Thief

The measured data for Thief places the Intel Core i5-14600KF and NVIDIA GeForce RTX 4070 Ti SUPER combination well inside playable territory across every preset below 4K Ultra, and even at 4K the pair manages triple-digit framerates on Low settings. This pairing of a 90th-percentile CPU and a 76th-percentile GPU produces a combo rank of 270 out of 3,720 tested combinations, which means the CPU is doing a large share of the heavy lifting in this 2014 stealth title. In the breakdown below, the data walks through how this rank breaks down, where the GPU's memory and clock resources matter, how resolutions scale, and exactly which preset delivers the best measured experience.

How This Combo Ranks

Within the 3,720 base combos recorded for Thief, this CPU-GPU twosome sits at position 270. Translated to a percentile, that is the top 7 percent of every recorded pairing. The CPU alone holds a 90th percentile position against all CPUs on the same benchmark suite, with a PassMark multi-thread score of 38,697 and a single-thread PassMark of 4,273. The GPU, meanwhile, sits at the 76th percentile of all GPUs in the benchmark, with a PassMark G3D score of 31,811. That disconnect between CPU percentile and GPU percentile shows a pairing where the compute side is overmatched for the title: in Thief the CPU is being used at a higher percentage than the GPU relative to their full databases, which is reflected in the low droplets at 4K versus the massive headroom in 2K and 1080p.

The 270-of-3720 rank is higher than the GPU percentile alone would suggest. At 1080p and 1440p, framerates climb well beyond 100 FPS even on High, which suggests the CPU fast enough to keep Thief's game logic uncapped. At 4K, the GPU steps in as the limiter, with a jump between Low (121 FPS) and Ultra (67 FPS) that indicates a much larger role for the memory bus than for the processor clocks. The deltaPct against the GPU's nearest benchmark rivals is small — the RTX 4070 Ti SUPER sits just 0.4% ahead of the M5000 and the GRID M60-1Q, and 1.9% ahead of the TITAN RTX — but in actual Thief frames, the gap in core count and texture rate matters less than the game's CPU-tick rate.

What the aggregate ranking cannot show is the distribution of the vote. The combo is ranked by a single metric per data row, but runs across four presets and three resolutions. In 12 measured configurations, the weakest outcome is the 4K Ultra 67 FPS average, and the strongest is the 1080p Low 209 FPS. That spread of 209 to 67 FPS shows that in Thief the combo effectively spans entry-level and high-end results purely by adjusting a menu setting, which gives hardware buyers a wide catch instead of a lazy bottleneck.

GPU Role

The RTX 4070 Ti SUPER pairs a 2340 MHz base clock with a 2610 MHz boost clock, theoretically topping out at 44.10 TFLOPS FP32 from 8,448 shading units, 264 texture mapping units, and 96 ROPs. Its memory is 16 GB of GDDR6X on a 256-bit bus, which reaches 672.3 GB/s. That GPU's role in Thief is partly about bandwidth: in 4K Low the card shows 121 FPS while 4K High yields 70 FPS. The difference of 50 FPS comes from the texturing and shading load rather than from VRAM; Thief, released in 2014, does not stretch the effective bandwidth of a 672 GB/s part under a 10-year-old title. The VRAM capacity of 16 GB is far beyond what the game can realistically use, qui trivially for the page because the data only records a few subs—and no VRAM cap is hit. The GPU's role is then to supply 121 FPS at 4K Low with basic texture but without ray tracing (Thief's release predates that feature), and to provide the same 4K frame drop when moved to High.

The GPU's own benchmark table paints a part of the picture: PassMark DirectX 10 reads 181, DirectX 11 reads 278, and DirectX 12 reads 119. In a 2014 DirectX 11 stalwart such as Thief, the effective execution is dominated by the last stages of the pipeline and by bandwidth, not by the raw compute numbers. The 44.10 TFLOPS figure impresses on paper, but the measured fps at 4K High (70) shows how far the game's CPU thread is from saturating that compute. The data's headline result is that the GPU is at least the determining component at 4K: from 92 FPS at Medium down to 70 FPS at High, and 67 FPS at Ultra, demonstrating almost complete reliance on the shader density when presets climb. At 4K, the GPU efficiency drops again since the CPU is at 90th-percentile only occasionally.

