AMD Ryzen 9 9950X3D Review: Stunning gaming performance meets top-tier productivity

AMD's 3D V-Cache technology has proven to be the killer tech that delivers dominating performance over Intel in gaming, and the second-gen revision propels the 16-core, 32-thread $699 Ryzen 9 9950X3D to a whole new level of gaming performance — all while preserving the lion's share of performance in productivity applications. Our CPU benchmarks show the 9950X3D is an incredible 37% faster than Intel's flagship Core 9 285K on average in 1080p gaming. It also beats Intel’s fastest competing gaming chip, the Core i9-14900K, by 26% on average in our test suite. In fact, on average, the 9950X3D even ties the hands-down best CPU for gaming on the market, the Ryzen 7 9800X3D.

Naturally, performance will vary by title due to the vagaries of the 3D V-Cache tech, but the 9950X3D is impressive in a wide range of workloads, and its higher 170W TDP also unlocks exceptional Intel-beating performance in many of our productivity benchmarks.

AMD's game-boosting 3D V-Cache tech is now in its third generation, and this time around, AMD moved the vertically-stacked 64MB L3 cache chiplet from the top to the bottom of the compute chiplet to enable higher thermal headroom (details here). As a result, the 9950X3D has the same 170W/230W TDP threshold as the standard Ryzen 9 9950X model instead of the 50W lower rating the company used with the previous-gen 7950X3D model, which ultimately hampered performance.

AMD's goal is to master both disciplines — gaming and productivity — with its dual-chiplet X3D processors, but that's a tricky technical challenge. To build on the success of the first-gen models, the company also refined its chipset drivers to deliver more accurate thread scheduling in some corner cases. This helps reduce the performance delta between its single- and dual-chiplet X3D models. AMD also whipped in full overclocking support for the 9950X3D, which we’ve tested, allowing you to eke out more performance in gaming and productivity work.

AMD also launched the lower-tier $599 Ryzen 9 9900X3D today, saying its new drivers help reduce the vast differences between the two last-generation models. We don't yet have the 9900X3D for testing, but we're working to obtain a sample.

If you're looking for a quick take on performance, see the image album below, which breaks down performance in gaming and productivity applications. These overall measurements are backed up by our benchmarks on the following pages, so be sure to check those out. As you can see, the Ryzen 9 9950X3D is a tremendous chip in nearly every single category.

Image 1 of 4

For now, all eyes are on the Ryzen 9 9950X3D, and from what we've seen, it's another grand slam against Intel. Team Blue still hasn't developed a counter to AMD's 3D V-Cache technology, effectively ceding the gaming crown to its rival in the desktop PC market. AMD's new optimizations serve widen the gap even further. Let's take a quick look at the new tech, then move on to the benchmarks.

AMD Ryzen 9 9950X3D Pricing and Specifications

As with the prior-gen 7000X3D chips, the two new 9000X3D chips use two compute dies, with one die featuring a 3D-stacked V-Cache chiplet that increases L3 cache capacity to 128MB. This keeps game data close to the execution cores, boosting performance in many types of titles (but not all). Meanwhile, the other standard chiplet can boost to higher frequencies to deliver more performance in both single- and multi-threaded tasks. Overall, the Ryzen 9 9950X3D is built on the same foundation as the Ryzen 9 9950X; it just has a single L3 SRAM chiplet placed under one of the compute dies.

The Zen 5-powered Ryzen 9 9950X3D comes armed with 16 cores, 32 threads, 144 MB of total cache, and a peak boost clock rate of 5.7 GHz. This chip has a 170W TDP, 50W higher than the prior-gen model. That's because AMD’s previous-gen 3D V-Cache chips employed a 3D-stacked chiplet on top of the compute die, which trapped heat and constrained the operating temps/voltages, ultimately forcing AMD to lower the TDP.

The new models now have the L3 SRAM die on the bottom of the chip, removing a barrier between the hotter compute die and the chip cooler. That unlocks higher thermal headroom. AMD extracts the additional headroom with the higher 170/230W TDP, allowing higher clocks and longer sustained boost activity.

The Ryzen 9 9900X3D sports 12 cores, 24 threads,140 MB of total cache, and a 5.5 GHz boost. This chip has the same 120/162W TDP as its predecessor and its non-X3D counterpart, the Ryzen 9 9900X, so we expect significantly less performance than the 9950X3D across the board. AMD hasn't sampled this chip but says it has dramatically reduced the performance gulf between the two Ryzen 9 X3D models. We included the company's 9900X3D benchmarks at the bottom of the following page.

AMD's Updated Chipset Drivers 

Image 1 of 5

We've covered AMD's chipset drivers in-depth in the past. The drivers have a suite of components that enable its dual-chiplet processors to operate as close as possible to the single-chiplet comparables, despite having only a single L3 cache die under one of the compute chiplets.

AMD's first dual-chiplet X3D processors employed a new version of thread targeting that works in tandem with putting unneeded cores to sleep, thus forcing game code to run on the single chiplet that houses the performance-boosting L3 cache. However, this implementation made an irreversible change to the operating system that could hamper performance if the chip were later swapped out for a single-CCD processor, with the only remedy being a complete reinstallation of the operating system.

As you can see in the slides above, AMD has now fixed that issue with an updated Provisioning Packages Service (the engine that manages core parking and thread targeting). After roughly 15 minutes of idle time, this driver now automatically detects when a new processor has been installed in the system and adjusts the provisioning accordingly, so there's no need for an operating system reinstall. Things are essentially plug-and-play now, as they should be. AMD also addressed a shortcoming with its 3D V-Cache Performance Optimizer, so it now works when Virtulization-Based Security (VBS) is enabled on Windows 10.

Despite multiple components working in concert to ensure that games run smoothly on the dual-CCD models, some game titles remain problematic. To fix this, AMD revived its Application Compatibility Database (ACD), a technology that debuted with the first Threadripper CPUs. The ACD is a list-based feature that detects when certain games are launched (listed in the image above). It then triggers a mechanism that reduces the number of threads, thus hiding them from the operating system and fully preventing the game code from running on the unoptimized chiplet.

Internally, AMD has affectionately nicknamed this 'Core Lie' because the feature lies with the operating system about the number of cores available. This mechanism assures optimal performance with several of the more stubborn titles, helping to once again reduce the difference between the single- and dual-chiplet X3D models.

Let's see what all of this looks like in our gaming benchmarks on the next page.