After a long wait, the AMD Ryzen 9000 series processors, which are based on the new Zen 5 architecture, have finally hit the market. The most pressing question for consumers is whether they are worth purchasing. From AMD's introductions prior to the product launch, it appears that with the new Zen 5 architecture, the Ryzen 9000 series processors have the potential to outperform Intel's 14th generation Core processors in various processor performance, application performance, and gaming performance tests.

However, the Ryzen 9000 series processors still adhere to the core count specifications and working logic of the Zen 4 architecture, meaning that the Zen 5 architecture continues to use a full large-core design. In contrast, Intel's 14th generation Core processors have adopted a combination of Performance-core (P-cores) and Efficient-core (E-cores), and rely on a hardware thread scheduler for task allocation. With the addition of a large number of E-cores, Intel's processors theoretically have a significant lead in core count over the mainstream processors of the Ryzen 9000 series, such as the 14th generation Core i7 processor, which has 12 more E-cores than the Ryzen 7 series. So, in real-world testing, which architecture logic (pure large cores vs. large cores + numerous small cores) performs better? Let's continue to find out.

Advertisement

Significant Enhancement of Single-Core Performance: A Brief Analysis of the Zen 5 Processor Architecture

To defeat processors with more cores using fewer cores, the most critical aspect is obviously to enhance the performance of each core. Therefore, when AMD introduced the Zen 5 microarchitecture, they stated that the architecture aims to achieve four major goals: first, to execute more instructions per cycle, second, to have a wider overall width, third, to double the cache data bandwidth, and fourth, to accelerate AI. To this end, AMD has made the following main improvements to the Zen 5 processor architecture. Since "Microcomputer" has already provided an in-depth analysis of the Zen 5 processor architecture in its August 2024 issue, we will only provide a brief review and summary here.

The focus of the Zen 5 architecture improvements lies in the decoder being changed to two 4-width decoders.

Firstly, on the front end, the focus of Zen 5's improvements is the change to two 4-width decoders, which can achieve 8-width decoding at the same time. Since the Zen architecture began, AMD has used a single 4-width decoder for multiple generations, including the Zen 4 architecture, which also used a 4-width design. In Zen 5, this has finally been thoroughly improved. The corresponding branch prediction capabilities have also been significantly enhanced, including lower latency, higher accuracy, and greater throughput. The instruction cache has also become faster and wider due to the adjustment of the decoder. Overall, Zen 5 now has a larger front end, which brings with it more optimizations and adjustments, especially in branch prediction.

The Zen 5 processor has a wider and larger dispatch and execution window.

The changes on the front end bring in more instructions, so the integer execution part of Zen 5 is larger and wider. The integer part of Zen 5 now has an 8-width dispatch and retirement system, and the scheduling unit has changed from the previous Zen 4's 4 small units to 2 large-scale units. The integer execution part includes 6 ALU BR and 3 ALU Mul multipliers, while the previous generation Zen 4 had 2 ALUs, 2 ALU BR units, and 3 AGUs. In terms of scale, Zen 5 is obviously larger, and Zen 5 has also increased the execution window size.

The L1 cache of each core in the Zen 5 processor has been increased to 80KB, and the transfer bandwidth has been doubled, with enhanced data prefetching capabilities.

The Zen 5 processor has complete AVX-512 support capabilities.In terms of caching and data processing, Zen 5 introduces a new 48KB L1 12-way data cache with a latency of 4 cycles, bringing the total L1 cache capacity per core to 80KB. In contrast, Zen 4 architecture processors have only 32KB of L1 data cache per core, with an 8-way design and a total L1 cache capacity of just 64KB. Additionally, the L1 cache bandwidth in the Zen 5 architecture is doubled, and the floating-point unit bandwidth is also doubled (corresponding to adjustments in the floating-point execution section), enhancing data prefetch performance. In terms of floating-point execution units, AMD has introduced a powerful SIMD 512-bit unit supporting AVX-512, with a queue depth of 384, featuring 6 FADD units with a 2-cycle latency, significantly boosting the overall floating-point instruction execution capability. The enhancement in floating-point capabilities will lead to performance improvements for AMD in AI and gaming computing, as well as in certain scientific computing tasks. Comparatively, while Zen 4 processors can also execute AVX-512, they achieve this by combining two 256-bit SIMD units, whereas Zen 5 has full AVX-512 support capabilities.

Zen 5 desktop processors feature compute cores manufactured using TSMC's 4nm process technology. In terms of production process, desktop processors based on the Zen 5 architecture, represented by the Ryzen 9000 series, use TSMC's 4nm process, with the CCD codenamed "Eldora," containing 8.315 billion transistors and an area of 70.6 square millimeters, resulting in a density of approximately 118 million transistors per square millimeter, a 26.8% increase in transistor density compared to Zen 4 processors. The IOD chip remains consistent with the IOD chip in the previous Zen 4 processors, still using TSMC N6 process technology, with an area of 122 square millimeters, 3.4 billion transistors, and a density of 27.9 million per square millimeter. Functionally, it still provides 24 PCIe 5.0 lanes, allowing users to simultaneously use one PCIe 5.0 x16 graphics card, two PCIe 5.0 x4 SSDs, and supports DDR5 5600 dual-channel memory, with the capability to support high-speed DDR5 8000 memory.

