Thief

Thief, released on 2014-02-24, runs a complex pixel‑lighting simulation on every object. The Intel Core i5-9400 and AMD Radeon RX 5700 XT combination delivers playable performance at 1080p and 1440p, while 4K requires a compromise on quality. At 1080p, Low settings produce an average of 119 FPS, which is the highest measured frame rate, while Ultra at 4K dips to 34 FPS. The Radeon RX 5700 XT, with its RDNA 1.0 architecture, drives most of the load: the GPU percentiles sit at 59 across all GPUs, and the CPU sits at 47 across all CPUs. The data also places the combo rank 2850th out of 3720 tested configurations for this title, indicating that while the pair is not near the top, it delivers competent performance in many scenarios.

Resolution Scaling

The most striking pattern in the measured frames comes from the resolution sweep. At 2560x1440 with High settings, the average is 60 FPS; at 1920x1080 it rises to 88 FPS, while at 3840x2160 the High preset drops to 35 FPS. The 1080p number is more than double the 4K number, and the 1440p sits in between, exactly as a pixel‑bound engine should behave. When the settings are lowered, the scaling becomes steeper: Low at 1080p reaches 119 FPS, at 1440p 104 FPS, and at 4K only 61 FPS. The gap between 1080p and 4K at Low is 58 FPS, which is larger than the corresponding 54 FPS gap at High. That larger absolute difference is a sign that the scene complexity, not just the pixel count, is changing the load.

Consider Medium settings: at 4K the average is 47 FPS with a recorded minimum of 40 and a maximum of 54; at 1440p the average is 79 FPS with a minimum of 67 and maximum of 91; and at 1080p the average is 107 FPS, minimum 91, maximum 124. The quarter‑HD resolution scale‑up from 1080p to 1440p yields a drop from 107 to 79 FPS, a reduction of 28 FPS. The jump to 4K yields another 32 FPS decline, hitting 47. Because the grads are similar across the two steps, the measurable behavior indicates a nearly linear pixel‑throughput bottleneck. The GPU’s 448.0 GB/s of memory bandwidth and 121.9 GPixel/s pixel fill rate are the likely limits, but the data set provides no separate memory-or-Rasterizer analysis.

Ultra settings behave more gently: 1080p Ultra runs at 84 FPS, 1440p Ultra at 57 FPS, and 4K Ultra at 34 FPS. The percentage reduction from 1080p to 1440p is roughly 32% (84 to 57), and the next step is roughly 40% (57 to 34). This increasing relative decline, as resolution climbs, is typical of a GPU that become the constraining factor at high pixel counts. For a game of Thief’s age (2014), even the i5-9400’s six cores are enough to feed the RX below 120 FPS. The frame drop from 84 to 34 at Ultra is a big enough shift to be attributed to the graphics pipeline.

The median frame behavior at 1080p High – 88 FPS – still allows a 60 Hz refresh monitor to be saturated, but the 4K High number (35) forces the user to consider lower presets. At 4K, only the Low preset reaches an average above 60 FPS, at 61 FPS. The scene’s rendering complexity at Low avoids the GPU’s bandwidth saturation, while the Ultra preset drops to 34, which will be far from a playable experience for many. The measured FPS table, derived from the official data, confirms that the 4K segment requires the most careful preset choice.

GPU Role

The AMD Radeon RX 5700 XT is the main driver in this test, built on the RDNA 1 architecture with a base clock of 1605 MHz, game clock 1755 MHz, and boost clock 1905 MHz. Those clocks, combined with 2560 shading units, 160 TMUs, and 64 ROPs, give the card a pixel rate of 121.9 GPixel/s and a texture rate of 304.8 GTexel/s. The data includes synthetic benchmark scores: 16277 points in Passmark_G3D, 9542 in Geekbench OpenCL, 68310 in Geekbench Vulkan, and 74647 in Geekbench Metal. The percentile against all GPUs is 59, meaning it beats 58% of all reviewed GPUs. Its nearest rival in the raw average score is the AMD Radeon Pro 5600M with a delta of +0.1% (that is, the Pro 5600.2% performs slightly higher, but the RX 5700 XT is within a hundredth of a percent). The comparison also includes the NVIDIA RTX PRO 6000 Blackwell (‑0.3%), the AMD Radeon PRO W7500 (‑0.3%), and the NVIDIA GeForce RTX 5090 D V2 (‑0.9%). These rival cards are not used in the actual measured game results, but the the average score parity indicates that the RX 5700 XT sits near the median of current GPUs.

