AMD Ryzen 9 7950X3D review: the new fastest gaming CPU

AMD’s Ryzen 7 5800X3D CPU was a remarkable addition to the AM4 platform, boasting superior gaming performance compared to other Ryzen 5000 processors, thanks to its innovative 3D V-Cache design. The Ryzen 9 7950X3D and 7900X3D are the latest successors to this groundbreaking CPU, aiming to replicate its success. The Ryzen 7 7800X3D is also set to launch in April.

Can the three impressive performers dethrone Intel’s 13900K as the best gaming processor? With a superior socket, faster DDR5 RAM, and advanced manufacturing process, will the larger L3 cache have the same impact on Ryzen 7000 as it did on Ryzen 5000? Our testing of the flagship Ryzen 9 7950X3D, boasting 16 Zen 4 cores and 3D V-Cache, priced at £699/$699 – the same as the original 7950X – will reveal the answer.

Before delving into the gaming benchmarks and content creation, it’s important to understand what sets the hardware of the 7950X3D apart. Unlike previous Ryzen CPUs, which utilized a chiplet design with low to mid-range parts using a single chiplet and high-end parts using two, the 7950X3D features an asymmetric design. One of its chiplets has been upgraded with 3D V-Cache, while the other retains the smaller cache size and frequency of the 7950X it’s based on. This upgrade comes at a slight cost to the maximum frequency of the chiplet, but it’s a fascinating development nonetheless.

While the asymmetric design may seem unusual and has its limitations, it also offers distinct benefits. Firstly, producing a CPU with a single 3D V-Cache chiplet is more cost-effective than one with two, resulting in lower prices. Additionally, applications that require a larger cache, such as games, can operate primarily on those cores, while tasks that do not benefit can utilize the higher frequencies provided by the non-V-Cache cores.

AMD indicated during its briefings that incorporating a 3D V-Cache in both CCDs of a design may not be worth the cost and flexibility drawbacks, despite the potential performance benefits.

AMD has made significant additions to its chipset software to ensure that games and apps are properly assigned to the appropriate cores. When installing the new chipset drivers, if a two-CCD 7000X3D CPU is detected, users will receive a background service, a background process, a new entry in the device manager, and other features. However, if the same chipset driver version is installed without an X3D CPU and then swapped in later, the additional functionality will not be activated unless the chipset drivers are reinstalled. This has been confirmed through personal experience.

AMD has implemented a clever technique that utilizes the Windows Game Bar to detect when a game is active. This technique involves ‘parking’ the frequency cores, specifically numbers 16-31 in the case of the 7950X3D, to ensure that Windows prioritizes the V-Cache equipped cores for games. By using the Game Bar overlay, users can inform Windows that any application is a game and run it on the high cache cores. Additionally, there are registry entries that can disable this behavior for specific games, although only League of Legends is currently called out. Users can also modify this behavior in the BIOS by manually selecting auto, cache, or frequency cores as preferred for everything. This functionality was tested on page six to determine the extent of the 3D V-Cache advantage.

Although initially intended for the 7950X3D and 7900X3D, the core parking behavior could potentially enhance the performance of the 7900X and 7950X CPUs. This is particularly relevant as the single-CCD 7700X outperforms these CPUs in certain games. During an AMD briefing, I proposed this idea and received a positive response indicating that it could be a logical progression. Therefore, it is worth keeping an eye on any developments in this area.

The table below displays the positioning of the three Ryzen 7000X3D CPUs in comparison to their counterparts. It’s important to note that each of these Zen 4 designs boasts several enhancements over their Zen 3 predecessors, including a 13% increase in instructions-per-clock (IPC), improved execution engine, and better branch predictor. Additionally, they support DDR5 and PCIe 5.0, feature 5nm CCDs and a 6nm I/O die, and require the new AM5 socket for increased power and performance. However, upgrading to these CPUs will necessitate a new motherboard, RAM, and potentially new cooling.

To conduct our tests, we have utilized the identical fundamental configuration as highlighted in our Ryzen 7600 analysis, albeit with the most recent AMD chipset drivers integrated. This encompasses an ASRock X670E Taichi motherboard, G.Skill Trident Z5 Neo DDR5-6000 CL30 RAM, and Asus’ RTX 3090 Strix OC for our graphics card. The cooling system is furnished by a 240mm Alphacool Eisbaer Aurora AiO, which is conveniently compatible with the new AM5 socket.

To store our games, we have opted for three PCIe 4.0 NVMe SSDs, namely a 4TB Kingston KC3000, a 1TB PNY XLR8 CS3140, and a 1TB Crucial P5 Plus. Our setup is powered by a 1000W Corsair RM1000x power supply.

In order to ensure consistency with previous benchmarks, we have opted to utilize a pre-22H2 edition of Windows 11, as it has demonstrated superior performance for Ryzen 7000 CPUs. Additionally, we have employed the most recent BIOS for our ASRock X670E Taichi motherboard, version 1.15.SMU215, along with the latest AMD chipset drivers during the testing phase.

To test various platforms, we utilized top-of-the-line motherboards such as the Asus ROG Crosshair 8 Hero for Ryzen 5000 testing, the Asus ROG Maximus Z590 Hero for 11th-gen Intel testing, the Asus ROG Z690 Maximus Hero for 12th-gen testing, and the Gigabyte Z790 Aorus Master for 13th-gen testing. For DDR4 motherboards, we used G.Skill 3600MT/s CL16 memory, which is considered the optimal choice for DDR4. On the other hand, DDR5 motherboards were tested using DDR5-6000 CL30 for primary testing and DDR5-5200 CL30 to demonstrate baseline (near-JEDEC) performance. The results achieved at 6000MT/s are clearly labeled in the graphs that follow, while unmarked results refer to DDR5-5200 or DDR4-3600 as previously mentioned.

To provide context, we will begin with two brief content creation benchmarks: a Cinebench R20 3D render and a Handbrake video transcode, before delving into the gaming benchmarks featured on pages two to six.

As expected, the performance of the 7950X is slightly lower than the higher-frequency 7950X3D in Cinebench and Handbrake, as these applications do not benefit from additional cache. The Handbrake export is about 10 percent faster on the 7950X, while the Cinebench multi-core test is approximately 7.5 percent faster. However, in the single-core test, the 7950X3D can utilize its higher-frequency cores, resulting in a negligible one percent difference compared to the 7950X.

Despite falling behind the 13900K in multi-core workloads, the 7950X3D is only around 10 percent slower in Handbrake and approximately 13 percent slower in Cinebench compared to the top Intel CPU. However, the 7950X3D still outperforms any previous-generation competitor by a significant margin, with a 36 percent faster score in Cinebench compared to the 5950X and a 32 percent faster score compared to the 12900K.

The power usage of the Ryzen 9 7950X3D is impressive, as it draws significantly less power from the wall (232W) compared to the 7950X in the same HEVC encode workload (368W), which compensates for its lower performance. Moreover, it outperforms the Intel Core i9 13900K (473W) in terms of power usage. Although the margin in power usage for non-AVX workloads, such as games, is expected to be smaller, the power usage is still relevant for content creation, especially in the current economic climate. This highlights the efficiency of AMD’s latest processors in general.

Now, let’s get into the fun stuff – the games. We’ve tested a range of titles, so pick out your favourites from the links below or just hit that next page button to continue the journey.