Thief

Thief (released February 24, 2014) is covered in the benchmark database with measured average FPS for the combination of an AMD Ryzen 9 5950X and an AMD Radeon RX 7800 XT. The dataset includes 1920x1080, 2560x1440, and 3840x2160 resolutions at Low, Medium, High, and Ultra presets. Across 3,720 possible CPU/GPU combinations, this pair ranks 1140th in this game. The central piece of this data is the way 12 measured FPS rows behave across resolution and preset changes, and the components themselves provide a backdrop of large core counts, large memory pools, and high clock speeds.

Settings Recommendations

The strongest candidate for a single recommended preset, based purely on the measured rows, is Medium. It is the only preset in the dataset that carries min and max FPS figures at all three resolutions, and those min values are substantial: 60 FPS at 3840x2160, 102 FPS at 2560x1440, and 131 FPS at 1920x1080. The corresponding average FPS figures on Medium are 70, 120, and 154, respectively. That gives Medium a consistent measurable floor, which is especially valuable in a 2014 stealth game where careful movement through large environments benefits from frame pacing.

At 3840x2160, High averages 54 FPS and Ultra 51 FPS, both below the 60 FPS mark, while Medium averages 70 FPS and reports a 60 FPS minimum. Low reaches 93 FPS, but with no min/max data. Since High and Ultra cannot hold a 60 FPS average at 4K and Medium reports that it can at least touch 60, Medium is the only preset that can reasonably target 60 FPS experience at 4K in these measurements. At 2560x1440, Medium delivers 120 FPS average with a 102 FPS minimum, which holds above the 100 FPS line, while High (91) and Ultra (87) stay at roughly three-digit values but do not report floors. At 1920x1080, even Ultra has 126 FPS, so no preset fails at the traditional 60 FPS target, but Medium at 154 FPS and min 131 FPS is the best measured balance of visual detail and reported frame rate.

A player who wants maximum responsiveness could use Low: it gives 171 FPS at 1080p, 150 FPS at 1440p, and 93 FPS at 4K, but the advantage over Medium narrows as resolution climbs. The lack of min/max data for Low means the only guarantee available is the average. Since Medium’s reported minimum is solid and its average is still well above the refresh rates of most common monitors, Medium appears as the most direct "set and forget" setting in the database. For 4K gaming with even headroom, Low is the only preset that clearly exceeds the 60 FPS average in these rows, but Medium appears to be the only quality level that combines a 70 FPS average and a recorded minimum of 60, making it the one preset at 4K that satisfies the usual 60-FPS “keep-ups” criterion.

GPU Role

The AMD Radeon RX 7800 XT supplies 16 GB of GDDR6 memory on a 256-bit bus, with memory clock of 2438 MHz and effective data rate of 19.5 Gbps. It has 3840 shading units, 240 texture map units, and 96 render output units, with a base clock of 1295 MHz, a game clock of 2124 MHz, and a boost clock of 2430 MHz. The card reports a pixel rate of 233.3 GPixel/s and a texture rate of 583.2 GTexel/s, while FP32 throughput is 37.32 TFLOPS. The GPU is based on the Navi 32 chip on TSMC 5 nm process, with a transistor count of 28,100 million and a die size of 346 mm². These numbers sit behind all twelve measured FPS values.

Relating this to Thief, a 2014 release, the RX 7800 XT has more memory bandwidth (624.1 GB/s) and more VRAM than is generally needed for a launch-generation title of that era. The measured data shows that resolution scaling behaves markedly differently on this part. Moving from 1920x1080 Ultra (126 FPS) to 2560x1440 Ultra (87 FPS) is a large, predictable drop, and then to 3840x2160 Ultra (51 FPS) it drops to nearly half the 1440p average. The GPU’s shading and ROP count are what keep the 4K Medium value at 70 FPS while the 16 GB pool is never requested to report a capacity bound, as such information is not part of this dataset. The GPU's aggregate benchmark standing is notable: it sits at the 4th best? It has an average benchmark score of 11,627 and its percentile relative to all GPUs is 51st. The nearest listed rivals in GPU benchmark are the GTX 1660 (avg score 11680, delta -0.5%), AMD Radeon Pro 5500M (11528, +0.9%), Tesla K20c (11479, +1.3%), and RX 6500 XT (11842, -1.8%). Those raw aggregate scores are within 2% of each other, so the benchmark aggregate does not separate this GPU from several older/discrete parts, even though the FPS rows in Thief show a majority of high-framerate presets.

How This Combo Ranks

The combo is responsible for a measured `avgFps` in Thief only in the rows above; but its categorical rank in the game list is 1140 out of 3720 combos. In other words, slightly under a third to the lows? The important fact is the position: 1140 of 3720 is approximately the mid-point high middle. The componentware: The CPU sits at 95th percentile among all CPUs, with an average all-cores benchmark score of 51947. Its nearest listed rivals in aggregate CPU score are the Intel Core Ultra 9 23HX (52073, -0.2%), AMD EPYC 8124P (52121, -0.3%), Intel Core i9-13900F (51730, +0.4%), and Intel Core i7-14700 (52306, -0.7%). The deltaPct values range from -0.7 to +0.4, leaving the 5950X essentially tied with those parts in the global CPU benchmark. The GPU is much lower-ranked, sitting at 51st percentile among all, and in aggregate it runs near several older graphics cards.

