Tech enthusiasts all over the world have waited for this day. We're finally allowed to show you benchmark performance results for AMD's 3DV-Cache technology on Zen 4 Ryzen 7000. The company is kicking off their product stack with the Ryzen 9 7950X (this review), and the Ryzen 9 7900X, which we'll cover in a future review. The most-affordable model, Ryzen 7 7800X3D will launch on April 6. Since this quite a long wait, we simulated the 7800X3D by disabling the second CCD on the 7950X3D, in a separate article that's live now, too.

While everyone expected AMD to build the 7950X3D using two 3DV-Cache dies, one on each 8-core CCD of the processor, the design today is using a different approach. AMD has placed 3DV-Cache on one CCD only, and the other one runs without additional cache. The theory behind this approach is that a 3DV-Cache-enabled CCD has to run at lower clock and voltage, which results in a performance tradeoff. You're getting a ton more cache, but you'll be losing frequency, which will negatively impact workloads that aren't cache-sensitive. Having a second CCD that runs without 3DV-Cache, but at full frequency, is a good idea, the million-dollar question is "how to make sure everything ends up on the best cores."

Intel has integrated "performance" and "efficiency" cores in their processor lineup for two generations now, and their approach is to use a hardware component called "Thread Director" that interacts with the operating system, telling it where to put the various threads on a running system. Intel designed Thread Director to be a black box--you have almost no knowledge of what it does, and why, and you certainly can't customize its behavior. AMD uses a different solution that is more open, you get plenty of control, making it very easy to fix application misdetections or change the behavior for your specific use-case. On the other hand, additional software installation is required, you need Microsoft Game Bar running and Game Mode has to be enabled in the OS, too. Just to clarify, all this applies to the dual-CCD designs only (Ryzen 9 7950X3D and 7900X3D). Ryzen 7 7800X3D has just one CCD, which of course has 3DV-Cache, so there's no such challenges for this processor, but this also means it will run at a lower effective clock frequency, because it can't offload work to high-frequency cores.

Throughout this review we're presenting two additional results "Scheduler set to Prefer Cache" and "Scheduler set to Prefer Frequency". These numbers are based on a BIOS option that lets you customize the scheduling behavior in the processor. You may either use the default (controlled by AMD's drivers with input from game detection) or override that mechanism and push all workloads on the cores with 3DV-Cache first ("prefer cache"), and only when all those cores are busy use the high-frequency cores on CCD2. The third option is to run all tasks on the high-frequency cores first, and put additional load on the 3DV-Cache cores ("prefer frequency"). These results also serve as useful indicator to determine what apps benefit from more cache, or which are better served with more MHz.

Let's start with application performance results first. Here we see the 7950X3D achieve the same performance as Ryzen 9 7950X, an important result. On one hand there's highly intensive workloads like rendering and encoding where the 7950X3D falls behind a bit, because of its lower frequency on the second CCD and the fact that these tasks don't benefit from more cache. On the other hand there's lots of productivity apps that run better with more cache, enough to make up for the loss in frequency. Intel's Core i9-13900K is still a tiny bit faster (2.2%), mostly because it runs at higher frequency, especially in low-threaded workloads. Except for some specific tests, the differences are small enough to call it "even."

Next up, gaming performance. Here the 7950X3D is a clear winner compared to the Ryzen 7000 CPUs without 3DV-Cache. In some games the differences are shocking, especially at lower resolution, like Battlefield V (+34%), Far Cry 6 (+27%), Elden Ring (+24%), Cyberpunk (+23%). On average we see a 14% FPS uplift at 1080p for the 7950X3D vs the 7950X, which is comparable to what 5800X3D gave us vs the 5700X (+16%). It seems that the game detection failed in some titles like Spider-Man and God of War, which post higher numbers when we override the AMD software and use the "prefer cache" BIOS setting. If you encounter such a case, it's really easy to fix though. Press Win+G to open Microsoft Game Bar overlay, click on settings and tick the "remember this is a game" checkbox. Done! No fiddling with executable names or the registry. Still, I suspect that a vast majority of people will not be aware of this solution, or even the fact that the game detection might not work on all games, that's why AMD must ensure they track down these cases and fix them quickly. Thanks to their software-based approach, pushing out fixes should be trivial and can be done in a matter of days, while Intel needs to go down a much more complicated route. I'm not sure if they ever updated anything in Thread Director yet, there certainly isn't any public record of it.

Gaming performance compared to Intel's Core i9-13900K is slightly higher at 1080p, but at 1440p and 4K, the Intel CPU wins, by a small margin. Here again, the 13900K's higher operating frequency is what makes the difference. Of course, if the game scales well with more cache, like the titles I mentioned before, there's simply no stopping the 7950X3D. At 4K, where the bottleneck is with the GPU, the differences are fairly small, and just 5% separates a big range of processors, including the 5800X3D, which is still a fantastic solution for gamers, especially if they already have an AM4 platform setup.

