Roughly a year ago we put the Ryzen 9800X3D and Core Ultra 9 285K through a marathon head-to-head across 45 games, and the result was decisive: on average the Ryzen chip was 24% faster.
Since then Intel has issued a steady stream of firmware and tuning updates, including a new one-click overclocking feature called Boost 200S. That makes this a good moment to revisit the matchup and see whether any of those changes have meaningfully shifted the landscape for Intel's flagship part.
For this updated round of testing we narrowed the list to a dozen titles, most of them newly released or recently updated, and evaluated each with two quality presets, typically medium and ultra.
The Core Ultra 9 285K is paired with an Asus ROG Maximus Z890 Hero running BIOS version 2201 and CUDIMM DDR5 8200 memory from G.Skill. We are testing two configurations: a standard out of the box setup using XMP, and a second using Boost 200S. That mode raises memory frequency to 8,000, adjusts key voltages, and clocks up the Die-to-Die and Uncore Fabric interconnects.
In theory this is a one-click tuning option designed to extract extra performance without manual tweaking. In practice, support varies. Of the two 285K samples we have on hand, one runs 200S without issue, while the other routinely crashes. Intel does not void the warranty when it is enabled, but it is also not guaranteed to work, nor to deliver tangible gains.
On the AMD side, some AM5 motherboards include an unofficial one click alternative called X3D Turbo. AMD's official boost mechanism remains PBO (Precision Boost Overdrive), which we have already tested extensively.
For this comparison we are including X3D Turbo results captured from the Gigabyte X870E Aorus Master with its latest BIOS and DDR5 6000 CL30 memory. As with 200S, results depend on silicon quality, and while we have yet to see instability on our hardware, there will certainly be chips in the wild that do not behave as cleanly.
Gigabyte's approach disables the second CCD on dual CCD parts such as the 9950X3D, while on single CCD chips like the 9800X3D it disables SMT and tightens memory timings, which can improve gaming performance. Whether it consistently does so is something this testing set will answer.
All CPU benchmarks are driven by an RTX 5090 to ensure the GPU is not the limiting factor. The aim here is to isolate CPU-bound gaming performance. With that context in place, let us get into the numbers.
Test Setup
| CPU / Motherboard / Memory | AMD Ryzen 7 9800X3D Gigabyte X670E Aorus Master [BIOS F38e] G.Skill Trident Z5 RGB 32GB DDR5-6000 CL30-38-38-96 |
|
| Intel Core Ultra 9 285K Asus ROG Maximus Z890 Hero [BIOS 2201] G.Skill Trident Z5 CK 32GB DDR5-8200 CL40-52-52-131 CUDIMM |
||
| Graphics Card | Asus ROG Astral RTX 5090 OC Edition | |
| ATX Case | Antec Flux Pro | |
| Power Supply | Kolink Regulator Gold ATX 3.0 1200W | |
| Storage | TeamGroup T-Force GE Pro M.2 PCIe Gen5 NVMe SSD 4TB |
|
| Operating System | Windows 11 24H2 | |
| Display Driver | GeForce Game Ready Driver 581.57 WHQL | |
Benchmarks
Cyberpunk 2077: Phantom Liberty
With the medium preset, the 9800X3D was 42% faster than the 285K, delivering 234 fps on average compared with 165 fps. The one click OC options did not significantly change the average frame rates, though 200S improved 1% lows on the 285K by 5% while X3D Turbo improved them by 12% on the 9800X3D.
When switching to the ray tracing results, which are more CPU intensive, we still run into GPU limits even at 1080p with the RTX 5090. As a result the 9800X3D lead shrinks to 8%.
Marvel Rivals
For the Marvel Rivals and Rainbow Six Siege testing we also included CAMM2 data using DDR5 9600 memory with the Core Ultra 9 285K. Unfortunately, this memory speed forces Gear 4 mode, which significantly increases latency. Although bandwidth is higher, the worse latency harms overall gaming performance, which is why results go backwards.
