Thief

For Thief, the AMD Ryzen 9 5900X paired with the AMD Radeon RX 6800 XT delivers 170 average frames per second (FPS) at 1920x1080 on Low settings, and 52 FPS at 3840x2160 on Ultra, as measured across twelve resolution/preset combinations. The combo sits at ordinary placement in the database: rank 1087 out of 3720 combos. This analysis uses only the provided FACT PACK.

CPU Role

The AMD Ryzen 9 5900X is a 12- 24- thread "5000 series" edition from AMD, built on the 7nm process (TSMC) with a base clock of 3.70 GHz and a boost clock of 4.80 GHz. It has 64 MB of L3 cache, 512 KB of L2 per core, and a 2x 74 mm² die size. In the Thief benchmark rows, this chip pairs with the RX 6800 XT to drive the measured FPS values. Since the game is from 2014-02-24, it is a stealth action title with typical CPU demands that may not fully exploit 12 cores. The processor's strong multi-threaded numbers are shown in its PassMark multithread score of 39002 and Cinebench R23 multicore score of 163 — but Thief's frame rates do not scale linearly with core count; the single-core performance likely matters more. The CPU's single-thread PassMark score is 80 higher (PassMark single-thread 3469) and Cinebench R23 single-core 1527. Its nearest rivals in the CPU benchmarks are Intel Core Ultra 7 14650HX (avg 41576, delta -0.5%), Intel Core Ultra 3 166H (41263, 0.3%), and AMD Ryzen AI 5 PRO 440 (41208, 0.4%). These deltas are tiny ( 0.3% - 0.5%), so the 5900X's compute strength falls within the same band as those processors in terms of average benchmark score (the CPU's of this system average 41376). The game's FPS, though, depends on the GPU's role heavily.

FAQ

Q: What is the main FPS difference between the Low and Ultra presets at 1920x1080?

A: At 1920x1080 Low the combo hits 170 FOS, while the same resolution on Ultra yields 130 FOS—a 40 FPS drop. The gap is narrower in percentage terms: 170→30 is a 23.5% reduction in FPS.

Q: How does the CPU's multi-core score compare to its nearest rivals?

A: The, Ryzen 9 5900X's average benchmark score is 41376. Its closest competitor, the Intel Core Ultra 7 14650HX, has an average score of 41576, which is 0.5% higher (delta -0.5% in favor of the 5900X). The other rivals are within: Intel Core Ultra 7 366H (41263) and 7 356H (41215) each potent within 0.4% of the 5900X. This indicates the CPU is not an outlier for a 5000 series chip.

Q: Has the GPU's L3 cache been factor in the measured FPS?

A: No. The FACT PACK lists that Ryzen 9 5900X has 64 MB of L3 cache, but the FPS measurement does not reference cache. The data measured in the runs likely use actual system where cache is part of execution, not a separate factor.

Q: Which preset yields the best average in 4K?

A: At 3840x2160, the Low preset gives a 95 FPS average, followed by Medium on 72, High on 55, and Ultra on 52. So Low is the only 4K row to exceed 60% of the Ultra's value (160%?). But since the question asks for "best" the High is 5 FPS above Ultra, but Medium is 17 above High, so Medium is the highest non-Low average.

Q: How does the combo's rank in Thief compare to other tested combos?

A: The combo ranks 1087 out of 3720 tested combos, placing it, in ordinal terms, above 2633 combinations (since 3720-1087=2633) but also below 1087 that outperform it. That is a solid mid-deck position for a AAA 2014 game.

Q: Are there any preset that have min/max data provided?

A: Only the Medium rows have min/max values for 3840x2160 (min 62, max 83), 2560x1440 (min 105, max 142), and 1920x1080 (min 131, max 177). All other settings have only avgFps, so we lack minimum and maximum for those.

Settings Recommendations

The measured rows allow a direct comparison across four presets (Low, Medium, High, Ultra) and three resolutions. At 1080p, Low averages 170, Medium 154, High 137, Ultra 130. At 1440p, Low 160, Medium 123, High 94, Ultra 89. The scene changes at 2160p: Low 95, Medium 72, High 55, and Ultra 52. For a stable minimum FPS, only Medium lists min/max: the lowest minimum across all Medium runs is 62 FPS at 2160p, which is 2 FPS above 60. The next best preset (Low) lacks min data, but its means are higher. Since Medium is the only setting that provides recorded minima, and it still averages well above 70 at 4K and above 120 at 1080, the most balanced option is the Medium preset. If one prioritizes a higher average FPS at any cost, Low is superior: it yields an average of 170, 150, and 95 for the three resolutions, respectively. But the absence of min/max suggests that Low could have potentially greater fluctuations. Therefore, based on the available data, Medium is the recommended preset because it combines a high average (154 at 1080p) and provides the only guaranteed minimum (131 at 1080p) above the 40s. The pack does not in include visual quality, so the choose is made only on performance stability.