Zen 5 desktop processors still feature integrated RDNA 2 architecture display cores. Additionally, the first batch of Ryzen 9000 series processors also comes with integrated display cores based on the RDNA 2 architecture, meaning that every Zen 5 processor can function normally even without a dedicated graphics card. Of course, the scale of the integrated display core is not large, with only two CU compute units and 128 stream processors, but it supports AV1 hardware decoding, as well as hardware decoding and encoding for H.264 and H.265, and supports 4K@60Hz display, sufficient for general office and video applications. Compared to regular Zen 4 processors, the Ryzen 9000 series processors have added support for USB4 technology, providing up to two USB4 interfaces with a bandwidth of 40Gbps each. However, this requires users to use a 600 series motherboard with a USB4 interface on the board or the X870E, X870 motherboards expected to be available in September this year to enjoy the benefits of USB4 technology.

Activating the PBO (Precision Boost Overdrive) feature on the Ryzen 9000 series desktop processors can bring up to a 15% performance improvement. In terms of processor overclocking, in addition to the advantages of unlocked multipliers, AMD has also optimized the PBO (Precision Boost Overdrive) technology. If users have excellent cooling configurations and prioritize performance, especially multi-threaded performance, then activating the PBO feature on the Ryzen 9000 series desktop processors can increase the processor's operating frequency, leading to a performance improvement of up to 15%, which is very similar to Intel's Turbo Boost feature. AMD mentions that if the processor's default TDP is lower, the performance improvement obtained after activating the PBO feature will be more noticeable.

Curve Shape allows users to fine-tune the processor under multiple frequency and temperature conditions. In addition to the traditional PBO updates, AMD has also introduced an updated version of Curve Optimizer and a new feature called Curve Shaper for overclocking adjustments. Curve Shaper is primarily used to address issues related to power consumption, temperature, and frequency under different states of the processor. Curve Shaper allows users to set 15 combinations of frequency and temperature working modes based on the processor's lowest frequency, low frequency, medium frequency, high frequency, and ultra-high frequency, as well as three different temperature states of low, medium, and high. This enables the processor to reduce frequency, temperature, and power consumption as much as possible during gaming or light daily applications, and also achieve the highest possible frequency with more reasonable voltage in high-performance mode.

Thanks to the improvements in the Zen 5 microarchitecture, the overall IPC (Instructions Per Cycle) has seen an average increase of 16% compared to the previous generation.Based on the aforementioned improvements, AMD has stated that the overall IPC (Instructions Per Clock) increase of the Zen 5 microarchitecture compared to the previous generation is an average of 16%. This sets the stage for the Zen 5 processors to compete with Intel's 14th generation Core processors. Similar to the Ryzen 7000 series, the Ryzen 9000 series processors based on the AMD Zen 5 architecture will initially launch with four models: 6-core, 8-core, 12-core, and 16-core.

There are four Ryzen 9000 series processors based on the Zen 5 architecture at launch:

- The flagship Ryzen 9 9950X sees no significant changes from the Ryzen 9 7950X, maintaining a 16-core, 32-thread design with a top boost frequency of 5.7GHz. The combined L2 and L3 cache remains at 80MB, which is 1MB of L2 cache per core, totaling 16MB of L2 cache.

- The Ryzen 9 9900X, with 12 cores and 24 threads, is very similar in specifications to the Ryzen 9 7900X, including the same top boost frequency of 5.6GHz and a combined L2 and L3 cache capacity of 76MB. The only difference is a significant reduction in the processor's TDP (Thermal Design Power) by 50W, now only 120W.

- The Ryzen 7 9700X, an 8-core, 16-thread processor, continues to use the Socket AM5 package and is equipped with an octopus-shaped IHS (Integrated Heat Spreader) cooling cover.

- The Ryzen 7 9700X has increased its top boost frequency from 5.4GHz of the Ryzen 7 7700X to 5.5GHz, while the TDP has been reduced from 105W to 65W. This highlights the improved energy efficiency of the Zen 5 processors. The combined L2 and L3 cache capacity remains unchanged at 40MB for both processors.

- The Ryzen 5 9600X, configured with 6 cores and 12 threads, also sees a 100MHz increase in its top boost frequency, from 5.3GHz of the Ryzen 5 7600X to 5.4GHz. The TDP has been reduced from 105W to 65W, and the combined L2 and L3 cache is consistent at a total of 38MB. AMD has matched these processors with corresponding competitors; for example, the Ryzen 5 9600X is set to compete with the Core i5-14600K, the Ryzen 7 9700X targets the Core i5-14700K, the Ryzen 9 9900X will mainly compete with the Core i9-14900K, and the Ryzen 9 9950X currently has no direct competitor.