In Thief, the GPU’s 8 GB of GDDR6 memory is never exceeded—the game’s profiles fit within that frame buffer. The core’s boost clock (1905 MHz) is particularly relevant because Thief’s levels contain many dynamic lights and shadow calculations, and the shader load scales with resolution. The performance difference between Low and Ultra at the same resolution shows the core count utilization. At 1080p, Low achieves 119 FPS while Ultra gets 84 FPS, a 35 FPS gap. At 4K, that Low‑to‑Ultra gap narrows to 27 FPS (61 vs. 34). The smaller gap at 4K is due to the GPU hitting its fill‑rate ceiling earlier, leaving less headroom for the extra shader work that Ultra requires.

The RX 5700 XT’s 448 GB/s memory bandwidth also matters for Thief, because the game streams large textures with a 256‑bit bus. The memory clock of 1750 MHz results in 14 Gbps effective data rate. The measured frame times at 4K Medium (40‑54 FPS range) hint that the memory subsystem is not the limiting factor at the lower quality, but at Ultra the GPU falls below the 30 FPS floor. Since the CPU does not show a large role in the percentile comparison (the i5‑9400 is near the 47th percentile of CPUs), the real bottleneck for this setup is the GPU’s 35‑59 FPS range at 4K, not the processor.

CPU Role

The Intel Core i5‑9400 is a six‑core, six‑thread Coffee Lake ‑ 945, base clock 2.90 GHz and boost clock up to 4.10 GHz. Its L3 cache is 9 MB shared, and it uses the Intel Socket 1151 board design. The CPU’s synthetic benchmarks: 668 in Cinebench R15 multi‑core, 94 in single‑core; 2784 in Cinebench R20 multi, 392 single; 6629 in Cinebench R23 multi, 935 single; Geekbench multi of 5141, single of 1389. These raw numbers put it in the 47th percentile of all CPUs, with an average benchmark score of 2254. Its nearest rivals include the Intel Core i7‑1068NG7 (average 2259, delta ‑0.2%), the Intel Core i7‑10710U (2248, delta +0.3%), the AMD Ryzen 5 PRO 1500 (2246, delta +0.4%), and the Intel Core i3‑103405 (2264, delta ‑0.4%). In other words, the i5‑9400 is within half a percent of those four chips.

In Thief, the CPU’s role is to prepare the graphics commands and handle AI patrolling, which is part of the gameplay. The fact that the CPU scores are modest— the 47th percentile—indicates that it is not a workstation monster, but it handles a frame‑rate oriented title like Thief without presenting a 4‑core limitation. The measured frame data show that at 1080p Low, the configuration exceeds 120 FPS, a rate that a six‑core at 4.1 GHz is comfortable feeding. At 1440p and 4K, the rates drop, but that is attributable to the GPU, not the CPU. When we reduce the settings to Low, the CPU is still able to produce 104 FPS at 1440p and 61 FPS at 4Kp, which is a strong indicator that the CPU can sustain the rendering engine at these resolutions.

Thief’s engine is not surprisingly thread‑heavy; the i5‑9400’s six logical processors are more than needed for this title. The L3 cache is 18 MB (9 MB per quad? Actually per spec says 9 MB shared). The number of threads is six, which matches the game’s typical thread split. The measured FPS at all settings and resolutions show no hard CPU bottleneck: even at 1080p Ultra, the system holds 84 FPS, which is a perfectly good result for an aging game. The only time the CPU could become a constrain is at very high refresh rates above 200 FPS, but none of the measured presets get close to that. The Low preset hits 119 at 1080p, which is exactly the maximum frame; but even that is not limited by the six cores.

The table of measured frames, with the minimum and max at Medium settings only, further supports the CPU’s sufficiency. At 1080p Medium the minimum is 91 FPS, at 1440p Medium the minimum is 67 FPS, and at 4Kp Medium the minimum is 40 FPS. If the CPU were too slow, we would expect the minimum frame rates to drop suddenly as the scene complexity rises, regardless of the GPU. The data show no such anomaly, so the CPU is not the binding constraint in this game.

Extensive FAQ

Q: What is the lowest average FPS measured in the test?

