ASUS TUF GAMING Z890-PLUS WiFi motherboard review: getting to know Intel Core Ultra 5 245K

The announcement stage is over, and today the sale of Intel Core Ultra 200 series desktop processors and motherboards for the new LGA1851 platform officially begins. We are reviewing ASUS TUF GAMING Z890-PLUS WiFi, a full-size model based on the top-end Intel Z890 chipset. To test its capabilities, we will use a potential hit – Core Ultra 5 245K, and for additional intrigue, we will compare the results with those for Ryzen 7 9700X.

Package contents

ASUS TUF GAMING Z890-PLUS WiFi is offered in a medium-sized cardboard box. It comes with a short instruction manual for assembling the system. A detailed description of features and capabilities is available on the manufacturer’s website – a typical recent development that saves paper.

Also included with the board are two SATA interface cables, an additional rubber mount for M.2 drives, a booster antenna for the wireless module, and a set of branded stickers.

Once again the board comes to us for review in an “extended configuration” – as a part of an already assembled system from Artline. More on the PC configuration later.

Design and layout

ASUS TUF GAMING Z890-PLUS WiFi is made in the ATX form factor with classic dimensions for this standard – 305×244 mm. The exterior design of the board stylistically corresponds to the design of related models of the TUF series – strict, minimalistic and recognizable. Of course, the PCB, coolers, and most connectors are black. Only some of the connectors have a gray base, so the designers are following the TUF concept here.

ASUS TUF GAMING Z890-PLUS WiFi is based on the Intel Z890 chipset. At the initial stage of the launch of the new LGA1851 platform, Intel offers only top-of-the-line hub chips, so there are no alternatives. Accordingly, the board allows the use of any processors for the LGA1851 socket. The starting range of CPUs is traditionally not large, but it consists of powerful enthusiast models with “K/KF” indices in their names.

The processor power stabilization subsystem has a 20-phase circuit (16+1+2+1) and includes DrMOS power assemblies with an operating current of up to 80 A. Despite the fact that the Core Ultra 200 line chips have potentially lower power consumption than previous generations, the top-end platform should provide a certain power reserve for further experiments and be ready for the next generation of CPUs. Therefore, additional capacity will definitely not be superfluous.

For VRM cooling, rather large radiator blocks are provided. Moreover, the aluminum profile that covers the elements along the interface panel is additionally enlarged, so it also serves as a decorative cover for this area. Larger dimensions – larger dissipation area – lower operating temperatures of the assemblies. This is how it works.

The additional power is connected via two 8-pin EPS12V connectors. The manufacturer emphasizes that the group has solid metal contacts (ProCool Power Connectors).

ASUS TUF GAMING Z890-PLUS WiFi has four connectors for memory modules. The Core Ultra 200K/KF processors received DDR5-6400 support in the basic version. Therefore, one should have expected that the boards for such CPUs would allow a little more than in the case of CPUs of previous generations. Indeed, ASUS TUF GAMING Z890-PLUS WiFi claims support for DDR5-9066+ kits, which obviously means CUDIMMs with their own clock. At the time of writing, the manufacturer’s website does not yet have a list of validated kits.

Four slots allow the use of 48GB modules, so the total memory capacity can be up to 192GB. The connectors themselves have a one-way DIMM locking mechanism for easy installation in a pre-assembled system.

The layout of the ASUS TUF GAMING Z890-PLUS WiFi slots for additional expansion cards goes somewhat against recent trends. Despite the fact that their number is usually reduced, the board offers five PCI Express slots at once. The main PCI-E 5.0 x16 for connecting a video card always operates in x16 mode, using PCI Express 5.0 processor lanes.

Additional metal “armor” to increase mechanical strength is already a common occurrence for connectors in this position. However, an additional lever mechanism (PCIe Slot Q-Release) with a plastic bracket that allows for easier release of the graphics adapter from the slot has only recently been introduced.

