Thief

The AMD Ryzen 9 7900X and NVIDIA GeForce RTX 5090 combination delivers a compelling, though not entirely linear, performance curve in Thief. The data reveals a system where the GPU’s immense capabilities are progressively constrained by resolution, while the CPU provides a robust foundation that rarely bottlenecks. This analysis breaks down the measured frame rates to understand where the balance of power lies.

Resolution Scaling

The transition from 1920x1080 to 3840x2160 exposes a classic shift in system limitation. At 1080p, the combo achieves its highest frame rates, with the Low preset hitting 209.1 FPS. This figure suggests the CPU and GPU are working in concert, but the relatively modest scaling from Low to Ultra at this resolution — 209.1 FPS down to 168.8 FPS — indicates that the graphics card is not yet fully saturated. The processor’s strong single-thread and multi-thread performance, evidenced by its Cinebench R23 scores of 6168 and 43695 respectively, is sufficient to feed the RTX 5090 without creating a significant bottleneck.

Moving to 2560x1440, the frame rates remain high but begin to show more pronounced GPU dependency. The Low preset drops to 181.6 FPS, while Ultra falls to 144 FPS. The delta between the lowest and highest settings narrows slightly compared to 1080p, suggesting the graphics card is starting to take on a more dominant role as pixel count increases. The data shows a performance drop of approximately 13% from 1080p to 1440p at the Low setting, which is a substantial but expected cost for the resolution increase.

The most dramatic shift occurs at 3840x2160. Here, the GPU is clearly the limiting component. The Low preset yields 153.1 FPS, but the High and Ultra presets plummet to 84.5 FPS and 85.1 FPS respectively. This represents a 45% reduction in frame rate from Low to Ultra at 4K, compared to a 19% reduction at 1080p. The scaling pattern confirms that at 4K, the RTX 5090’s raw throughput is the primary determinant of performance. The Ryzen 9 7900X’s 12 cores and 24 threads are ample, but the sheer pixel workload at 3840x2160 overwhelms the gains from lower settings, highlighting the GPU’s dominance in this resolution domain.

GPU Role

The NVIDIA GeForce RTX 5090’s specifications are central to understanding these results. With 32 GB of GDDR7 memory on a 512-bit bus, the card possesses a memory bandwidth of 1.79 TB/s. This is a critical asset for Thief, a game that can be demanding on memory throughput at higher resolutions. The data suggests that the 4K Ultra preset, which likely utilizes more texture data and higher-quality assets, benefits from this expansive memory pool, avoiding potential stuttering or texture thrashing that could occur with lesser cards.

The GPU’s clock behavior also plays a role. The base clock of 2017 MHz and boost clock of 2407 MHz, combined with its 21760 shading units, provide the computational muscle for the observed frame rates. At 1080p, the GPU is not working at its maximum potential, as evidenced by the relatively close frame rates between Medium (185.7 FPS) and Low (209.1 FPS). This indicates that the card has headroom to spare. However, at 4K, the GPU is pushed harder, and the gap between Low (153.1 FPS) and Medium (115.8 FPS) widens significantly, demonstrating that the shading units and texture units are being more fully utilized.

The RTX 5090’s FP32 performance of 104.8 TFLOPS is a theoretical maximum, and the measured results in Thief show a practical application of this power. The card’s percentile ranking of 94 against all GPUs in the database underscores its position at the top of the performance hierarchy. The data indicates that this GPU is not just fast; it is disproportionately fast at lower resolutions where CPU overhead is relatively higher, but it truly shines when tasked with the heavy lifting at 4K, where its architecture can flex its full capabilities.

How This Combo Ranks

Within the database of 1642 tested combinations, this pairing of the AMD Ryzen 9 7900X and the NVIDIA GeForce RTX 5090 achieves a rank of 22. This places it in the top 1.3% of all tested combos, a position that reflects the exceptional performance of both components. The rank is a composite measure, but the measured FPS data suggests the ranking is heavily influenced by the GPU’s dominance, particularly at higher resolutions where the CPU’s influence wanes.

The CPU’s own standing is solid, with a percentile of 93 against all processors. Its nearest rival, the Intel Core Ultra 5 235, has an average benchmark score that is a negligible 0.1% lower, indicating that the 7900X is at the very top of its class. Similarly, the GPU holds a 94th percentile ranking, with its nearest rival, the NVIDIA GeForce RTX 5090 D, scoring just 0.1% higher. This near-parity between the CPU and its rivals, and the GPU and its rivals, suggests that the combo’s rank of 22 is a result of synergy rather than a single component outpacing its peers by a wide margin.

The ranking is not the absolute top, but it is a remarkable position for a combination that features a processor from the 7000 series and a graphics card from the 50-series. The data implies that to achieve a higher rank, a combo would need either a CPU with even faster single-thread performance or a GPU with an even more substantial lead over the RTX 5090, which the benchmark data for the latter’s rivals does not show. The combo rank of 22 is a strong statement of the system’s overall capability in Thief.

Settings Recommendations

Based on the measured FPS rows, the optimal settings balance depends heavily on the target resolution and desired frame rate. At 1920x1080, the High preset at 167.9 FPS offers a smooth experience with likely better visual fidelity than Medium (185.7 FPS). The Ultra preset at 168.8 FPS provides a marginal improvement over High, suggesting that the extra graphical features have a negligible cost at this resolution, making Ultra a viable choice for users with high-refresh-rate 1080p monitors.

For 2560x1440, the High preset at 148.8 FPS appears to be the sweet spot. It offers a significant visual upgrade over Medium (168.9 FPS) while maintaining a frame rate well above 144 FPS, which is ideal for standard 144Hz monitors. The Ultra preset at 144 FPS is also playable, but the drop from High is more pronounced, and the visual difference may not justify the performance loss for competitive play.

At 3840x2160, the choice is more challenging. The High preset at 84.5 FPS is playable but does not hit the 90 FPS mark that some users prefer for smoothness. The Medium preset at 115.8 FPS is a more balanced option, offering a substantial performance boost over High and Ultra while likely retaining most of the visual impact. The Low preset at 153.1 FPS reflects the GPU’s power, but it sacrifices too much visual quality for a 4K experience. For most users at 4K, the Medium preset provides the best trade-off between fidelity and frame rate, with Ultra being reserved for those who prioritize visuals above all else.

Measured FPS Breakdown

The complete dataset provides a clear picture of performance across all tested configurations.

At 1920x1080, the Low preset achieves 209.1 FPS, which is the highest frame rate recorded for this combo. Medium follows at 185.7 FPS, a 11% reduction. High and Ultra are closely matched at 167.9 FPS and 168.8 FPS, respectively, showing that the most demanding settings have a similar cost at this resolution.

At 2560x1440, the frame rates are lower but still robust. Low delivers 181.6 FPS, Medium drops to 168.9 FPS, and High falls to 148.8 FPS. Ultra is the slowest at 144 FPS. The scaling from Low to Ultra is a 21% decrease, a more significant penalty than at 1080p, which underscores the growing GPU load.

At 3840x2160, the pattern changes dramatically. Low yields 153.1 FPS, Medium drops to 115.8 FPS, and High falls to 84.5 FPS. Ultra is nearly identical to High at 85.1 FPS. The gap between Medium and High is 27%, highlighting the substantial cost of higher-quality textures and effects at 4K. The data confirms that the RTX 5090 can handle 4K at high frame rates, but only when settings are managed carefully.