It is worth noting that the release dates for the four Zen 5 processors vary. The lower-end Ryzen 5 9600X and Ryzen 7 9700X will launch first on August 8, 2024, while the higher-end Ryzen 9 9900X and Ryzen 9 9950X will be released later on August 15. The Micro Computer testing lab has received the Ryzen 5 9600X and Ryzen 7 9700X test samples first. In terms of appearance, apart from the different model numbers printed on the front, they are indistinguishable from the previous Zen 4 processors. Both processors still use the Socket AM5 package and feature an octopus-shaped IHS cooling cover with the same number of bottom contacts, 1718. Whether the Ryzen 5 9600X and Ryzen 7 9700X can accomplish the task of defeating their respective competitors will be verified through testing.

Unleashing the full performance of the Zen 5 processors with a high-specification ROG X670E motherboard.Due to the X870 series, B850, and B840 motherboards designed for the Zen 5 processors not being expected to hit the market until September of this year, the existing 600 series motherboards can also perfectly support Zen 5 processors. Therefore, in this test, to maximize the performance of the Ryzen 7 9700X and Ryzen 5 9600X processors, we have specially paired them with a high-specification X670E motherboard: the CROSSHAIR X670E HERO from ROG.

The motherboard features a luxurious 18+2 phase power circuit, coupled with an oversized one-piece I/O+VRM heat sink with built-in heat pipes.

To perfectly support high-end Ryzen processors, the CROSSHAIR X670E HERO motherboard has seen significant upgrades in various aspects. Firstly, in terms of the power circuit, ROG has enhanced the scale of the motherboard's power circuit, increasing from the 14+2 phase design of the CROSSHAIR VIII DARK HERO to an 18+2 phase power circuit. At the same time, each phase of the power circuit has been upgraded from Power Stages MOSFETs supporting 90A loads to MOSFETs supporting 110A loads.

The motherboard's Polymo dynamic lighting display can synchronize its illumination with other hardware that supports AURA SYNC lighting technology.

In terms of thermal design, its heatsink surface is mirror-polished, and both the motherboard chipset heatsink and the ROG "Eye of the Tiger" logo feature a dot-matrix design style. Combined with the motherboard's Polymo dynamic lighting display, it gives the impression of being a dream piece of equipment from a cyberpunk world. Of course, in addition to its stunning appearance, this motherboard's heatsink is also meticulously designed. Its cooling module consists of an oversized one-piece I/O+VRM heat sink and a large area chipset heatsink, which can effectively increase the heat dissipation surface area to achieve rapid cooling. Furthermore, through the ROG water cooling control area and a rich array of fan headers, it achieves comprehensive thermal control.

The DDR5 memory slots of the ROG CROSSHAIR X670E HERO motherboard can support high-speed DDR5 memory with speeds above DDR5 6400.

Considering that since the AMD Zen 4 Ryzen 7000 series processors, AMD has also begun to adopt the new generation of DDR5 memory, and currently, the PMIC (Power Management Integrated Circuit) of many mainstream DDR5 memories limits the operating voltage of the memory, the ROG CROSSHAIR X670E HERO motherboard specially provides an AEMP setting that can unlock the voltage lock of DDR5 memory, allowing DDR5 memory to use higher voltages, thereby overclocking the memory to above DDR5 7000 or higher. Of course, one can also choose to enhance memory performance by reducing latency. It is worth mentioning that the ROG CROSSHAIR X670E HERO also has some killer features to better utilize the performance of Zen 5 processors.

The PBO enhancement option in the motherboard BIOS can further enhance the performance of Zen 5 processors, and we recommend everyone to enable this setting on ASUS or ROG motherboards.

Users can enable the "Dynamic OC Switcher Hybrid Dual-Mode Overclocking" feature to enhance the single-thread and multi-thread performance of the processor during overclocking.

In its BIOS, one can see that its PBO (Precision Boost Overdrive) options include not only "Enabled, Turn On" but also an "Enhancement" (PBO Enhancement) option. The motherboard BIOS also includes an automatic processor performance enhancement feature, and enabling these can fully unleash the performance of Zen 4 and Zen 5 processors, allowing them to run at higher frequencies. At the same time, the motherboard continues the "Dynamic OC Switcher Hybrid Dual-Mode Overclocking" technology from the C8DH motherboard, which can intelligently switch between DOCP+PBO and full-core overclocking modes based on the user's preset current and temperature thresholds. In simple terms, it means activating DOCP+PBO for higher single-core performance during low loads and switching to full-core overclocking mode for stronger multi-thread performance during high loads. Additionally, this motherboard also features AI Smart Optimization 2.0 technology, including AI Smart Overclocking, AI Smart Cooling 2.0, AI Smart Networking, and two-way AI Noise Reduction, among other functions. Among them, AI Smart Overclocking can predict and assess the CPU overclocking potential and system cooling environment, providing tuning suggestions to help ordinary users break through the CPU frequency limits.The motherboard comes with a ROG Hyper M.2 expansion card, allowing the motherboard's PCIe 5.0 M.2 SSD slot count to reach 3.