Resolution Scaling

Framerate scaling across resolutions clarifies the limiting component. Starting at 1080p Low, the system averages 209 FPS. At 1440p Low, it cuts to 184 FPS, and at 4K Low it falls to 121 FPS. Using the 1080p number as the baseline, the 1440p result is about 1.2× slower and the 4K result about 13% of the original. The rate of loss between 1440p and 4K is more than triple that between 1080p and 1440p, well illustrating the GPU headroom at 1440p making the bottleneck appear only at 4K.

Looking at High setting: 1080p High averages 171 FPS, which drops to 119 FPS at 1440p, and then to 70 FPS at 4K. Here the jump from 119 to 70 is roughly 41 percent of the 1080 Free domain — a steep cliff that points to the GPU's memory bandwidth being the binding constraint under the high preset at 4K. Using Medium: 1080p Medium is 189 FPS, 1440p Medium 156, and 4K Medium 92. The percent changes are: 1080p to 1440p loses about , 1440p to 4K loses over 40% of the intermediate number — again pointed down the GPU curve. At Ultra, the 1080p number 165 FPS slides to 113 FPS at 1440p and then 67 FPS at 4K. The trend in every presets a backwards L: a gentle slope from 1080 to 1440 and a much steeper slope from 1440 to 4K.

The implication is that the core of this combo — the CPU with its 14 cores and 20 threads — is never the bottleneck until the fps exceeds the capable region. At 1080p Low the score 209 FPS, and the CPU still sitting at its 90th percentile benchmark. The scaling from 1440 to 4K phenomenon means the game is pixel-bound, not geometry-bound. There is no quadrupling of frames from 4K to 1080p; the 1080p results are always less than double the 4K results for the same preset, so the rendering pipeline holds enough processor count to stay out of the way early, while demonstrating that the keep is on the GPU as pixels mount.

Measured FPS Breakdown

Below are the average framerates (and min/max where recorded) for each resolution and preset:

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

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

| 3840x2160 | High | 70 | – | – |

| 3840x2160 | Low | 121 | – | – |

| 3840x2160 | Medium | 92 | 78 | 106 |

| 3840x2160 | Ultra | 67 | – | – |

| 2560x1440 | High | 119 | – | – |

| 2560x1440 | Low | 184 | – | – |

| 2560x1440 | Medium | 156 | 133 | 180 |

| 2560x1440 | Ultra | 113 | – | – |

| 1920x1080 | High | 171 | – | – |

| 1920x1080 | Low | 209 | – | – |

| 1920x1080 | Medium | 189 | 161 | 217 |

| 1920x1080 | Ultra | 165 | – | – |

The Medium preset is the only one for which the data logs min/max values across all three resolutions, and those rows show a useful band width. At 1080p Medium, minimum FPS stays above 161 with a maximum of 217 — a ~56 FPS swing that sits in fluid territory. At 1440p Medium, the min is 133 and the low is 180, still a tolerable margin for a 2014-fashioned action title. At 4K Medium, min 78 and max 106 produce an average of 92, with the 78 minimum just above the point the frame pacing can show, but acceptable in the game's slower parts.

Reading resolutions separately, the 1440p band is still strong enough to run Medium at 156 or High at 119. The simultaneous 183 in Low barely makes a difference in frame. At the 1080p end, all presets present 100+ fps; even the Ultra night of 165 FPS is well within measure for a fast response. At the 4K end, only Low breaks past the 100 FPS mark (121), Medium sits at 92, and High and Ultra squeeze into sub-80 territory (70 and 67). The 4K Ultra 67 FPS is the single lowest average in this data, a fact that matches the general principle: the oldest generation title still passes the GPU a large shader load at maximum settings.

Settings Recommendations

Given the measured numbers, there is a clear division between what the hardware can push at 4K and what it can sustain at 1080p. On 1920x1080, any preset is comfortable — the difference from Low (209 FPS) to Ultra (165 FPS) is about 20 percent of the low loading, so a user might choose Ultra for visual fidelity without reaching anywhere near a display refresh that's below 144. Medium at 1080p offers the best measured balance of the three available bands: 189 FPS average, with a min of 161 and a max of 217, providing both the frame cushion and the headroom to a recommended one.

