79 fps
  • Final Fantasy XV: 14.55 fps
  • Handbrake: 882.277 fps
uhd graphics 770 stock benchmark performance

When running Furmark GPU Stress Test, the average GPU Slice clock is 1550 MHz and GPU Unslice clock is 1350 MHz with 1.081 volts. The average memory clock is 2401 MHz. The average GPU temperature is 37 degrees Celsius, the average GPU power is 14.47 watts, and the average water temperature is 27 degrees Celsius.

uhd graphics 770 stock furmark gpu stress test

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 Clear CMOS button

  • The Safe Boot button temporarily applies safe settings to the BIOS while retaining the overclocked settings, allowing you to modify the settings causing a boot failure
  • The ReTry button forces the system to reboot in case it locks up during the boot process where the Reset button is rendered useless. It will not change anything to your BIOS settings.

Both Safe Boot and ReTry buttons sit right next to each other at the bottom of your motherboard.

Pressing the Clear CMOS button will reset all your BIOS settings to default. This is useful in case 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: MCE + XMP

In our first overclocking strategy we simply take advantage of ASUS MCE and Intel XMP 3.0 technologies.

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. On Z690 motherboards, ASUS has enabled this option by default.

A major difference with the performance impact of MCE on Alder Lake versus the previous Intel generations is that Intel by default is very generous with the maximum power for Alder Lake.

In previous generations, the sustained power limit was always set to the specified TDP. So that’s 125W for K-SKU CPUs. MCE would enable the power limit to run at maximum, so by enabling it you’d get 280W to 300W of sustained power consumption with the Core i9 (when you have sufficient cooling of course).

For Alder Lake K CPUs the sustained power limit has been set higher than TDP. For the 12900K it’s 241W, for the 12700K it’s 190W, and for the 12600K, it’s 150W.

Unlocking the Alder Lake Turbo Boost power with MCE will therefore yield less of a performance increase over default compared to previous platforms. That said, considering the integrated graphics doesn’t use more than 15W at default, enabling MCE was never going to make a huge impact.

Intel Extreme Memory Profile

Intel Extreme Memory Profile, or XMP, is an Intel technology that lets you automatically overclock the system memory to improve system performance. It is an extension to the standard JEDEC specification that allows a memory vendor to program different settings onto the memory stick. The settings include the memory frequency, the memory timings as well as the memory voltage.

There are two types of XMP certification:

  • XMP ready: the module was programmed with an uncertain, but stable, profile
  • XMP Certified: the module was programmed with settings that have passed supplier tests for the CPU and motherboard.

You can find the list of XMP Certified products on Intel’s website.

Intel XMP 1.0 was developed for DDR3 and was later superseded by XMP 2.0 for DDR4. With the launch of DDR5 memory, there is now an XMP 3.0 too. As Alder Lake CPUs supports both DDR4 and DDR5, it also supports Intel XMP 2.0 and Intel XMP 3.0 technology.

With Alder Lake it’s the first time we’re running an integrated graphics with DDR5. DDR5’s main performance benefit is vastly improved memory bandwidth and higher operating frequencies. Both should help us achieve better performance than with DDR4.

I almost always run XMP on my systems because it is an incredibly easy and safe way to improve system performance. Do note that some motherboards may adjust CPU memory controller voltages to support extremely high-frequency memory.

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.

  • Geekbench 5 OpenCL: 9,723 points
  • Geekbench 5 Vulkan: 9,718 points
  • Furmark 1080P: 925 points
  • AI Benchmark: 1,361 points
  • 3DMark Night Raid: 13,571 marks
  • Unigine Superposition: 5,974 points
  • Spaceship: 14.2 fps
  • CS:GO FPS Bench: 53.39 fps
  • Final Fantasy XV: 14.74 fps
  • Handbrake: 909.666 fps
uhd graphics 770 mce and xmp enabled benchmark performance

Enabling XMP and unlocking the Turbo Boost 2.0 limits has a very limited impact on our benchmark performance. We see a performance increase up to 3% in Handbrake, but on average we only improve by a little over 1%.

When running Furmark GPU Stress Test, the average GPU Slice clock is 1550 MHz and GPU Unslice clock is 1350 MHz with 1.079 volts. The average memory clock is 3100 MHz. The average GPU temperature is 37 degrees Celsius, the average GPU power is 14.47 watts, and the average water temperature is 27.3 degrees Celsius.

uhd graphics 770 mce and xmp enabled furmark gpu stress test

OC Strategy #2: MCE + XMP + GPU 2100

In our second overclocking strategy we finally get around overclocking the integrated graphics. As it turns out, it’s really quite simple and straight-forward. And kinda weird.