The rest of the slots are served by the Intel Z890 chipset. This includes another full-size PCI-E x16 4.0 slot that works in x4 mode, two compact PCI-E 4.0 x1 slots, and another PCI-E 4.0 x4 slot that has a corresponding chassis. In a classic gaming PC configuration with a separate graphics card, some of the slots will be blocked directly by the graphics adapter cooler. In this case, PCI-E x1 will be the first to be sacrificed. If the video card has a triple-blade design, then PCI-E 4.0 x4 and PCI-E x16 4.0 (x4) will still be available.

Despite the fact that there are so many expansion slots on the board, the developers managed to find space on the PCB to accommodate four M.2 storage bays. Above the main PCI-E x16 slot is the M.2_1 port, which is serviced by the processor and always operates in PCI-E 5.0 x4 mode. This eliminates the need to split the CPU’s high-speed links when connecting an SSD. The available number of PCI Express lanes in the new Intel chips allows you to simultaneously get PCI-E 5.0 x16 for a video card and M.2 PCI-E 5.0 x4 for a drive.

In addition, the M.2_1 port is equipped with a large cooler with a complex profile, which is very appropriate, given the power requirements and heating of the highest speed M.2 PCI-E 5.0.

The second port (M.2_2) is located opposite the PCI-E x1 slots. It is also serviced by the processor, but it is designed for M.2 PCI-E 4.0 x4 mode. This is a drive connector that does not require additional cooling or is equipped with its own radiator block.

Two more M.2 ports are hidden behind a solid aluminum plate that serves as a common cooler for the respective SSDs. This pair of connectors is already on the chipset’s “balance sheet”. M.2_3 corresponds to PCI-E 4.0 x4 and is notable for the fact that it allows the use of both standard 2280 and extended format models – 22110. In turn, M.2_4 is distinguished by the ability to use SATA SSDs in addition to PCI-E x4 drives.

Therefore, ASUS TUF GAMING Z890-PLUS WiFi offers a wide variety of options for connecting drives. As for SATA devices, the board has four 6Gb/s SATA ports.

Note that there are no mutually exclusive combinations in this case, so all drives can be connected simultaneously, and this does not affect the capabilities of the available PCI-E expansion slots.

The board has seven 4-pin connectors (PWM/DC) for customizing the cooling system. In addition, the proprietary Q-Fan technology allows you to fine-tune fan algorithms based on temperature sensor readings.

There is only a tiny “pixelated” TUF logo in the upper right corner. For the rest, there are three connectors for connecting ARGB Gen 2 addressable ribbons. In the same area, we also note the presence of Q-LED startup diagnostic indicators, which this time have been moved to the upper edge of the PCB. Prolonged illumination of the light at the start is an indicator of a problem with a particular subsystem.

ASUS TUF GAMING Z890-PLUS WiFi offers an internal USB Type-C port for the corresponding port on the case panel. It is USB 3.2 Gen 2×2 with a bandwidth of up to 20 GB/s, but this time without additional amplification with Power Delivery support.

The Realtek ALC1220P codec and several specialized capacitors are used for the sound subsystem, which is isolated from the general PCB array. DTS Audio support is also implemented. A similar scheme performed well during testing of ASUS TUF GAMING X870-PLUS WIFI. Subjectively, the overall impressions of sound quality are identical.

The sound is quite typical for modern built-in audio: the sound is detailed, clear and bright. Depending on your personal preferences, if you like a softer and more “relaxed” sound, some may find it a little harsh here, with too much bass – but this can be easily corrected with the help of the equalizer in the DTS utility. Here you can also choose one of the sound templates if you are not satisfied with the standard sound.

The board’s networking capabilities are provided by an Intel 2.5G controller for wired networks and a Wi-Fi 7 module (MediaTek Wi-Fi 7 MT7925) with a maximum bandwidth of up to 2.9 GB/s.

The interface panel of the board, as expected for models of this class, is immediately equipped with a metal cap with informative port labels. As for the latter, the highlight here is Thunderbolt 4 (Type-C) with a bandwidth of 40 GB/s, which will be a mandatory companion of at least Intel Z890 chipset motherboards. Processors for LGA1851 have a built-in Thunderbolt 4 controller, so only the appropriate wiring is required for implementation. The second Type-C port on the panel complies with the USB 3.2 Gen2×2 (20Gb/s) standard and supports DP Alt Mode for image output.

