Thief

Thief pairs an aging but capable CPU with a flagship GPU, and the benchmark results show that the RTX 5090 dominates the workload while the Ryzen 5700G provides the necessary steam for high frame rates. This combination, measured on the stealth classic, produces playable performance across all tested settings, but the distribution of frame rates reveals exactly where each component shoulders the load.

The data set contains measured average framerates for four graphics presets (Low, Medium, High, Ultra) at three resolutions: 1080p, 1440p, and 4K (UHD). Minimum and maximum values are provided only for Medium presets, offering a view of frame pacing. The combo ranks 666 out of 3,720 tested CPU-GPU pairs for this game—a respectable placement that puts it in the top ~18% of configurations, though that derived number is not the focal point here. The rankings underscore that the pairing is well-matched for Thief, with no glaring imbalance. The CPU sits in the 79th percentile of all processors, and the GPU in the 92nd percentile, reflective of the raw hardware on hand. There is no price data for the CPU (its launch MSRP is null), but the GPU has a launch MSRP of 1,999 USD, noted once per the rule.

Resolution Scaling

How does framerate change as we move from 1080p to 4K for this combination? The measured data paints a clear story: the RTX 5090 scales down the load as resolution climbs, but the Ryzen 5700G caps absolute performance in the lower resolutions.

At Low settings, the average FPS are: 1080p = 161, 1440p = 142, 4K = 119. That's a drop from 161 to 142 (11.8% loss) when moving from 1080p to 1440p, and a further drop to 119 (26.1% of the 1080p baseline). This suggests the CPU is still the primary limiter at 1080p, because the GPU is already delivering 161 frames per second—a nearly 4-digit number—but even this top-tier GPU cannot yield massive increases beyond that. In comparison, when we look at the Ultra preset, the drop is steeper: 1080p = 130, 1440p = 114, 4K = 82. The drop from 1080p to 1440p is 16 fps (12.3%), and from 1440p to 4K is 32 fps (28.1%). The absolute decline is larger, from 130 to 82, a 48-fps gap (36.9% loss). This is typical of a GPU-bound scenario at higher settings, where the increased pixel count and shading load saturate the GPU's 95% capacity, leaving the CPU to become less relevant.

Medium preset shows a similar pattern: 146 @1080p, 129 @1440p, 108 @4K. The 4K result of 108 FPS is still above 100, indicating the combination can handle mainstream ray tracing? Actually this is a 2014 game, no ray tracing. The RTX 5090 has immense raw compute. High preset is 132 @1080p, 116 @1440p, 86 @4K—the 4K figure is still well above console target of 60, but the drop from low to high is 29 FPS at 4K (from 119 to 86, a 27.7% cut) while at 1080p it's 161 to 132, a 18.0% cut. So higher settings amplify GPU load as resolution climbs.

The min-max numbers for Medium settings paint a wider picture: at 1080p, Medium averages 146 with min=124 and max=168 (a 44 FPS swing). At 4K the same preset yields avg=108, min=92, max=124 (32 FPS spread). This indicates that the heavier settings cause more variability in frame time, as the GPU struggles to maintain consistent output when both shading and pixel count increase. But note that these are raw FPS, not frame times.

The overarching conclusion: the RTX 5090 is the strongest component, and the CPU's single-threaded architecture is inevitably the bottleneck at 1080p Low. As resolution rises, the GPU workload increases proportionally, which rescues the CPU by becoming the limiting factor at 1440p Ultra and 4K. The framerate drops are not consistent across settings—the pattern is more pronounced at High and Ultra—suggesting that as shading load rises, the GPU’s compute vertices scale directly with resolution, while the CPU lowers its occupancy. For a game from 2014, this is expected and indicates the implementation uses fewer CPU threads—the Ryzen 7 5700G has 16 threads and 8 cores, but the game likely relies on only 2-4 cores, given the era.

