Review Ryzen 7 9850X3D: The New Gaming CPU King - By a Whisker

The embargo has been lifted on the Ryzen 7 9850X3D and there's two ways to look at the arrival of this new gaming-focused processor from AMD. On the one hand, it can rightfully be claimed that the best has just got better: X3D has been a revelation in the gaming space and the 9850X3D is no different. On the flip side, it is the most incremental of incremental bumps with AMD's claims of a two to three percentage point boost in gaming workloads borne out by our data. Still, in MSRP terms at least, the price bump is also incremental: $499/£439 for the new chip is close to the 9800X3D's original recommended cost.

However, in marketing terms, there is an added wrinkle. As "RAMageddon" continues, what should we make of AMD's claims that you can pair an X3D processor with cheaper memory with no performance drop? What AMD seems to be suggesting is that games have less need to access system memory (subject to various factors such as bandwidth and latency) and so you enjoy the same performance. In practice, however, that should also apply to Ryzen 7 9800X3D and that definitely can be impacted by using less highly specced RAM.

Time for some testing then, with AMD supplying us with a complete review kit. The Ryzen 7 9850X3D arrives in the same unassuming packaging as any other CPU, lacking a cooling solution - as is the norm. I bought in a Thermalright Phantom Spirit 120 EVO cooler, complete with "digital display cover" to complete a set-up based on an Asrock X870E Taichi motherboard, GSkill Trident Z5 Neo RGB memory (2x 16GB DDR5, 6000MT/s, CL28) and Samsung 9100 Pro SSD. All tests were conducted with an Nvidia RTX 5090 FE on stock clocks.

AMD were in contact concerning Cinebench R23 scores, which the firm uses to verify that our systems are working properly. When setting up EXPO RAM overclocking in the Asrock BIOS, I noticed that the 105W TDP power limit was removed, so we should be seeing the processor at its fastest. Compared to a 9800X3D, a 2241 single core, 22248 multi-core score from the new chip faced off against a 2079/22029 result. That's a 7.8 percent boost on one core up against no real difference at all with multi-threading.

Onto the gaming tests then, and bearing in mind that the Ryzen 7 9800X3D is such a known quantity, I stuck with two games. Our custom benchmark run through the Cherry Blossom Marketplace in Cyberpunk 2077 would test multi-core performance, while the classic Far Cry 6 would focus instead on single-core results (the Dunia engine is almost legendary for its aversion to spreading workloads over many threads).

Cyberpunk first and at 1080p resolution with DLSS performance mode on the ultra RT setting - effectively bringing the game's most challenging CPU workload to the fore - the Ryzen 7 9850X3D managed a three percent lead over the 9800X3D. That reduced to just 1.7 percent at 1440p, but the deficit there would have been down to the lightest of GPU interactions influencing the result. Regardless, I'd say that the modest score boosts promised by AMD were borne out. Meanwhile, on Far Cry 6, our tools tallied up a 3.2 percentage lead for the 9850X3D over 9800X3D at ultra settings at native 1080p, with literal point-for-point matching at 1440p.

A quick note about one percent and 0.1 percent lows. This data is highly volatile and when you see how bad the frame-time graphing looks in the visualisations, it's hard to see how this data is useful for trying to quantify a gameplay scenario. I think a better consistency metric is required that would be measured over a much, much longer sample of matched gameplay.

Next up, We've got to talk about AMD's claims that the Ryzen 7 9850X3D can be paired with slower RAM for no performance impact - here's an article with an overview of what's being suggested along with the inevitable bar charts. Looks great, right? A less than one percent hit in using awful DDR5-4800 up against more expensive DDR5-6000. In the current climate, AMD is suggesting you could save hundreds of dollars/pounds/euros by buying bottom of the barrel modules.

Even with our data based on just two games, we can see that there's a lot more nuance to this discussion and there are some eyebrow-raising elements to AMD's numbers. The cited Cyberpunk 2077 frame-rates are so high (over 270fps), we can only assume that AMD did not test on a particularly challenging area with high settings. Meanwhile, our best result in a CPU-limited scenario is 97.93fps.

That result came at 1080p on DLSS performance mode on the ultra RT setting with the 9850X3D paired with AMD's supplied GSkill Trident 6000MT/s CL28 sticks. Turning off EXPO and running the kit at 4800MT/s CL40 saw a reduction over 9.9 percent in frame-rate terms and looking at the graphing, both frame-times and PC latency are more volatile. Next, I increased the frequency to 5600MT/s, again at CL40. That gap reduces down to 4.1 percent. Better, but still a noticeable drop. Interestingly though, I then tweaked the GSkill memory back to 6000MT/s frequency, but adjusted the latencies to match a standard Corsair CL36 kit. Here, the end result was exactly the same as running the kit at CL28.

It's just one game, but going into the testing, I did note some user reports suggesting that frequency still plays a role with X3D processor performance, but latency? Not so much. I repeated the test with Far Cry 6 and here, AMD's claim was borne out. On all four memory configurations, the game delivered nigh-on identical results with just a 0.5 percent max variance. My guess? Some titles can stay within the confines of the X3D processor's cache hierarchy (in their benchmark scenes at least). Others can't and there will be a difference. The closer you can get to the 6000MT/s target - a sweet spot for Ryzen 9000 in general - the better, it seems.

Ultimately, AMD has delivered exactly what it said it would, a clock-boosted Ryzen 7 9800X3D. The new 9850X3D averaged at 5413.5MHz during the Cyberpunk 2077 1080p bench, up against 5198.39MHz on the 9800X3D, which explains the slim delta in results. Far Cry 6? 5449.5MHz on the new processor up against the 5199.7MHz of the old, both chips running with all power limits removed. So, an official 400MHz spec boost from the new processor manifested as something closer to 250MHz in our results.

The Ryzen 7 9850X3D launches tomorrow. If UK recommended prices of £439 hold true, you can currently buy the established 9800X3D for around £400. Obviously, the existing product offers the better deal, but in terms of premium pricing for a halo CPU with diminishing returns in terms of pure performance, I've certainly seen worse and there may be some long term bonus in eventual resale value. Whichever one you choose, just make sure you get at least reasonable memory to pair with it. It's just one game result, but Cyberpunk suggests 5600MT/s CL40 minimum - but 6000MT/s remains the sweet spot.