We also see three USB 3.2 Gen2 (10 GB/s) and USB 3.2 Gen1 (5 GB/s) ports in the classic Type-A format. Full-size HDMI 2.1 and DisplayPort 1.4 are provided for connecting screens, and an Ethernet socket allows you to connect to a wired network.

ASUS TUF GAMING X870-PLUS WIFI supports BIOS Flashback firmware update technology, the corresponding activation button is also located on the panel. To connect the amplifying antenna of the Wi-Fi module, Q-Antenna connectors with threadless mounting are used. The board also offers five audio jacks for speaker switching and has an optical S/PDIF digital output.

PC configuration

• Motherboard: ASUS TUF GAMING Z890-PLUS WIFI (Intel Z890, ATX)
• Processor: Intel Core Ultra 5 245K (6P+8E; 4.2/5.2 GHz + 3.6/4.6 GHz)
• Cooling: ASUS TUF Gaming LC II 360 ARGB
• Memory: Kingston Fury Beast DDR5-6400 RGB 64 GB (KF564C32BBAK2-64)
• Video card: ASUS TUF GAMING GeForce RTX 4070 Ti SUPER 16 GB (TUF-RTX4070TIS-O16G-GAMING)
• Storage devices: Kingston FURY Renegade 1 TB (SFYRS/1000G), Kingston NV2 1 TB (SNV2S/1000G)
• Body: ASUS TUF GAMING GT502 PLUS

A Core Ultra 5 245K processor was used to evaluate the board’s capabilities. It is the youngest model in the galaxy of enthusiast chips for LGA1851.

The processor has a 14-core configuration – 6 productive (P-Cores) with Lion Cove architecture and 8 energy-efficient (E-Cores) based on Skymont.

Productive cores have a frequency formula of 4.2/5.2 GHz, while the operating range of energy-efficient cores is slightly lower – 3.6/4.6 GHz. Each P-Core is equipped with 3 MB of L2 cache, while the energy-efficient ones have 4 MB per 4-core cluster. Thus, the total capacity of the second level cache is 26 MB. The chip also offers a total L3 of 24 MB.

As a reminder, for Core Ultra 200 chips, the manufacturer has abandoned support for Hyper-Threading technology, relying on architectural improvements to the cores to compensate for the lack of HT in multithreaded tasks.

The Core Ultra 5 245K processor has a Processor Base Power (PBP) of 125W, while the Maximum Turbo Power (MTP) is set at 159W.

To cool the Core Ultra 5 245K, we used the ASUS TUF Gaming LC II 360 ARGB 3-fan liquid system, which has already proven itself well during the Ryzen 7 9700X tests. Despite the fact that Arrow Lake-S chips claim a significant reduction in power consumption, we are still dealing with a chip from the enthusiast family, so the cooling should be appropriate.

According to the specification, Core Ultra 200-S chips support DDR5-6400 RAM. Therefore, it is logical that for LGA1851 PCs, kits with these modes should be the starting ones. So, the Kingston Fury Beast DDR5-6400 RGB 64GB dual-channel kit (KF564C32BBAK2-64) is quite appropriate here.

GeForce RTX 4070 Ti SUPER 16 GB graphics cards are a great start for advanced gaming configurations, so the choice of ASUS TUF GAMING GeForce RTX 4070 Ti SUPER 16 GB (TUF-RTX4070TIS-O16G-GAMING) was predictable. In addition, this model was used during the Ryzen 7 9700X tests, so it will allow for a more correct comparison of the results.

A pair of Kingston drives was used to install the OS, applications, and test games. In general, the system unit was assembled in the ASUS TUF GAMING GT502 PLUS chassis. It is a practical big guy with a panoramic view of the interior, which was pleasantly pleasing to the eye during the first acquaintance with this case.

Core Ultra 5 245K vs. Ryzen 7 9700X

The best addition to the motherboard review will be a small local confrontation between the direct opponents – Core Ultra 5 245K and Ryzen 7 9700X. Such a comparison will give you a first idea of the capabilities and allow you to evaluate the potential of Intel’s new generation chips.