With the upgrade of AMD processors and the X670 chipset, the expansion capabilities of the ROG CROSSHAIR X670E HERO motherboard have also been greatly enhanced. It features two graphics card slots that support the PCIe 5.0 standard and support x16 or x8+x8 modes, which implies its potential to support the construction of a dual-card SLI parallel system with PCIe 5.0 graphics cards. In terms of storage, it provides up to 5 M.2 SSD slots. Among them, there are two PCIe 5.0 M.2 SSD slots on the motherboard that are provided by the processor with PCIe 5.0 x4 bandwidth, as well as two PCIe 4.0 x4 M.2 SSD slots provided by the X670 chipset.

ROG also includes a ROG Hyper M.2 expansion card with the motherboard, which, when plugged into the second PCIe 5.0 graphics card slot, can expand to provide an M.2 SSD interface with PCIe 5.0 x4 bandwidth (16GB/s). However, it should be noted that since the bandwidth of the CPU's PCIe is divided after plugging into the second PCIe 5.0 graphics card slot, the bandwidth of the first PCIe 5.0 graphics card slot will only be PCIe 5.0 x8, which will have a slight impact on the graphics card performance.

Below the memory slots, the motherboard still provides a "Q-Release" graphics card easy-release button.

In addition, the ROG CROSSHAIR X670E HERO motherboard also has a "Q-Release" graphics card easy-release button, which, when pressed, activates the card latch to move downward and eject the graphics card, making it more convenient for users to remove the graphics card. The Q-Latch convenient latch on the M.2 SSD interface allows users to insert and remove M.2 SSDs more conveniently, simply by rotating the latch to secure or remove the M.2 SSD, eliminating the need for additional installation screws.

The biggest upgrade of the motherboard I/O backplane is the provision of two USB4 interfaces, offering a 40Gbps transfer bandwidth and supporting 100W PD charging.

In other aspects, the ROG CROSSHAIR X670E HERO motherboard is also equipped with the Intel I225-V 2.5G wired network card, and is paired with Intel's latest Wi-Fi 6E AX210+ Bluetooth 5.2 wireless module. Compared to the previous Wi-Fi 6, Wi-Fi 6E adds a 6GHz frequency band, with a frequency range of 5925～7125MHz, offering more channels, greater capacity, and significantly increased throughput. The motherboard backplane provides various latest expansion interfaces, including two USB4 interfaces that can provide a 40Gbps transfer bandwidth, support 100W PD charging, dual 4K video output, and are provided by the Intel JHL8540 controller. When connected to a USB4 mobile SSD, the sequential read speed can exceed 3100MB/s.

The motherboard also comes with a front USB3.2 Gen 2x2 Type-C interface, which supports QC 4.0+ fast charging technology, providing up to 60W of power output.

Of course, various traditional interfaces are also indispensable, such as the motherboard being equipped with an HDMI interface that supports 4K@60Hz display, a front USB3.2 Gen 2x2 Type-C interface, and supports QC 4.0+ fast charging technology, providing up to 60W of power output. In terms of audio, the ROG CROSSHAIR X670E HERO's SupremeFX 7.1 audio system is equipped with the latest Realtek Codec: the Realtek ALC4082 7.1 channel Codec with an output signal-to-noise ratio of 120dB and a recording signal-to-noise ratio of 113dB, and is paired with the "harmonic distortion + noise" of only -114 dB ESS SABRE 9218PQ four-channel DAC decoding chip, Nichicon audio capacitors, and various other high-quality components.

Unleashing the maximum performance of Zen 5, the Kingston FURY Beast DDR5 RGB 6800 32GB kit comes to assist.Despite the nominal memory support speed of the Zen 5 processor being DDR5 5600, considering the enhanced capability of the Zen 5 processor to support high-speed memory, in order to maximize the performance of the Zen 5 processor, we have also used the Kingston FURY Beast DDR5 RGB 6800 32GB memory kit in our testing, which supports AMD EXPO and Intel XMP one-click overclocking technologies. This memory kit, in terms of design style, is similar to the Kingston FURY Beast series, neither too tall nor too short, with a large silver KINGSTON FURY logo embedded on the left side of the memory front, complemented by a distinctive white "BEAST" on the right, giving the memory an appearance that is grand and high-end.