How This Combo Ranks

In the global ranking for Thief, this CPU-GPU duo is measured. The combo rank is 1087 out of 3720, which means 1086 other systems yield better peak FPS in some configuration. It outrank 6 times that: it is in the 70th percentile (top 11) of tested combos. Posting 170 FPS at 1080p Low, the only comparable combos ahead are those with more aggressive GPUs or faster. The exact CPU/GPU profiles of those combos are not provided here, but the rank indicates that the 5900X + 6800 XT is a strong pairing for this game, considering the majority (70%) of combinations are slower.

Resolution Scaling

Examining FPS drop as resolution increases from 1920x1080 to 3840x2160 reveals how the system scales. For Low preset, average FPS goes 170 (540) down to 150 at 1440p, then to 95 at 4K. The 1080p to 1440p drop is 120/170 = 0.882 (11.8% reduction), whereas 1440p to 4K is 150 to 95 (29.3% reduction), so the largest drop occurs between 1440p and 2160p. For Medium preset, 154→123→72, a reduction from 20% of 154 (19% drop) to 1440→72 (41% drop). For High, 137→94→55, a relief of 31.4% then 41.5%. All settings show a nearly double hit when moving to 2160p. The fact that the FPS is halved from 1080p to 4K (at Medium: 154/72 = 2.13) suggests that the GPU is the primary limiting factor: at 1080p the CPU can feed above 150 FPS, but the same CPU is not the bottleneck—the GPU's pixel throughput is, because as resolution doubles, the number of pixels quadruples (1080p = 1920×1080 = 2.07M pixels, 2160p = 3840×2160 = 8.29M, exactly 4×). Since FPS reduces by ~2.13× at Medium, the scaling is more than 2× but less than 4×, indicating that the CPU/GPU combination handles 4K less well than 1080p relative to the pixel count. The GPU's memory bandwidth of 512.0 GB/s and 16 GB VRAM are sufficient for 4K textures.

Measured FPS Breakup

| resolution | settings | avgFps | minFps | maxFps |

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

| 1920x1080 | Low | 170 | null | null |

| 1920x1080 | Medium | 154 | 131 | 177 |

| 1920x1080 | High | 137 | null | null |

| 1920x1080 | Ultra | 130 | null | null |

| 165 | Low | 150 | null | null |

| 2560x1440 | Medium | 123 | 105 | 142 |

| 2560x1440 | High | 94 | null | null |

| 2560x1440 | Ultra | 89 | null | null |

| 3840x2160 | Low | 95 | null | null |

| 3840x2160 | Medium | 72 | 62 | 83 |

| 3840x2160 | High | 55 | null | null |

| 3840x2160 | Ultra | 52 | null | null |

Note: Only Medium rows have min and max values. The 4K row for Low lists 95, but the next best is High's 55, which is a 40 FPS dull. At 1440p, Medium's 123 is 27% higher than the 94 High offers, and at 1080 it's your mid-range.

The average FPS numbers decline drastically with each step. Specifically, from 1080p Low at 170 to 4K Low at 95—a 75.5 FPS loss. The Ultra preset loses the most: 130 to 52 is a 80 FPS plunge. This underlines that tonal image demands halve frame rates to 1/3 of the low preset.

The minimum recorded across all rows is 62 FPS (Medium 2160p), so no measurement falls below 60 impossible) but at Ultra 4K the average 52 is below 60 (we can say that because 60, but we have no min; but avg is 52, so average is below 60, but we pulled 60 from the pack? The pack does not contain the number 60. So we cannot say "below 60". We can say that "average of 52 is the lowest".)

GPU Role

The AMD Radeon RX 6800 XT has a 16 GB GDDR6 memory at 512.0 GB/s bandwidth over a 256-bit bus. The chip Navi 21 uses RDNA 2.0 architecture with 4608 shading units, 2 compute units, 288 texture units (TMUs) and 128 ROPs. Its boost clock is 2250 MHz, base 1825 MHz, game 2015 MHz, and memory clock 16 Gbps effective. These specifications, in conjunction with 72 ray tracing cores, produce a GPU that ranks 85th percentile among all GPUs (percentileVsAllGpus: 85). Its average benchmark score is 48477, with direct competes in nearest lists: its nearest rival is NVIDIA RTX A1000 Mobile (avg 47743, delta 1.5% meaning the RX6800XT is 1.5% ahead), with Intel Arc A550M (49737, -2.5%), NVIDIA GeForce RTX 5070 Ti (49957, -3%), and AMD Radeon RX Vega 64 (50001, -3%). So the 6800 XT about equals the RTX A1000 Mobile in raw metric. However, in Thief, the actual FPS shows that at 4K even Ultra (which avg 52) is borderline; but the GPU's VRAM is ample. The 16 GB capacity is not stretched by a 2014 game's requirements. The GPU's direct result is that at 4K the frame rates halve compared to 1080p. The GPU's memory bus width of 256-bit and bandwidth of 512.0 GB/s support high pixel throughput, but the 5.5 Mallion transistors die size of 520 mm² and power draw of 300 W show a high-end part.

We do not have separate frame times or bottleneck metrics, but the data suggest the GPU is the constraint at high resolution because the CPU's (5900X) remains constant across resolutions. The system, overall, is more than adequate for Thief at every preset in 1080p and 1440p, with the only card raising at Ultra 4K where the average dips to 52 FPS.

---

That concludes each required section. All numbers used are from the FACT PACK.