As I mentioned earlier in the video: the maximum supported Graphics Ratio is 42. We can set this ratio without increasing the voltage and max out the integrated graphics. Overclocking has never been this easy!

Do note that this only increases the Slice frequency and not the Unslice frequency.

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 Max. CPU Graphics Ratio to 42

Then save and exit the BIOS.

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

  • Geekbench 5 OpenCL: 12,782 points
  • Geekbench 5 Vulkan: 12,983 points
  • Furmark 1080P: 1,246 points
  • AI Benchmark: 1,825 points
  • 3DMark Night Raid: 17,211 marks
  • Unigine Superposition: 7,692 points
  • Spaceship: 18.7 fps
  • CS:GO FPS Bench: 71.13 fps
  • Final Fantasy XV: 18.64 fps
  • Handbrake: 911.320 fps
uhd graphics 2100 mhz benchmark performance

Obviously increasing the frequency of the UHD Graphics 770 from 1550 MHz to 2.1 GHz is going to improve the system performance. We see a performance increase of up to 36.13% in Geekbench 5 Vulkan.

The QuickSync performance as measured by Handbrake remains at about 3%. That’s because unlike with Rocket Lake, the Slice and Unslice frequency is decoupled. By setting the Graphics Ratio we only overclock the Slice. Since the media encoding functions like QuickSync are part of the Unslice, the performance is not impacted.

When running Furmark GPU Stress Test, the average GPU Slice clock is 2100 MHz and GPU Unslice clock is 1350 MHz with 1.088 volts. The average memory clock is 3100 MHz. The average GPU temperature is 41 degrees Celsius, the average GPU power is 19.83 watts, and the average water temperature is 28 degrees Celsius.

uhd graphics 2100 mhz furmark stress test

OC Strategy #3: MCE + XMP + GPU 2255 + BCLK 110

I our third overclocking strategy we finally get to break the limits as we need to use base clock frequency overclocking to improve our overclock. We will increase the base clock frequency to 110 MHz and combined with a Graphics Ratio of 41 this results in a GPU frequency of 2255 MHz.

Other than using the base clock frequency, we also increase the VccGT voltage using adaptive voltage mode. Furthermore, we also need to touch a little on BCLK Aware Adaptive Voltage and By Core Usage CPU Ratios.

BCLK Frequency

In terms of the clocking topology, Alder Lake inherits the CPU internal clock generator from Tiger Lake. The standard Alder Lake platform has a 38.4MHz crystal as a reference clock to the PCH. The PCH then generates 3 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. This is different from Rocket Lake where the PCH PLL would generate the 100MHz base clock frequency with no interference from the CPU.

The main reason for including the internal clock generator is to provide a cost-saving opportunity for low-cost systems or platforms.

However, just like Rocket Lake, on high-end desktop motherboards, you’ll still find an external clock generator feeding the 100MHz BCLK frequency to the CPU. Whichever way you get the 100 MHz BCLK, this base clock frequency is multiplied with specific ratios for each of the different parts in the CPU.

The GT frequency or graphics frequency is based on the same 100MHz BCLK but is first divided by 2, and then multiplied with the GT ratio. As mentioned before, the Slice and Unslice are decoupled meaning they run at different ratios. However, both still use the 100MHz divided by 2 base clock frequency.

Alder Lake clock topology

As you now understand, increasing the base clock frequency will impact a lot of parts inside our CPU. So apart from adjusting the Graphics Ratio we also need to adjust the default CPU ratios, Ring ratio, and DDR5 memory ratio.

By Core Usage CPU Ratio

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

Sync All Cores sets 1 ratio that is applied to all cores. This is very much the traditional way of overclocking.

By Core Usage 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. This is also how Intel configures its CPU frequencies according to the Turbo Boost 2.0, Turbo Boost Max 3.0, and other turbo technologies.

Since we are planning to increase the base clock frequency by 10%, we will adjust the default Turbo Ratio configuration using the By Core Usage CPU Ratio options to mimic, as much as possible, the standard configuration. For example, to achieve the maximum single core Turbo Boost 2.0 frequency of 5200 MHz we will set the ratio to 47X instead of 52X.

VccGT Voltage Rail

From the voltage perspective, Alder Lake resembles Rocket Lake more than Tiger Lake.

Compared to Tiger Lake, Alder Lake transitions away from using FIVR for the Cores, Ring, and integrated graphics. Instead, power gates are used. However, unlike Rocket Lake some parts of the Alder Lake CPU are powered using a FIVR. If we only consider the Alder Lake CPU and disregard the chipset, there are a total of 7 different voltage inputs.