Using the medium preset, the 285K reached 194 fps with DDR5 8200, and enabling 200S improved performance by 1.5% to 197 fps.
The 9800X3D on the other hand achieved 242 fps, making it 25% faster than the 285K when comparing stock configurations with memory profiles enabled. Turning on X3D Turbo boosted the 9800X3D to 257 fps, a further 6% gain, which means that when comparing the one click OC options directly, the 9800X3D was 30% faster.
Switching to the ultra preset results in a heavy GPU limit in this title, which greatly reduces the performance difference between the CPUs. Here the 9800X3D was just 7% faster, and the one click OC features had minimal impact.
Rainbow Six Siege X
With the medium preset, the 285K averaged 488 fps, and enabling 200S improved performance by 4%. Under the same conditions, the 9800X3D was 18% faster. Turning on X3D Turbo provided a 10% uplift for the Ryzen chip, bringing it to 634 fps on average, which meant the 9800X3D was 25% faster when comparing both CPUs with their OC options enabled.
Even with the maximum quality preset, where the gap narrows due to GPU load, the 9800X3D still held a 14% lead.
Assassin's Creed Shadows
Assassin's Creed Shadows is generally more GPU limited, even at the medium preset. Despite this, the 9800X3D was 18% faster, and the one click OC options only added around 2% for each CPU.
At the ultra high preset the 9800X3D extended its lead to 25%, though once again the OC features did not meaningfully change results.
Horizon Zero Dawn Remastered
Using the built in benchmark, the 9800X3D was 34% faster at the medium preset. The one click OC options had almost no effect for either CPU.
At the very high preset the 9800X3D still led by 35%. In this test, 200S helped the 285K by 5%, but even with both CPUs overclocked the 9800X3D maintained a 29% advantage.
Space Marine 2
Next we have Space Marine 2, and with the medium preset the 285K tops out at just over 100 fps, whereas the 9800X3D reached 145 fps, a 37% improvement. Enabling 200S does almost nothing for the 285K, while X3D Turbo improves the 9800X3D by 9%, making it 48% faster when comparing both processors with their overclocking features enabled.
The ultra preset shows a similar trend. The 9800X3D is slightly more GPU limited, but it still ends up 30% faster in stock form or 42% faster when comparing the one click OC options.
The Last of Us Part II Remastered
Performance in The Last of Us Part II Remastered is far more competitive, allowing the 285K to effectively match the 9800X3D. Using the medium preset, the 9800X3D was just 8% faster on average, although 1% lows were 17% better.
With the very high preset, the 285K is able to roughly match the 9800X3D when 200S is enabled, as this improves performance by around 5%.
Spider-Man 2
In Spider-Man 2, the 9800X3D once again leads comfortably, averaging 245 fps, a 28% advantage over the 285K. The 200S feature boosts the 285K by only 2%, and X3D Turbo provides the same 2% increase for the Ryzen processor.
Switching to the ultimate ray tracing preset increases CPU load, but the game becomes heavily GPU bound. As a result, both CPUs perform much the same.
Mafia: The Old Country
In Mafia: The Old Country using the medium preset, the 9800X3D is just 7% faster for the average frame rate, though it is 25% faster for 1% lows. The one click OC features provide no meaningful change for either CPU.
When switching to the Epic preset the benchmark becomes more GPU limited. The 9800X3D leads by just 6% on average, though it maintains a 16% advantage for 1% lows.
Assetto Corsa Competizione
ACC has always favored AMD's 3D V Cache processors, and the 9800X3D continues that trend. Using the medium preset it was 81% faster than the 285K. The 285K gained 6% from 200S, while X3D Turbo improved the 9800X3D by 5%.
Using the Epic preset delivers nearly identical scaling. The 9800X3D was 89% faster, averaging 269 fps compared with 142 fps on the 285K.
