AMD Ryzen 9 9950X3D2 review: More cache, more cash

The Ryzen 9 9950X3D was already a power-hungry CPU, and the dual-cache version pushes power demands further. AMD hasn’t crossed the line of thermal demands that Intel did with its Raptor Lake and the subsequent Refresh chips, but where the Ryzen 9 9950X3D sat around 200W in all-out multithreaded workloads, the dual-cache model inches closer to 250W, peaking as high as 278W in applications and as 318W in an artificial workload via Prime95 with AVX disabled.

Image 1 of 13

Starting with that peak in Cinebench 2024 with a multi-core render, the Ryzen 9 9950X3D2 consumed 25% more power than the Ryzen 9 9950X3D and its non-X3D counterpart. Compared to the Zen 4 version of this chip, the 9950X3D2 consumes 81% more power. Still, the 9950X3D2 is 12% behind where the Core i9-14900K peaked in this test.

We can see similar disparities in Blender, and in the Junkshop scene, the 9950X3D2 actually exceeded the Core i9-14900K by a few watts. In encoding workloads via Handbrake, the Ryzen 9 9950X3D2 is in lockstep with its single-cache counterpart in all but the x264 10-bit encode.

In the single-threaded y-cruncher test with AVX enabled, the Ryzen 9 9950X3D2 actually drew 11.6% less power than the Ryzen 9 9950X3D, likely due to lower peak clock speed. We disable PBO for our testing (more on that next), so these power draw figures would likely equal out with PBO turned on. There’s clearly enough headroom here for more power in a single-threaded scenario.

In idle scenarios, the Ryzen 9 9950X3D2 is decent. In a true idle state, it matches the Ryzen 7 7800X3D, Core i9-14900K, and base Ryzen 9 9950X3D, showing AMD’s improvements in idle power draw when moving from Zen 4 to Zen 5. In an active idle scenario (YouTube playback), however, power consumption goes up to 38W, which is 11.7% (or 4W) ahead of the Ryzen 9 9950X3D.

Image 1 of 4

Given the tight margins in performance, I expected worse efficiency for the Ryzen 9 9950X3D2 compared to the single-cache model, but that’s not the case. We can see basically identical efficiency in Handbrake and Linpack, the latter of which makes sense given the similar power draw between the two CPUs. In Cinebench, the dual-cache version was 16% less efficient than the Ryzen 9 9950X3D, which is a solid result given that 25% increase in power consumption we saw in this test.

Image 1 of 3

Another way to visualize efficiency is with a scatterplot, which you can see above. For these charts, the bottom-right corner is the best, and the top-left corner is the worst. You can see the Ryzen 9 9950X3D and 9950X3D2 in lockstep. The data here isn’t different from the bar graphs above, just a different way to visualize it.

Test Setup

We use identical test beds to ensure accurate data across our test pool. That includes the same hardware configuration, short of the motherboard and CPU, and the same software configuration. We use a frozen OS image based on Windows 11 24H2. We’re testing the same version of the same apps across our test pool to ensure minor version changes won’t skew our data.

On the hardware side, we use the RTX 2080 FE for application testing. It doesn’t do anything other than provide a display output for our application testing, but it uses the same driver as the RTX 5090 FE, which we use for game testing. This allows us to maintain our frozen OS image with mucking things up with driver reinstalls.

In the BIOS, we make a few tweaks to optimize performance in a way that’s still covered under warranty. We run tests with XMP/EXPO enabled, alongside Resizeable BAR. Windows Virtualization-Based Security is disabled. We manually disable any automatic boosting features that aren’t covered by warranty, as well, including AMD’s Precision Boost Overdrive and Intel’s Extreme power profile.

• MORE: CPU Benchmark Hierarchy
• MORE: AMD vs. Intel
• MORE: AMD Ryzen 7 9800X3D review