Judging by the technical specifications alone, it should be an interesting “version”. To balance the odds, we are using Ryzen 7 9700X exclusively in the 105W TDP mode. The processor was initially designed with this level of power consumption in mind, and the process of switching from 65W to 105W is now as simple as possible and usually requires only one mouse click in the BIOS.

Before starting the tests, we suggest that you familiarize yourself with the performance of the RAM subsystem. In both cases, we used DDR5-6400 kits with the same timings – 32-39-39-80.

The new Intel chips have a very efficient memory controller, offering 100 GB/s transfers already in DDR5-6400 mode. The performance of the Ryzen 7 9700X is much more modest, especially in terms of read and copy speeds, which are the weaknesses of single-CCD models. However, these are only selected indicators that we shouldn’t take into account, but we don’t draw conclusions. Architecture features can compensate for certain omissions. In addition, the Ryzen 7 9700X platform has the best overall memory latency – 74.4 ns versus 87.9 ns.

So, let’s get to the benchmarks. First, let’s get acquainted with the results of Core Ultra 5 245K, which are presented in the screenshots.

It should also be noted that with the default BIOS settings, under multi-threaded workloads, productive cores accelerated and ran at 5000 MHz for a long time, while energy-efficient E-Cores ran at 4600 MHz. The supply voltage is 1.13 V.

The processor test stages of 3DMark favor the Core Ultra 5, and it is quite significant. On the CPU Profile stage with the maximum number of threads, it is about 25%. In the Time Spy computational stages, the new Intel chip received 10% more points.

Synthetics with CPU-Z again returns to the situation with a quarter advantage. Obviously, in certain situations, the 6P+8E combination with 14 physical fins is better than 8/16. Intel has managed to compensate for the lack of Hyper-Threading primarily due to the tangible acceleration of E-Cores. However, we note that in single-threaded mode, the Ryzen 7 9700X has a slight advantage.

In Cinebench tests with scene rendering, the Intel 14-core also proves to be faster. While the advantage in R23 is 11%, in the new Cinebench 2024 version it increases to 18%.

In Geekbench 6, the Core Ultra 5 245K processor is 7% faster than its opponent in multi-threaded mode, but in single-threaded mode, the advantage is again for the Ryzen 7 9700X. And this time, the difference is already a tangible 11%.

In Blender, Intel’s new product proved to be 6% more productive.

But this balance of power is not a dogma. This was demonstrated by the V-Ray 6 test, where the Ryzen 7 9700X proved to be 2% better, and in the Z-Zip archiver, the advantage of the AMD chip was already 7%. It would seem that tasks where the number of cores is of fundamental importance, but architectural features are sometimes more important.

Performance in games

There is a lot of interest in the ratio of processor power in games. Although we don’t expect a significant difference in performance, let’s look at the actual results.

As in the case of applied tasks, the situation here is not clear-cut. In certain projects, the Core Ultra 5 245K platform has an advantage, while in others, the Ryzen 7 9700X system has an advantage. Although the overall difference in performance is usually not that significant, even in Full HD mode, it is still there.

In 1440p mode, the ratio of forces is generally maintained, although the difference in performance is further reduced due to the increased dependence on the capabilities of the video card.

Integrated graphics

Arrow Lake-S processors have integrated Intel Graphics with Xe-LPG architecture. In the case of Core Ultra 5 245K, 4 clusters of Xe-core computers are used, operating in the frequency range of 300-1900 MHz. Ryzen 9000 chips also have an integrated video core, but it’s a rather modest Radeon Graphics with two compute clusters on RDNA2. A comparison of the capabilities of both solutions is shown in the diagrams, and it is quite eloquent.

Core Ultra 5 245K has a much more productive video core than the Ryzen 7 9700X. The advantage in synthetic benchmarks is three times, in real games – 2-2.5 times.

Energy consumption and heating

The energy confrontation between the two chips is very interesting. Improving efficiency was almost the main task of Intel’s developers. Exorbitant power demands of the previous generation of chips, especially during high multi-threaded workloads, were an important factor and a reason why competitors’ solutions were often chosen.

There is indeed a significant improvement here. The Core Ultra 5 245K consumed up to 144 watts of power at its peak during Cinebench R23 rendering. If we turn to the Ryzen 7 9700X, then using the TDP mode of 105 W, the actual level of CPU power consumption under identical conditions was almost similar – 140-142 W.