Baldur's Gate 3
The 9800X3D is a brute in Baldur's Gate 3, allowing for a completely unnecessary 262 fps with the medium preset, a 94% uplift over the 285K. In this case 200S helped the 285K's 1% lows by 12%, while X3D Turbo boosted the 9800X3D by 12%.
Even with the ultra preset enabled, the 9800X3D was still 88% faster on average and in this example X3D Turbo boosted performance by a further 6%.
Counter-Strike 2
In Counter-Strike 2 the 285K is capped at around 360 fps with both the medium and very high presets. The 9800X3D was 52% faster on average with the medium preset and 51% faster for the 1% lows.
Even when increasing GPU load with the very high preset the 9800X3D maintained a 26% advantage for the average frame rate and a 48% advantage for the 1% lows.
12 Game Average
Looking at the average performance across all 12 games tested, the 9800X3D was 35% faster using the medium settings and 26% faster when using the high or ultra settings.
On average we also saw a 3% gain from Intel's 200S Boost feature, while X3D Turbo provided a 3 to 4% uplift. Both features were underwhelming in practice.
Shader Compilation
One clear advantage the 285K holds over the 9800X3D is multi core performance. While the 9950X3D helps address this, it is not part of this comparison. When measuring shader compilation times the 285K was 28% faster in The Last of Us Part I, and 32% faster in both Stalker 2 and Marvel Rivals.
Brute Force Still Wins + Overclocking Bits
Nothing has meaningfully changed since the launch of the Core Ultra series. Intel has shipped patches and firmware updates that lift performance in a handful of titles, but the broader picture is essentially the same. At launch, the 9800X3D was 33% faster on average in heavily CPU bound testing. In our wider 45 game follow up, it held a 24% lead, a smaller gap explained by the inclusion of more GPU limited titles.
This latest round of testing lands in almost the exact same place. Across these 12 newer games, the 9800X3D was 35% faster using medium settings and 26% faster with higher quality presets that leaned more heavily on the GPU. The medium results mirror the day one review, and the higher preset results match the broader 45 game sample.
No matter how you frame it, the 9800X3D is still clearly ahead of the 285K for gaming. Intel's 200S Boost does not close the gap. High-speed CUDIMM isn't a silver bullet either, and this was proven way back when we first reviewed the Core Ultra series.
The same applies to CAMM2: the memory ratio shift required to run it introduces additional latency, which undermines the added bandwidth and ends up hurting gaming performance.
Another angle that often comes up with these CPU comparisons is overclocking, and unfortunately it has become a very divisive subject over the past few years, and as a result reliable information has become increasingly difficult to find, not just regarding benchmark numbers, but also about the settings used to achieve those numbers.
One of the best resources for this sort of information is SkatterBencher, an unbiased overclocking resource that not only provides results, but also acts as a guide for how to achieve those results.
Granted, SkatterBencher hasn't found any miracle settings that turn the 285K into a gaming beast, but there's certainly some degree of overclocking headroom, provided you have good enough silicon, but most of the impressive gains are enjoyed by productivity applications rather than games.
Following their guide on a strong 285K sample yielded modest improvements, and nothing meaningfully beyond what 200S Boost already provides. The current recommendation from SkatterBencher is simply to enable 200S Boost, raise the all core ratio, and undervolt slightly – there is no hidden tuning profile you can tinker with that will make the 285K go vroom. Memory tuning helps a little more than anything else, but even with DDR5 8200, the uplift remains small and still puts the 285K well behind the 9800X3D.
The Core Ultra 9 285K does retain a meaningful advantage in productivity performance. The Ryzen 9 9950X3D can match it, but it typically costs over 30% more. This gives the 285K some legitimate value as an all rounder, and the Z890 platform is excellent from a features perspective.
The problem is platform longevity. Intel's track record here is not reassuring, and LGA1851 is widely expected to end after a single Arrow Lake refresh. That makes recommending the platform today a difficult sell. For anyone building now, we would either wait to see Intel's next move, or move to AM5 and look ahead to Zen 6 and potentially Zen 7.