The one relevant for overclocking the UHD Graphics 770 is called the VccGT. This voltage powers the GT or integrated graphics including the Slice, Unslice, and Display block. It supports both override and adaptive voltage mode.

alder lake voltage topology

In Override voltage mode the voltage will remain constant, whereas using Adaptive voltage mode, the graphics voltage will dynamically change depending on the workload. When overclocking and using adaptive mode, the CPU will interpolate the required voltage for a given frequency between the base frequency and our set maximum frequency.

Normally you can set the adaptive voltage directly, which maps against the configured boost frequency, or you can use a global adaptive voltage, which offsets the entire voltage-frequency curve. For example:

If our base frequency is 300 MHz at 0.9V and our maximum boost frequency is 1550 MHz at 1.05V, then the interpolated voltage for 1000 MHz will be somewhere between 0.9V and 1.05V. If the V/f curve is linear, that would be 0.96V.

If we overclock our system to 2.3GHz and use an adaptive voltage offset of 300mV, the base frequency will be 300 MHz at 1.2V and the maximum boost frequency will be 2300 MHz at 1.35V. The interpolated voltage for 1000 MHz will be somewhere between 1.2V and 1.35V. If the V/f curve is linear, that would be 1.26V.

alder lake adaptive mode igp interpolation

BCLK Aware Adaptive Voltage

A crucial setting when overclocking the BCLK frequency is BCLK Aware Adaptive Voltage. This option has been on Intel CPUs since Kaby Lake back in 2016.

To understand its function, you can refer to the Adaptive Voltage Mode explanation from my Alder Lake launch article. Fundamental to understand about Adaptive Voltage mode the factory-fused voltage-frequency curve of a CPU core maps a voltage against a ratio. For example, 45X may be mapped against 1.175V for and 52X may be mapped against 1.35V. When the CPU boosts to a high frequency, it references the voltage-frequency curve using the configured ratio to know which voltage to apply.

alder lake voltage frequency curve adaptive voltage mode

When you overclock using the BCLK frequency the voltage may not be suitable for the resulting frequency.

In this specific case, we have increased the BCLK frequency to 110 MHz and have decreased the CPU ratios to match the default specification. Without enabling BCLK aware adaptive voltage, the CPU would look up the voltage for the 47X ratio, in this case, 1.212, then use this for 5.2 GHz. This would obviously not work out well.

alder lake bclk aware adaptive voltage disabled

By enabling this setting, the CPU takes account of the adjusted base clock frequency, and the CPU will use the appropriate voltage based on the effective frequency as opposed to the configured ratio.

alder lake bclk aware adaptive voltage enabled

BCLK Aware Adaptive Voltage works for all V/f power domains in the CPU and produces a voltage based on frequency as opposed to the multiplier. The V/f power domains include the CPU cores, the Ring, and the integrated graphics.

Upon entering the BIOS

  • Go to the Extreme Tweaker menu
  • Set Ai Overclock Tuner to XMP II
  • Set BCLK Frequency to 110
  • Set ASUS MultiCore Enhancement to Enabled – Remove All Limits
  • Set DRAM Frequency to DDR5-6160MHz
  • Set Performance Core Ratio to By Core Usage
    • Set 1-Core Ratio Limit to 47
    • Set 2-Core Ratio Limit to 46
    • Set 3-Core Ratio Limit to 45
    • Set 4-Core Ratio Limit to 45
    • Set 5-Core Ratio Limit to 44
    • Set 6-Core Ratio Limit to 44
    • Set 7-Core Ratio Limit to 44
    • Set 8-Core Ratio Limit to 44
  • Set Efficient Core Ratio to By Core Usage
    • Set Efficient 1-Core Ratio Limit to 35
    • Set Efficient 2-Core Ratio Limit to 35
    • Set Efficient 3-Core Ratio Limit to 35
    • Set Efficient 4-Core Ratio Limit to 35
    • Set Efficient 5-Core Ratio Limit to 34
    • Set Efficient 6-Core Ratio Limit to 34
    • Set Efficient 7-Core Ratio Limit to 34
    • Set Efficient 8-Core Ratio Limit to 34
  • Set Max. CPU Cache Ratio to 34
  • Set Max. CPU Graphics Ratio to 41
  • Set BCLK Aware Adaptive Voltage to Enabled
  • Set CPU Graphics Voltage to Offset Mode
    • Set Offset Mode Sign to +
    • Set CPU Graphics Voltage Offset to 0.15

Then save and exit the BIOS.