You will certainly see reviews today where the 7950X3D is considerably faster on average than in our review. If we only tested Far Cry, Elden Ring and Cyberpunk the delta would be +18% vs 13900K, and if we had Age of Empires, Mount and Blade II Bannerlord and Spider-Man only, the difference would be -19% vs 13900K. Test scene selection is also very relevant, AMD's own reviewer's guide lists the 13900K as faster than 7950X3D in Cyberpunk 2077, I see the opposite, same in Far Cry 6. Maybe they're using the integrated benchmark, I use actual gameplay in all tests. At the end of the day what matters to you is the games that you play. I tried to pick popular titles from various genres, and also included "RT on" testing in two titles for this new 2023 test suite, which also uses RTX 4090 and Windows 11 22H2 with latest drivers and software. All 40 (!) comparison CPUs were retested on this setup in January and February.

A very important performance result that might be overlooked by many is energy efficiency. The Ryzen 9 7950X3D is an extremely power efficient design that really impresses when it comes to application and gaming performance. With performance that's roughly identical to 13900K, it handles all these tasks with half (!) the power consumed, or less. For example in the highly multi-threaded Cinebench test, every watt consumed will give you 253 points—the 13900K guzzles power and only yields 131 points per watt. In gaming things are even more impressive, here we saw 200 W on the 13900K in Cyberpunk, and the 7950X3D is using only 55 W, wow! The same is true for many other games. If you want a high-end gaming experience and would like to reduce your energy usage, carbon footprint, or the heat dumped into your room then the Ryzen 9 7950X3D is what you need. It will be interesting to see whether the 7800X3D can improve on this, because it runs just with a single CCD and also at lower clocks. In single-threaded workloads the X3D doesn't do so well though, because it has to power the two CCDs and all that cache, neither of which have a meaningful impact on performance of a single thread in Cinebench.

I've been complaining about extremely long boot times in previous Zen 4 reviews, and they are still not fixed in a meaningful way. One workaround is to enable the "Memory Context Restore" BIOS option, which saves some memory training info after the first attempt and reuses that on subsequent reboots. I always wondered why it's turned off by default, now I know. I enabled the option like I always do and was wondering why I kept getting random bluescreens in Windows. Everything worked fine under load, for hours, but when the PC was idle and I stopped moving the mouse cursor, boom, bluescreen. I spent a whole day changing memory timings, voltages, swapping components, reinstalling Windows, until I found out that memory context restore was the problem. This will probably get fixed soon, like several other small bugs that caused hangs during POST after changing a setting in BIOS or Ryzen Master. With memory context restore disabled, it will take 35 seconds from pressing the power button until you are seeing the first output on your monitor from POST. Booting into Windows adds another 10-15 seconds on top of that.

While the original 5800X3D was locked for overclocking, AMD is giving us some control through Precision Boost Overdrive with the 7000X3D CPUs. You still can't use multiplier-based overclocking, but you may raise the various power limits and use Curve Optimizer to adjust the voltage-frequency curve. At stock, temperatures are no problem as long as you use a decent cooler, despite the fact that AMD lowered the thermal limit to 89°C, to protect the 3DV-Cache CCD from too much heat. Once you unleash the PBO settings, you'll be running into the thermal throttle basically all the time. Switching from the Noctua NH-U14S to our Arc Liquid Freezer II AIO made only minimal difference, which means the problem is with the thick IHS, which limits the heat transfer, so no matter what cooling you put on top of the CPU, your temps will always be high. Still, we saw around 2% gains in applications and gaming, which helps in the fight against Intel's offerings.

When I first heard that 7950X3D would come with V-Cache on a single CCD only, I thought it would be a big failure. Instead AMD has achieved the unthinkable, their new flagship processor works great in both applications and games. This is good news for gamers that use their rig for various content-creation tasks, too—the 7950X3D will give you the best of both worlds. In terms of pricing, the Ryzen 9 7950X3D is expensive, as expected. With $700 it matches the launch price of the 7950X, which is definitely nice, but the 7950X has dropped down to $590 now. Saving those $110 makes perfect sense if you do not game at all. If you run a mix of workloads, both gaming and demanding applications, then I would pick the 3DV-Cache version any day over the 7950X. If your core focus is gaming and you only run light applications and browsing, then the Ryzen 9 7950X3D is certainly overkill and you should wait for the 7800X3D in April. Intel's Raptor Lake lineup is still very strong, the 13900K offers outstanding performance in applications and games for $570, the $130 difference to 7950X3D could go towards a faster GPU. If you want a CPU that's "good enough" for 4K with a high-powered GPU, then Intel Core i5-13600K for $290 is hard to beat. If you already have a AM4 setup, then upgrading to the 5800X3D is a great alternative, because you only need a new CPU and can continue using motherboard and memory. It will be interesting to see what happens next, how 7800X3D can affect the AM5 value proposition, and whether upcoming A620 chipset motherboards can help bring the platform cost down.