The Kingston FURY Beast DDR5 RGB 6800 32GB kit features an attractive RGB lighting effect. Additionally, the top of this memory is equipped with RGB LEDs and light guides, which can display up to 18 different lighting effect modes, including "Rainbow, Prism, Light Speed, Teleport, Flame," and supports lighting synchronization technologies from four motherboard manufacturers: ASUS AURA SYNC, Gigabyte RGB FUSION, ASRock Polychrome SYNC, and MSI Mystic Light Sync. This allows it to synchronize and glow with other hardware that supports these lighting effects, creating a more spectacular and dazzling visual experience.

The Kingston FURY Beast DDR5 RGB 6800 32GB kit supports AMD EXPO and Intel XMP one-click overclocking technologies. Paired with the Zen 5 Ryzen 7 9700X processor, this memory can easily be overclocked to DDR5 7200, providing higher memory bandwidth and lower memory latency.

The Kingston FURY Beast DDR5 RGB 6800 32GB kit we tested consists of two memory modules, each with a capacity of 16GB. Each module features a single-sided 8-chip design, using SK Hynix's A-die chips, which are the same as those used in most high-speed DDR5 7600, DDR5 8000, and other high-speed memory modules on the market. This also implies that the memory has excellent overclocking potential. The SPD of this memory provides an EXPO-DDR5 6800 configuration and an EXPO-6400 configuration. Users can simply enable the AMD EXPO memory one-click overclocking feature in the motherboard BIOS, select the configuration, and instantly boost the memory speed to the maximum DDR5 6800, with the memory voltage increased to 1.4V, operating at a low latency setting of 34 (CL) - 45 (tRCD) - 45 (tRP) - 90 (tRAS) @ 2T. According to our actual measurements, on the AMD Zen 5 platform, it can also easily use the 1.4V, 34-45-45-90 latency settings to run stably at DDR5 7200, providing higher memory bandwidth and lower memory latency.

Compare the processor without using Intel Default Setting settings! How do we test?

Test Platform

Motherboard: ROG CROSSHAIR X670E HERO, ROG MAXIMUS Z790 DARK HERO

Processor: Ryzen 7 9700X, Ryzen 5 9600X, Core i5-14600K, Core i7-14700KMemory: Kingston FURY Beast DDR5 RGB 6800 32GB kit @ DDR5 7200

Storage: Yangtze Memory TiPro7000 Trinity Edition 1TB

Graphics Card: GeForce RTX 4080 Super

Power Supply: ROG THOR 1200W

Operating System: Windows 11

Compared to Intel processors, we do not use the Default Setting but instead adopt the ASUS Advanced OC Profile configuration, which means unlocking current and power consumption restrictions.

Next, we will conduct a detailed test on the Ryzen 7 9700X and Ryzen 5 9600X processors using the Zen 5 architecture, focusing on whether they can compete with the 14th generation Core processors with more cores as described by AMD. For instance, can the Ryzen 7 9700X match the Core i7-14700K, which is designed with 20 cores and 28 threads, and can the Ryzen 5 9600X compete with the Core i5-14600K, which has 14 cores and 20 threads?

Although currently, Intel's 13th and 14th generation Core K-series processors have been pushed to the "forefront" due to stability issues, and motherboard manufacturers have introduced BIOS settings that significantly reduce performance but improve stability known as Intel Default Setting, in order to showcase the maximum default performance that Intel processors can achieve, we will still follow the previous testing methods for Intel K-series processors and not use the Intel Default Setting. For example, in this test, for the 14th generation Core processors, we will use the ROG MAXIMUS Z790 DARK HERO motherboard and select the "ASUS Advanced OC Profile" setting to unlock the processor's current and power consumption restrictions, with a temperature wall maintained at 100°C to maximize its default performance.

Similarly, for the AMD Zen 5 processors, we will enable the PBO (Precision Boost Overdrive) technology in our tests but will not activate overclocking feature options such as Curve Optimizer and Curve Shaper. The temperature wall is set at a safer 90°C for general users, which is less likely to cause processor damage, focusing on showcasing their performance under default specifications. At the same time, to release the maximum performance of the four processors as much as possible, we will also pair them with the Kingston FURY Beast DDR5 RGB 6800 32GB kit and overclock it to DDR5 7200, as this memory can stably operate at a higher rate, and high-speed memory can bring higher data transfer bandwidth, enhancing the processor's work efficiency.

Zen 5 processors have superior single-thread performance.Next, we first tested the performance of the four processors using benchmarking software such as Geekbench 6.2.1, CINEBENCH R23, PerformanceTest 11.0, 3DMark, and CPU-Z. The results showed that the Ryzen 7 9700X and Ryzen 5 9600X indeed have better single-core performance, outperforming in most of the benchmark tests. For instance, in the single-core test of Geekbench 6.2.1, the single-core performance of the Ryzen 7 9700X was 3414, which is 10.3% higher than its competitor, the Core i7-14700K. The single-core performance of the Ryzen 5 9600X was 3342, leading its rival, the Core i5-14600K, by as much as 14.5%. In the 3DMark processor test, the single-thread performance of the Ryzen 7 9700X was about 7.1% ahead of the Core i7-14700K, and the single-thread performance of the Ryzen 5 9600X was 12% stronger than the Core i5-14600K.