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

  • Geekbench 5 OpenCL: 13,646 points
  • Geekbench 5 Vulkan: 13,683 points
  • Furmark 1080P: 1,338 points
  • AI Benchmark: 1,949 points
  • 3DMark Night Raid: 18,065 marks
  • Unigine Superposition: 8,093 points
  • Spaceship: 19.9 fps
  • CS:GO FPS Bench: 75.43 fps
  • Final Fantasy XV: 19.69 fps
  • Handbrake: 974.459 fps
uhd graphics 770 manual overclock benchmark performance

Thanks to the increase in base clock frequency, the integrated graphics now runs at 2255 MHz. This provides a significant performance increase over our previous overclocking strategy. We see a performance increase of up to 44.65% in Furmark 1080P.

The most interesting performance increase we see in Handbrake. The base clock frequency directly impacts a lot of parts inside the CPU, including the Unslice portion of the integrated graphics. Thanks to the 10% increase in base clock frequency, the Unslice frequency increased from 1350 MHz to 1485 MHz. This also results in a substantial increase in QuickSync performance as measured by Handbrake.

When running Furmark GPU Stress Test, the average GPU Slice clock is 2255 MHz and GPU Unslice clock is 1485 MHz with 1.211 volts. The average memory clock is 3080 MHz. The average GPU temperature is 48 degrees Celsius, the average GPU power is 24.63 watts, and the average water temperature is 28 degrees Celsius.

uhd graphics 770 manual overclock furmark gpu stress test

OC Strategy #4: MCE + XMP + GPU 2380 + BCLK 116

In our fourth and final overclocking strategy we try to squeeze the most performance out of this system. We will increase not only the graphics voltage but also overclock the 12900K P-cores and E-cores.

Since this video is specifically about overclocking the UHD Graphics 770, I won’t dig too deep into the P-core and E-core configuration. I just wanted to show that you can combine a CPU overclock with an IGP overclock in one configuration. For the CPU P-core and E-core overclock, I simply copied the settings from my Alder Lake launch article. You can check out that video for more details on the configuration.

By increasing the VccGT voltage to 1.375V I was able to further increase the base clock frequency from 110 MHz to 116 MHz and the associated Slice and Unslice frequencies to 2378 MHz and 1567 MHz.

Upon entering the BIOS

  • Go to the Extreme Tweaker menu
  • Set Ai Overclock Tuner to XMP II
  • Set BCLK Frequency to 116
  • Set ASUS MultiCore Enhancement to Enabled – Remove All Limits
  • Set DRAM Frequency to DDR5-6264MHz
  • Set Performance Core Ratio to By Core Usage
    • Set 1-Core Ratio Limit to 48
    • Set 2-Core Ratio Limit to 48
    • Set 3-Core Ratio Limit to 47
    • Set 4-Core Ratio Limit to 47
    • Set 5-Core Ratio Limit to 46
    • Set 6-Core Ratio Limit to 45
    • Set 7-Core Ratio Limit to 44
    • Set 8-Core Ratio Limit to 44
  • Enter the Specific Performance Core submenu
    • Set Performance Core0 Specific Ratio Limit to 48
    • Set Performance Core1 Specific Ratio Limit to 48
    • Set Performance Core2 Specific Ratio Limit to 47
    • Set Performance Core3 Specific Ratio Limit to 48
    • Set Performance Core4 Specific Ratio Limit to 47
    • Set Performance Core5 Specific Ratio Limit to 48
    • Set Performance Core6 Specific Ratio Limit to 47
    • Set Performance Core7 Specific Ratio Limit to 48
  • Leave the Specific Performance Core submenu
  • Set Efficient Core Ratio to By Core Usage
    • Set Efficient 1-Core Ratio Limit to 35
    • Set Efficient 2-Core Ratio Limit to 35
    • Set Efficient 3-Core Ratio Limit to 35
    • Set Efficient 4-Core Ratio Limit to 35
    • Set Efficient 5-Core Ratio Limit to 34
    • Set Efficient 6-Core Ratio Limit to 34
    • Set Efficient 7-Core Ratio Limit to 34
    • Set Efficient 8-Core Ratio Limit to 34
  • Enter the Specific Efficient Core submenu
    • Set Efficiency Core Group0 Specific Ratio Limit to 36
    • Set Efficiency Core Group1 Specific Ratio Limit to 35
  • Leave the Specific Performance 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 2
  • Leave the AVX Related Controls submenu
  • Enter the AI Features submenu
    • Set Regulate Frequency by above Threshold
  • Leave the AI Features submenu
  • Set Max. CPU Cache Ratio to 34
  • Set Max. CPU Graphics Ratio to 41
  • Set BCLK Aware Adaptive Voltage to Enabled
  • Set CPU Core/Cache Voltage to Adaptive Mode
    • Set Additional Turbo Mode CPU Core Voltage to 1.425
  • Set CPU Graphics Voltage to Offset Mode
    • Set Offset Mode Sign to +
    • Set CPU Graphics Voltage Offset to 0.35

Then save and exit the BIOS.