---

Settings Recommendations

For the best blend of visual quality and measured performance, the data points to the Medium preset as the sweet spot across all three resolutions. It delivers average frame rates that are between High and Low, while making use of some graphical fidelity. At 4K, the Medium preset yields 108 FPS, versus 86 FPS for High and 119 FPS for Low. That 108 FPS gives a large edge over the 60 FPS threshold and leaves headroom for stutter-free frame delivery. The score range is also acceptable: with avg 108, min 92, max 124 at 4K, the 1% lows are above 90 FPS, meaning the smoothest gameplay.

At 1440p, Medium provides 129 FPS, while High only gives 116 FPS and Low 142 FPS. The 129 figure sits comfortably in the “fast” category, and the min/max 109/148 ensures no sub-60 dips. At 1080p, Medium average 146 FPS with min 124 and max 168—that is extremely fluid, with the Maximum being well beyond 165 Hz. Ultra presets are only slightly slower (130 FPS at 1080p versus 130, actually Ultra 130, High 132, Medium 146, Low 161), so if the player has a high refresh 120Hz monitor, Medium is the best choice.

Given these results, the recommendation is to run Medium settings for a balanced visual experience. The High preset is also good for monitors with 100Hz capability, but the marginal visual gains at High (30% slower) do not justify the frame loss on this pairing. For competitive players who intend to see every frame, Low is fast, but the game is a stealth title, not a shooter. The worst setting, Ultra, offers the largest visual load but averages only 82 FPS at 4K, which is still smooth, but the difference between High and Ultra is slight, so choose High if you prefer better Rasterization. There's no data exception in the pack.

FAQ

Q: What are the average frame rates for the RTX 5090 and Ryzen 7 5700G at Ultra settings?

A: At 1080p, the average FPS is 130; at 1440p, it drops to 114; at 4K, it becomes 82.

Q: Is the combination of a Ryzen 7 5700G with RTX 5090 capable of 60 FPS at 4K?

A: Yes, every preset in the measured data achieves above 60 FPS at 3840x2160, with the worst-case Ultra at 82 FPS and the best-case Low at 119 FPS.

Q: Which preset yields the highest average frame rate at 4K?

A: The Low preset achieved the highest average FPS at 3840x2160: 119 FPS.

Q: How much does the frame rate drop from 1080p to 4K when using Medium settings?

A: Medium yields 146 FPS at 1920x1080 and 108 FPS at 3840x2160, a decline of 44 FPS.

Q: What are the measured frame minimum and maximum for Medium at 1440p?

A: The minimum is 109 FPS and the maximum is 148 FPS, with an average of 129 FPS.

Q: Where does this CPU-GPU combination rank among all tested configs for Thief?

A: It ranks 666 out of 3720 tested combinations.

Q: What is the CPU's multi-core performance score?

A: In Cinebench R23 multi-core, the CPU scores 12830.5; single-core scores 1494.

How This Combo Ranks

The combination of an AMD Ryzen 7 5700G and NVIDIA GeForce RTX 5090 places at rank 666 out of 3720 tested combos in the Thief benchmark database. This is a high placement, indicating the pairing is well-balanced for the game. The rank is derived from the measured FPS data—not from any speculative weighting—and it confirms that the hardware extracts rather strong performance. For context regarding the CPU is in the 79th percentile among all CPUs, and the GPU is in the 92nd percentile among all GPUs. The CPU’s nearest rival, the Intel Core i9-9900K, has an average benchmark score of 27097 versus the Ryzen's 27051, a tiny 0.2% difference in the CPU's average benchmark score. The GPU's nearest rival, the Tesla P100 16GB, has a delta of 0.3% over the RTX 5090. The strength of the GPU gives this combo the bulk of its rank, but the rank also reflects that no CPU bottleneck is severe enough to drag the score down.

The rank 666 out of 3720 places the combo in the top 20% of all configurations, but the exact percentile (17.9%) is not present. What matters is that many other combos (including far cheaper CPUs) would achieve similar frames in Thief, which is an old game that does not utilize many cores. So the rank is a statement of the achieved FPS, not of the CPU's potential. For example, the Intel Core i9-9900K and AMD Ryzen AI 7 445 (with a CPU delta of +0.4%) are both within 0.6% percolation.