A: The lowest average is 34 FPS, occurring at 3840× The preset, with the Ultra settings.

Q: Which resolution and setting combos reach 60 FPS or higher?

A: The data shows 1920×1080 reaches 60+ in every preset (High 88, Low 119, Medium 107, Ultra 84). At 2560×1440, Low gives 104, Medium 79, High 60 (exactly 60), and Ultra is only 57 (below 60). At 3840×2160, only Low yields 61 FPS, while High gives 35, Medium 47, and Ultra 34.

Q: What is the average FPS at 2560×1440 with Low settings?

A: The buffer shows an average of 104 FPS at 2560×1440 with the Low preset.

Q: Are the minimum and maximum frames recorded only for a specific preset?

A: Yes, in the measured data, only the Medium setting at each resolution includes a `minFps` and `maxFps` field. At 1080p Medium, min is 91 and max 124; at 1440p Medium, min is 67 and max 91; at 4K Medium, min is 40 and max 54.

Q: Which GPU benchmark score is the highest for the RX 5700 XT?

A: The highest recorded GPU score is 74,647 in Geekbench Metal, compared to 9,524 in OpenCL, 68,310 in Vulkan, and 16,277 in Passmark G3D.

Q: What is the CPU’s percentile among all processors?

A: The Intel Core i5‑9400 sits in the 47th percentile of all CPUs, with an average benchmark score of 2254.

Q: Can the RX 5700 XT maintain 60 FPS at 4K without any compromises?

Yes, the only 4K preset that averages over 60 FPS is Low (61). The High at 4K is 35, Medium is 47, and Ultra is 34, so any graphical lowered option must be used to stay at 60.

Measured FPS Breakdown

| Resolution | Settings | Avg FPS | Min FPS | Max FPS |

| 4K | Low | 61 | – | – |

| 4K | Medium | 47 | 40 | 54 |

| 4K | High | 35 | – | – |

| 4K | Ultra | 34 | – | – |

| 1440p | Low | 104 | – | – |

| 1440p | Medium | 79 | 67 | 91 |

| 1440p | High | 60 | – | – |

| 1440p | Ultra | 57 | – | – |

| 1080p | Low | 119 | – | – |

| 1080p | Medium | 107 | 91 | 124 |

| 1080p | High | 88 | – | – |

| 1080p | Ultra | 84 | – | – |

The data are measured by the benchmarking site, not simulated. The average FPS is computed over a set of test sequences; the minimum and maximum are only available for the Medium preset.

Settings Recommendations

The optimal preset depends on the target display refresh rate and the user´s tolerance for lower frame rates. Starting at 4K, the only setting that tops 60 FPS is Low, which produces 61 FPS average. For a 60 Hz screen, that is the only playable choice at 4K. If the user is willing to accept roughly 47 FPS, Medium provides an average of 47 FPS with a minimum record of 40 FPS and a maximum of 54 FPS, so it is not consistent. Ultra at 4K is not advisable, as it gives only 34 FPS average.

At 1440p, the user can choose between the High preset (average 60 FPS, exactly at the 60‑Hz line) and Medium (79 FPS). Ultra is slightly below at 57 FPS. For a 165 Hz monitor, the Low preset gives 104 FPS, Medium 76, and High 60, again showing that the cost of extra visual quality is large. The best balance at 1440p for crispy textures would be Medium, which gives a strong 79 FPS and only lets down in Minimum (67). The game has the new generation. For a competitive or high‑Hz user, Low delivers the highest average (104) and avoids any stutter.

At 1080p, all settings exceed 70 FPS, so the user can opt for Ultra at 84 FPS or Medium at 107 FPS. The difference between Medium and Ultra is about 19 FPS, which is worth the extra visual effect if the display is 60 Hz. Low gives 119 FPS, but the visual quality is significantly lower. For most users, the sweet spot appears to be the Medium preset at 1080p, as it yields 107 FPS with a minimum above 90 FPS (which is far from the low average). At 1440p and 4K, the choice depends on refresh rate; the data clearly indicate that the Radeon RX 5700 XT is the bounding component, and lowering settings to High or below provides the stable gameplay expected of a stealth action. These are the only measured presets; no shadows or ambient occlusion adjustments were recorded, so the recommendation, strictly from the table, is to select Low for 4K, Medium for 1440p, and Medium for 1080p to achieve the best combination of frame and resolution.