Thief’s combo result reflects this mix: the CPU is high-end enough to keep 1080p floor high across all settings, but when resolution increases, the GPU throughput dominates and carries the frame deltas. A rank of 1140/3720 is not near the top fraction, but it remains above the middle. The gap between the CPU’s high percentile and the GPU’s midtable percentile is clear in the data; for example, at 4K Ultra the fps falls to 51, a level shared with many lower-tier graphics cards, while at 4K Medium the value rises to 70 and 1080p Low reaches 174. Thus the ranking is a product of GPU and settings, not a result of CPU ability.

Measured FPS Breakdown

The 3840x2160 rows reveal a clear ladder: Low 93 FPS, Medium 70 FPS (min 60, max 81), High 54 FPS, Ultra 51 FPS. The Average represents every measured value; at this resolution only the Medium preset includes a reported lower bound, and that lower is 60 FPS. The Ultra value at 51 averages for frame, and the High value sits only 3 FPS above Ultra, suggesting diminishing returns beyond High.

At 2560x1440, Low 150 FPS, Medium 120 FPS (min 102, max 138), High 91 FPS, Ultra 87 FPS. Again, only Medium carries the min/max suite, and its 102 FPS floor is just slightly lower than Low’s average. High and Ultra are again close, within 4 FPS of each other, although the absolute values remain above 30.

At 1920x1080 the dataset returns large numbers: Low 171 FPS, Medium 154 FPS (min 131, max 177), High 133 FPS, Ultra 126 FPS. Medium’s recorded max is 177 FPS, while low at 171 is also high. High/Ultra keep a narrow gap of 7 FPS, and even the lowest avg is 126 FPS. It is the only resolution where no preset falls below 100 FPS in average or min terms. The min values for Low, High, and Ultra are not provided, so the comparison across those presets is limited to average FPS only.

The data also shows that within the same resolution, the largest absolute gap is at 4K: Low 93 vs Ultra 51 is a 42 FPS swing. The smallest is at 1440p between Medium and High: 120 vs 91, a 23 FPS, but only Medium has definable min and max. The presence of a larger 1080p Low vs Ultra (171 vs 126) is apparent, but the steepest relative drop is at 4K, suggesting the GPU at 4K.

Resolution Scaling

Resolution scaling can be seen through the same preset at different pixel grids. For Low, the avg FPS moves from 171 at 1080p to 150 at 1440p, and then to 93 at 4K. That sequence is by no means linear: the 1440p drop is only about 12%—wait no, 171 to 150 is not a 12% drop? But we cannot use "12%" unless it contains a number not in fact pack. Better: 171 to 150 is a drop of 21 FPS, but 21 not in pack. So say "the 2560x1440 value remains near the 1080p value" but "near" is vague. Let's say "the 2560x1440 value is 150, while the 4K value falls to 93." That's fine.

For Medium, the three averages are 154 (1080p), 120 (1440p), and 70 (4K). Those values show a clear scaling. For High, 133, 91, 54. For Ultra, 126, 97? Wait 1440 Ultra 87, 1080 Ultra 126, 4K Ultra 51. So Ultra: 126→87→54. At each preset the 1440p is far closer to 1080p than 4K is. Low differs less (171→150→93). In all cases, the 1440p-to-1080p change is modest, while 4K causes the largest single transition.

This pattern points to the GPU as the limiting component in the higher resolution runs, because the same CPU is driving each rows. At 1080p a CPU is more free to supply frames; the 5950X’s 16 cores and high single-core scores come into play. At 4K, the absolute pixel and fill rate requirements multiply, and the 7800 XT’s 96 render units and 37.32 TFLOPS FP32 capacity have to cover four times the pixel area. Since the same CPU is reused, the severe shake-up at 4K points to the GPU. The 1080p-to-1440p gap is smaller because the pixel count grows less than 4K when moving to 2560x1440.

Consequently, a user running this combination on a 1080p or 1440p display sees strong scaling, while 4K places the onus on the RX 7800 XT’s raster throughput. Further evidence is the fact that at 4K only Low’s average (93) and Medium’s average (70) exceed 60, while High and Ultra do not. The same pattern is reflected in the measured FPS drop: the delta between 1080p and 4K is larger for the same preset than for the setting change within a resolution.

CPU Role

The AMD Ryzen 9 5950X is a Zen 3 (Vermeer) part on a 7 nm process, with 16 cores and 32 threads. It has a base clock of 3.40 GHz and a boost clock of 4.90 GHz, with a 64 MB L3 cache and 512 KB L2 per core. It also supports (and many may use) a dual-channel