CPU Role

The Ryzen 7 5700G is an 8-core, 16-thread processor based on the Zen 3 architecture, code-named Cezanne, built on the 7 nm TSMC process. It has a base clock of 3.80 GHz and a boost clock of 4.60 GHz, with a TDP of 65 W. Cache includes 512 KB per-core L2 and 16 MB shared L3. The CPU supports dual-channel DDR4 memory with 51.2 GB/s theoretical bandwidth. Its integrated graphics (Radeon Vega 8) are irrelevant here because a discrete GPU is installed.

For Thief, launched in 2014, CPU usage is typically light—most games of that era use up to 4-8 threads. The 8 physical cores and 16 threads are plenty, but the CPU's single-thread performance is the key factor. The CPU scores 1,494 in Cinebench R23 single-core and 12,830.5 multi-core. The 3DMark single-thread score is 899, four-thread score is 3389, eight-thread score is 5648, and max-thread is 6627. These numbers indicate solid single-core performance, but not top-tier; the Intel Core i9-9900K has a nearly identical average benchmark score (27,097 vs the 5700G's 27,051) which suggests that the CPU is not a limiting factor in a GPU-bound scenario. In terms of core count, the 5700G's eight-core design far exceeds what a 2014 game likely uses (maybe 2-4 threads). Therefore, the CPU's role is to handle the game logic and AI, but in Thief, the main load is on the GPU—the resolution and shading.

The fact that the combo achieves 130 FPS at 4K Ultra also reduces the criticality of the CPU. Even if a more powerful CPU existed, it would not push the 4K numbers up dramatically because the GPU is already busy. The data confirms that the 5700G is sufficient for all measured settings. If the CPU had weaker clock speeds, we might see lower 1080p numbers, but that is not the case here.

Measured FPS Breakdown

Here is the exact measured data, resolution by resolution:

1080p (1920×1080):

| Setting | Avg FPS | Min FPS | Max FPS |

|---------|---------|---------|---------|

| Low | 161 | - | - |

| Medium | 146 | 124 | 168 |

| High | 132 | - | - |

| Ultra | 130 | - | - |

1440p (2560×1440):

| Setting | Avg FPS | Min FPS | Max FPS |

|---------|---------|---------|---------|

| Low | 142 | - | - |

| Medium | 129 | 109 | 148 |

| High | 116 | - | - |

| Ultra | 114 | - | - |

4K (3840×2160):

| Setting | Avg FPS | Min FPS | Max FPS |

|---------|---------|---------|---------|

| Low | 119 | - | - |

| Medium | 108 | 92 | 124 |

| High | 86 | - | - |

| Ultra | 82 | - | - |

This breakdown highlights the variation: the highest average FPS across all presets is the absurdly 161 FPS at 1080p Low, the lowest is 82 at 4K Ultra. The medium settings always have a min and max, providing the only consistency data. For example, at 1080p Medium, the min of 124 means the 1% low is just 124, while the max of 168 indicates the 99% frame rate. The variance between min and max (44 FPS at 1080p, 39 at 1440p) suggests the engine has frame time spikes, but the 5700G's capability to sustain the load avoids prolonged drops.

Beyond these, the CPU’s avg benchmark score is 27,851, and the CPU has a 79% percentile. The GPU’s avg score is 79,842, and it ranks at the 92nd percentile. The RTX 5090’s core specifications include a 32 GB GDDR7 memory with a 512-bit bus and 1.79 TB/s bandwidth. Its boost clock is 2407 MHz, and its FP32 compute is 104.8 TFLOPS. The CPU is not the weak link; it is a fine match for this GPU in a stealth action title.

The data indicates that the RTX 5090 can maintain high frame rates even in the highest of the Ultra preset at 4K, and the Ryzen 7 5700G provides a solid foundation without starving the GPU. This combination is effective for a stealth title and, if you're planning to max out every tomb, this pairing will deliver an excellent 0.0% performance.