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

  • Geekbench 5 OpenCL: 14,410 points
  • Geekbench 5 Vulkan: 14,538 points
  • Furmark 1080P: 1,411 points
  • AI Benchmark: 2,046 points
  • 3DMark Night Raid: 18,965 marks
  • Unigine Superposition: 8,413 points
  • Spaceship: 22.7 fps
  • CS:GO FPS Bench: 78.49 fps
  • Final Fantasy XV: 20.48 fps
  • Handbrake: 1,015.663 fps
uhd graphics 770 benchmark performance manual overclock

With the integrated graphics now running 53% higher Slice and 16% higher Unslice clock frequency, at 2378 and 1567 MHz respectively, we see the highest performance in all benchmarks. The performance increase ranges from 15% in Handbrake to 61% in Spaceship.

When running Furmark GPU Stress Test, the average GPU Slice clock is 2378 MHz and GPU Unslice clock is 1566 MHz with 1.375 volts. The average memory clock is 3132 MHz. The average GPU temperature is 60 degrees Celsius, the average GPU power is 31.70 watts, and the average water temperature is 28.9 degrees Celsius.

uhd graphics 770 furmark stress test manual overclock

Intel UHD Graphics 770: Conclusion

Alright, let us wrap this up.

This is the same system I used to prepare my Alder Lake launch content. When I found out about the mega overclocking capabilities of the Alder Lake integrated graphics, I could not wait to once again dive into the topic of IGP overclocking. There are definitely a couple interesting things to note.

First, from the technology perspective turns out I was incorrect in my Alder Lake launch article. Just like on Rocket Lake, there’s no separate ratio control for the Unslice frequency. However, unlike Rocket Lake, the Slice and Unslice are decoupled and are running at different frequencies. Whereas on Rocket Lake you could increase the QuickSync encoding performance by increasing the graphics ratio, on Alder Lake you can only do so by increasing the base clock frequency as the Unslice ratio is fixed to 27X (so 1350 MHz for the Unslice frequency).

Second, the impact of enabling XMP and increasing the DDR5 memory frequency from DDR5-4800 to DDR5-6200 was not as much as expected. Our largest performance improvement was a mere 3.10% in Handbrake and only up to 1.95% in a 3D workload like Superposition. That’s much less than the +32% performance improvement we saw with the Rocket Lake UHD Graphics 750 and DDR4 after enabling XMP. I presume DDR5 provides more than sufficient bandwidth at DDR5-4800 already.

Third, even though the Alder Lake integrated graphics are architecturally the same as Rocket Lake, the overclocking capabilities have vastly improved. On Rocket Lake’s 14nm++ process we achieved only 1.75 GHz whereas with Alder Lake’s Intel 7 process technology we achieved almost 2.4 GHz. That’s an improvement of 650 MHz!

Fourth, the performance improvement after overclocking is equally spectacular. On average we see almost 50% improvement with a standout performance of 61% improvement in the Spaceship benchmark.

Overall, I had a lot of fun playing with the Alder Lake integrated graphics and I will definitely return to IGP overclocking with future platforms.

Anyway, that’s all for today!

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!

3 thoughts on “SkatterBencher #33: Intel UHD Graphics 770 Overclocked to 2378 MHz

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

    […] intricate details I 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 […]

  2. 459@HWBot

    I can definitely verify that your method works like a charm. With less than one hours’ worth of fiddling I was able to run Time Spy result of 1350 (1162 graphics) with 2281 MHz GPU clocks and DDR4-4004 speeds on 1:1 divider. That earned top spots on both 3DMark’s and HWBOT’s list at the time of running the benchmark on measly Noctua air cooling before putting any major effort on figuring out the max clocks for the GPU and optimal memory clock-timing combo. The performance isn’t even that far off from Raven Ridge Vega 8. It looks like that with decent memory subsystem the performance is completely GPU bound in both my and SkatterBencher’s scenario so anyone trying to max out the IGP could as well hunt for the last MHz for the GPU itself.

    1. Pieter

      Very nice, congratulations!

Leave a Reply