In the classic CINEBENCH R23 processor rendering test, the single-core rendering performance of the Ryzen 7 9700X was slightly ahead of the Core i7-14700K by about 1.1%, while the single-core rendering performance of the Ryzen 5 9600X had a significant advantage over the Core i5-14600K, leading by 5.5%. In the PerformanceTest 11.0 processor test, the single-thread performance of the Ryzen 5 9600X alone was enough to outperform both the Core i7-14700K and the Core i5-14600K.

The Core i5-14700K only managed to regain some ground in the CPU-Z 17.01.64 test, while the Core i5-14600K still fell short against the Ryzen 5 9600X in this test. We analyzed that this is mainly because the test is only an FP32 mathematical test using the SSE instruction set and does not assess the vector mathematical computing performance of SSE, which limits the test. Newer processor testing software like Geekbench, on the other hand, examines the processor's performance in various aspects such as file compression, image editing, object detection, background blur, text processing, and ray tracing; the single-thread test in PerformanceTest 11.0 assesses the processor's floating-point, sorting, and compression performance more comprehensively. 3DMark processing typically uses half SSSE3 instruction set and half the more advanced AVX2 instruction set, as games in actual operation usually use multiple instruction set calculations and are unlikely to use a single instruction set for all tasks. CINEBENCH R23, however, examines the actual image rendering capabilities of the processor, and both Zen 5 processors performed better in these tests, indicating that they indeed have better single-thread performance in real-world applications.

However, in terms of multi-threaded processor performance, due to the Core i5-14600K and Core i7-14700K having a significantly higher number of compute cores than the Ryzen 7 9700X and Ryzen 5 9600X, their multi-threaded performance is still superior. So, in the most user-concerned gaming applications, which of these four processors will perform better?

Zen 5 processors achieve a complete victory in gaming performance testing.

In the gaming tests, to exclude display bottlenecks and fully showcase the processor's gaming performance, we paired with a high-end GeForce RTX 4080 Super graphics card and tested at 1080p resolution with the highest quality settings. In the 12-game test composed of online games and popular AAA titles such as DOTA2, World of Tanks, Cyberpunk 2077: Phantom Liberty, Far Cry 6, etc., the two Zen 5 Ryzen 9000 series processors achieved a complete victory, defeating their respective Core counterparts. We are not surprised by this result, as games heavily rely on the processor's single-thread performance, and most games do not utilize all of the processor's compute threads. Even games that use 8 compute threads are rare, so naturally, the Ryzen 9000 series processors with stronger single-thread performance will have better performance.

Especially in Far Cry 6, a game where almost all Zen 4 processors would fall behind, the Ryzen 5 9600X and Ryzen 7 9700X completely changed the situation, with their average game running frame rates slightly ahead of the Core i7-14700K and Core i5-14600K. The root cause lies in their better processor single-thread performance. Even when we enable the highest quality settings like ray tracing ultra in Cyberpunk 2077: Phantom Liberty, the difference in processors can show a gap. The average frame rates of the two Core processors can only hover around 98fps, while the average frame rates of the Ryzen 5 9600X and Ryzen 7 9700X can both break through 102fps.In games that do not demand high graphics card performance and run at high frame rates, such as DOTA2, "World War Z: Aftermath," "Godfall," and "Serious Sam: Siberian Mayhem," the gap between the 14th generation Core processors and the Zen 5 Ryzen 9000 series processors becomes even more pronounced. For instance, the Ryzen 5 9600X runs DOTA2 at an average frame rate that is 26.7% faster than the Core i5-14600K, and in "Serious Sam: Siberian Mayhem," it leads by as much as 32.8%; the Ryzen 7 9700X runs "Godfall" at an average frame rate that is 16% faster than the Core i7-14700K, and in "World War Z: Aftermath," it leads by 15%. In summary, the stronger single-thread performance gives the Zen 5 processors a comprehensive advantage in gaming performance over the 14th generation Core products.

Zen 5 processors perform better in most real-world applications

Despite having more cores compared to Core processors, the two Zen 5 architecture-based Ryzen 7 9700X and Ryzen 5 9600X CPUs outperform their competitors in most software applications. For example, in the PCMark10 test that reflects the processor's performance in various fields such as video conferencing, productivity, spreadsheet processing, photo editing, content creation, and video editing, the Ryzen 5 9600X scores higher than the two Core processors. The key reason is that most everyday applications do not utilize all processor cores for computation. For instance, in PCMark10, when using the processor to add de-jitter effects to a video, the Ryzen 5 9600X can achieve a processing speed of 42fps, while the Core i7-14700K's processing speed is 34 fps. And if sharpening effects are added to the video through the processor, the Ryzen 5 9600X's processing speed is 97 fps, while the Core i7-14700K's processing speed is 86 fps. Additionally, in the facial recognition feature commonly used in video conferencing, the Ryzen 5 9600X's facial recognition speed can reach 140.69 fps, significantly leading the Core i7-14700K, which only has a speed of 106.74 fps.