Recall that the Core i5-14600K consumed ~180 watts in the drain on such tasks, and this was still considered a decent figure compared to what the top Core i9s (350+W) can afford.

So, with the Core Ultra 200, we have a significant progress in terms of energy efficiency. At least this is certainly the case with the tested Core Ultra 5 245K. Reducing power consumption also reduces the requirements for the efficiency of the cooling system, although you shouldn’t relax here. ASUS TUF Gaming LC II 360 ARGB kept the chip temperature under load at 65-70C, but this is the normal operating mode. There is still overclocking. In addition, it should be borne in mind that Arrow Lake-S chips have a slightly shifted hottest point, which is now located above the central zone.

We should also note that the power subsystem of the ASUS TUF GAMING Z890-PLUS WIFI board coped with the Core Ultra 5 245K load perfectly. VRM elements warmed up only to 36C. Of course, there is some merit in the good ventilation inside the system unit.

Overclocking Core Ultra 5 245K

Overclocking processors is an art that requires a serious immersion in the topic, painstaking study of the “stuff” and a lot of time for experiments. With a superficial review of the possibilities, we tried to accelerate Core Ultra 5 245K in express mode in the easiest way. Given the availability of unlocked multipliers for both types of processor cores, we achieved certain results by gradually increasing the coefficient.

Without an additional increase in the supply voltage, the processor remained stable when the operating frequency of productive cores was increased to 5200 MHz, and energy-efficient cores – to 4800 MHz. That is, in fact, for both types of cores we managed to get an additional +200 MHz under maximum load. Not as much as one might expect, given that TSMC’s 3-nanometer crystal (N3B) is used for CPUs.

However, we would like to emphasize that this is the most “lazy” approach that was used due to the lack of time for experiments. In addition, we were dealing with an engineering sample of the processor, which may have some differences from the copies that will be available for retail sale. Therefore, the frequency potential of Core Ultra 5 245K requires additional research.

Even though the increase in operating frequencies was relatively small, the performance gains are recorded and are significantly higher than the measurement error of 4-5%. The Ultra 5 245K Core responds well to the increase in P/E-Core operating frequencies, so it makes sense to explore the potential in more detail, especially for enthusiasts.

One more thing…NPU

All Core Ultra 200-S desktop chips will receive an NPU (Neural Processing Unit) hardware unit, which will potentially provide a significant increase in the efficient performance of machine learning tasks. The main idea of such a module is to reduce the load on the CPU/GPU when executing certain AI algorithms. Processing them on the NPU will not only free up resources of other units but also require less power to complete the task.

Arrow Lake-S received a third-generation NPU with a pair of neural computing engines consisting of Multiply Accumulate Arrays (MAC). Such operations are the basis of typical AI workloads. However, in the presented configuration, the maximum computing power is relatively small – 13 TOPS. However, it is still enough to perform AI tasks on a local system.

To test the capabilities of the NPU, we used the new Geekbench AI benchmark, which allows you to involve a “neural processor” in computing if one is available.

By choosing the OpenVINO AI framework, we obtain results using CPU, GPU, and NPU. As can be seen in the screenshots, the performance of all blocks at the single-precision stage (FP32) is quite close, while GPU and NPU offer 1.5-2 times higher performance than CPU at half-precision (FP16). At the same time, all units are similar in terms of quantization capabilities (INT8).

Thus, an additional NPU can either boost the overall computing power of AI tasks or perform them instead of the CPU/GPU at about the same speed but using less power.

As it turned out, the existing NPU unit can be further accelerated, and ASUS makes this one of the important features of the board, putting the corresponding option in the most prominent place in the A and Tweakers section. The NPU Boost option has three levels. Choose the maximum Level 3 and repeat the measurements in GeekBench AI.

The results for single precision remained almost unchanged, but the results for half and quantization increased by almost a third. And this is not bad. Especially with the proliferation of applications that will use these NPU computer reserves. At the moment, the availability of NPUs looks more like an option for the future, but it may well become a competitive advantage in the future, given that AMD does not offer such modules for desktop Ryzen 9000s.