At 2560x1440, the decision splits. Low at 1440p leaps to 184 FPS, but Medium brings the more balanced 156 FPS (min 133, max 180) and is the suggested baseline for users who value fps over a sharp framerate; when frame timing is the priority, High at 119 FPS still clears a 100-FPS threshold. Ultra at 1440p drops to 113 FPS, which is the lowest it gets at this resolution. For a precision game like Thief, the 70-80 FPS drops at high preset are halted by maximizing the Low, but the jump from High to Ultra is not worth it — the 155 FPS vs 113 FPS (difference of 42) shows Ultra costing 35% of the frametime headroom. The stated min/max at Medium (133 mach 180) suggests that a 144 Hz monitor will be satisfied at 1440p Medium. For 4K, there is no value choice beyond Medium: the average of 92 FPS is on the safe side for the same frequency of 60, and the min at Medium is a stable 78. High and Ultra on 4K sit in the 60-70 range, which can still be driven, but Medium is the recommended setting for a 4K run when matching with a 60 Hz display. Low at 4K pushes to 121 FPS, which is useful if the output target is a high-refresh 1440 monitor or a 4K 120 Hz panel but loses the fidelity.

In short, the sweet spot across the entire set is the Medium preset. It provides the lowest fps drop between resolution tiers and the most consistent min/max shape. Among the four record ranges, Medium is the only one to have min/max in all three resolutions, and its curve is the healthiest: 189→156→92 gives , at least predictable degradation from 1080 to 1440 and 1440 to 4K. High is a acceptable choice for 1440p (119 avg), and Low is the pick for 4K maximum framerate, but almost every player is best served between Medium and High.

CPU Role

The Intel Core i5-14600KF contributes 14 cores and 20 threads, a base clock of 3.50 GHz pushing to a 5.30 GHz boost, with 24 MB shared L3 cache and a 125 W TDP. Its architecture is Raptor Lake on a 10 nm process; it carries the ability to handle DDR4 and DDR5 memory, with dual-channel. In Thief's benchmark, the CPU's latest mark Cinebench R20 multi-core scores 13,668 (with single-core 1,929), Cinebench R23 multi-core is 32,544 and single-core 4,594, and Geekbench multi-core is 17,085 / single 1,424. The PassMark multi-thread score is 38,697, single thread 4,271. These numbers place diabetic i5 in the 90th percentile of all tested CPUs, which means that, in general productivity and compute, it is within the same 5% band as a series of AMD Ryzen PRO and Ryzen chips; its nearest rival, the AMD Ryzen 9 7900, reports an average score nearly identical (49,228 vs 49,394, delta just 0.3%).

What matters for Thief is the single-thread behavior. A 90th percentile CPU with a 7.3 GHz boost and strong single-thread Cinebench R23 at 4,594 points is well-above the game's original processing baseline from the era. When the data shows 1080p Low at 209 FPS, that is a byproduct of the CPU feeding the engine frames at a fast rate, not a GPU limitation — the engine's maximum vertex data is weak against the CPU's large core count. However, Thief does not scale beyond a couple threads, but the 14-core 20-thread environment still matters because the game's draw calls are small and the boosted single-core is the primary: R23 single at 4,594 could translate to possibly thousands of draw calls per second, well ahead of what the 2014 game needs.

The i5's memory compatibility with both DDR4 and DDR5, plus a dual-channel controller, is indirectly seen in the frame data. The L3 cache at 24 MB explains why the Trojaned score stays low even at 1080p where the GPU is not loaded; the CPU's cache was to hold the game's code path reducing the threads' shared L3 measures, maintaining high BPS. This relationship is why the top ranking at 270/3720 is substantially better than the GPU's 76th percentile position — the processor's high occupancy at 14/20 cores does not get tapped by Thief but its fast single-core figures feed the game's simulation at a rate that keeps GPU busy in the 70-209 FPS window.

Save for a GPU that limits at 4K, the grid is, the i5-14600KF is a full generation ahead of what any 2014 title can demand, and 4K Ultra is the only readout in the data that proves the CPU is not the issue. Overall, the combination's combined 270-of-3720 ranking, a 4K Medium walkway at 92 FPS, and a console-level (or higher) single-thread rating are the key takeaways: a mid-rank combo that performs near the top of the CPU spectrum while depending on the GPU for 4K fidelity.