In everyday application tests like WebXPRT4, the Ryzen 7 9700X and Ryzen 5 9600X processors also achieve a comprehensive lead, with the Ryzen 5 9600X defeating the two 14th generation Core processors in these tests. The test mainly includes the following content: image processing of two images using three image effects based on Canvas technology; facial detection and image classification using the Caffe and SqueezeNet models in OpenCV; calculation and display of stock portfolio graphical views using Canvas, SVG, and dygraph.js; encryption of notes in local storage and receipt scanning using OCR; calculation and display of multiple sales data charts using d3.js; and completion of scientific and English homework using Web Workers and Typo.js for spell checking. These applications do not call upon all processor cores, and our observations show that the Ryzen 5 9600X performs better in each task than the two Core processors, naturally resulting in a higher overall score.

Furthermore, we also used CrossMark from BAPCo to examine the processors' everyday application performance. CrossMark is a test software that supports the performance assessment and comparison of whole systems across different hardware and software platforms, as well as different operating systems. The software focuses on testing the overall performance in three major aspects: productivity, creativity, and responsiveness. The productivity test examines the system's document editing, spreadsheet processing, and web browsing performance; creativity assesses the system's photo editing, photo organization, and video editing performance; responsiveness examines the system's performance in program launching, file opening, and multitasking. The test results show that both the Ryzen 7 9700X and Ryzen 5 9600X processors perform better, each defeating their respective competitors.

In the 15-item image editing test of PhotoShop 2021, which relies on single-thread processor performance and requires color conversion and the addition of "palette knife" and "sponge" filters among 15 tasks, the Ryzen 5 9600X and Ryzen 7 9700X, with their better single-thread processing performance, also perform more excellently. Their task execution time of over 60 seconds indicates they are faster than the two 14th generation Core processors that take more than 70 seconds, making them more suitable for image editing tasks.

It is worth mentioning that in the AIDA64 ray tracing performance calculations at different precisions and AES data encryption tests, the Ryzen 7 9700X and Ryzen 5 9600X lead the two 14th generation Core processors by a significant margin. For example, the Ryzen 5 9600X's AIDA64 FP64 ray tracing performance leads the Core i7-14700K by 18%, and the Ryzen 7 9700X's lead is even greater at 51.5%. The fundamental reason is that the Zen 5 processors have complete support for the AVX-512 instruction set, which is used in these calculations. It can be said that support for the AVX-512 instruction set has given the Zen 5 processors a qualitative leap in related applications.

In light of the popularity of AI applications, we also tested the performance of the four processors when running large language models with 7 billion parameters like Mistral 7B. In the test, we input "tell me about micro computer" to examine the speed of word token generation and the time to generate the first token in response to the processor. During the test, we disabled GPU participation, and the processor computation thread count was set to the maximum thread count for each processor, such as 28 for the Core i7-14700K and 12 for the Ryzen 5 9600X.Based on the test results, the Mistral 7B large language model can utilize all thread counts for computation, which is why the Core i7-14700K has the highest token generation speed, but it is only slightly better than the Ryzen 7 9700X. The lead is not quite "worthy" of the much higher core count specification compared to the Ryzen 7 9700X and Ryzen 5 9600X, with its performance only being 4% and 8.7% faster than the Ryzen 7 9700X and Ryzen 5 9600X, respectively. The Core i5-14600K, with 14 cores and 20 threads, is at the bottom of the pack, and it also has a certain gap compared to the Ryzen 5 9600X. We speculate that the key reason is that Mistral 7B can support advanced instruction sets of the processor. The hardware requirements for Mistral 7 mentioned "having AVX, AVX2, AVX-512, and other CPU instruction sets can further improve performance," which is why, despite having fewer cores, the two Zen 5 processors have higher operational efficiency.

Especially in the first token generation time test, the performance of the two Zen 5 processors is very prominent, with both having significantly lower first token generation times compared to their Core counterparts. The Ryzen 7 9700X can generate the first token in just 7.92 seconds, while the Core i7-14700K requires users to wait 24.93 seconds before it starts answering questions. Therefore, even though the Core i7-14700K has a slightly higher token generation speed, the longer first token generation time results in a longer overall task completion time.

Of course, in multi-core intensive computing tasks that rely entirely on all processor cores for rendering and transcoding, the two Core processors with a considerable number of cores still have an advantage and are more suitable for such applications.

Low power supply requirements: Processor power consumption and heat dissipation measurements

Despite such excellent performance, this is achieved after enabling the PBO (Precision Boost Overdrive) technology. So, does the Zen 5 processor bring high power consumption and heat generation when running with PBO enabled? To test this, we also conducted tests on them in an environment with a 360 all-in-one water cooling system. Under default settings, the power consumption and temperature differences of the Ryzen 5 9600X and Ryzen 7 9700X are not significant, with their processor idle package power consumption around 32.8W. After running the AIDA64 FPU stress test, the full load temperatures of both processors quickly hit the 90°C temperature wall. The full load package power consumption of both processors is not high, with the Ryzen 7 9700X at only 142.15W and the Ryzen 5 9600X around 128.6W.

Thanks to their manufacturing process advantages, although their full load power consumption is higher than the nominal TDP specifications, it is much lower than that of the 14th generation Core processors. After all, the AIDA64 FPU stress test power consumption of the Core i5-14600K is around 177W, and the Core i7-14700K's stress test power consumption is around 300W. Therefore, if you use mainstream Zen 5 Ryzen processors, you don't need to buy high-specification power supplies unless you have a high-end graphics card. However, if conditions permit, you can still use a high-performance all-in-one water cooling system to reduce temperatures and allow the processor to accelerate to higher frequencies as much as possible.

The highest overclocking can reach a full core frequency of 5.65GHz

The Ryzen 7 9700X can achieve a full core frequency of up to 5.55GHz

The Ryzen 7 9700X can complete the CINEBENCH R23 multi-core rendering performance test at a full core frequency of up to 5.5GHzIn conclusion, we also attempted to overclock both processors. Given that the maximum boost frequencies of the two processors have already reached 5.4GHz and 5.5GHz respectively, their overclocking potential in a standard cooling environment would not be significant. After multiple attempts, the Ryzen 7 9700X can achieve a maximum all-core overclock of 5.55GHz, completing performance tests with light loads such as CPU-Z. When reduced to 5.5GHz, it can complete heavy-load tests like CINEBENCH R23, increasing the multi-core rendering performance from the default 24,224 points to 24,810 points.

The Ryzen 5 9600X can achieve a maximum all-core frequency of 5.65GHz. It can complete the CINEBENCH R23 multi-core rendering performance test at an all-core frequency of 5.6GHz. Furthermore, the Ryzen 5 9600X can achieve a maximum all-core overclock of 5.65GHz and complete the CPU-Z performance test. When reduced to 5.6GHz, it can complete the CINEBENCH R23 multi-core rendering heavy-load test, increasing the multi-core rendering performance from the default 18,546 points to 18,964 points. Due to its fewer cores, it can operate at higher frequencies.

With advanced technology architecture and process, more suitable for gamers and general users.

Considering the above tests, we believe that the Ryzen 7 9700X and Ryzen 5 9600X, two processors based on the Zen 5 architecture, are products with advanced technology architecture and production processes with moderate specifications. They have outstanding single-core, single-thread performance among the new generation of processors, offering faster operation speeds in online games, AAA titles, and most daily applications. For players who primarily use gaming applications and do not require multi-threaded intensive computing, they are among the most值得购买的processors. With full support for the AVX-512 instruction set, the Zen 5-based Ryzen 9000 series processors can outperform processors that do not support this instruction set in related computations and applications. Additionally, if users do not have a high-performance discrete graphics card but want to experience AI applications, these two processors, with their support for the AVX-512 instruction set, can also provide faster execution speeds in large language models like Mistral 7B. Their execution capabilities can match those of processors with many more cores but without AVX-512 support, allowing users with limited budgets to experience the latest AI technology.

For users with limited budgets, they can opt for motherboards like the ASUS TUF GAMING B650M-PLUS WIFI Heavy Artillery in the thousand-yuan price range.

At the same time, the two Zen 5 processors also continue the low-power advantage. Even with the PBO feature enabled, their maximum full-load processor power consumption is controlled within 150W, which does not put pressure on the power supply or motherboard. Therefore, even though we used a top-tier motherboard product like the ROG CROSSHAIR X670E HERO in this test, if users have a limited budget, they can also choose motherboards like the ASUS TUF GAMING B650M-PLUS WIFI Heavy Artillery in the thousand-yuan price range, which can also fully unleash the maximum performance of the processor.

Finally, it is worth mentioning that although they are new products, the launch prices of the Ryzen 7 9700X and Ryzen 5 9600X are also quite affordable. The official price of the Ryzen 7 9700X is 2,549 yuan, and the official price of the Ryzen 5 9600X is 1,949 yuan. For most working-class users, they are also affordable products. Moreover, these two processors are significantly cheaper than the launch prices of the previous generation's Ryzen 7 7700X at 2,999 yuan and the Ryzen 5 7600X at 2,249 yuan, yet they both possess top-tier gaming performance, application performance, and energy efficiency among current processor products. Clearly, the new generation of AMD Zen 5 processors not only brings a better experience but also makes users' money more valuable, which is the core competitive strength that allows AMD processors to continue to develop and grow.