Thief

The combination of the Intel Core i7-12700KF and the NVIDIA GeForce GTX 1660 SUPER yields a measured dataset for Thief that shows a clear pattern: performance is heavily dominated by the GPU at higher resolutions, while the CPU is never the primary constraint. Across all four settings presets, average FPS drops by roughly 40–50% when moving from 1080p to 1440p, and by a further 40–50% when stepping up to 4K. The most potent frames per second are seen at 1080p Low, peaking at an average of 120 FPS, while 4K Ultra produces only 27 FPS, indicating a system that scales predictably but cannot sustain high refresh rates at 4K Ultra. This analysis is based entirely on measured data from the benchmark dataset.

Resolution Scaling

The measured FPS for this CPU+GPU pair reveals a steep performance curve as pixel count rises. At every settings preset, the transition from 1920x1080 to 2560x1440 cuts the average frame rate by roughly one-third, and the leap to 3840x2160 halves it again. For example, at High settings the average FPS falls from 69 at 1080p, down to 47 at 1440p, and settles at 28 at 4K. Similar progressions hold across the other presets, as shown by the table (full details in the Measured FPS Breakdown section). At Low settings the 1080p average is 120 FPS, dropping to 82 at 1440p and 48 at 4K. Medium setting registers 91, 62, and 37 FPS for the three resolutions, while Ultra delivers 66, 45, and 27 FPS.

The shape of this scaling pattern is typical for a GPU-bound workload: the frame rate degrades proportionally to the increase in pixel count. The gap between 1440p and 4K is slightly larger than between 1080p and 1440p, which aligns with the multiplication of pixels when moving to 4K. When the GPU is the limiting component, changing resolution compounds the number of pixels processed, and the GTX 1660 SUPER's 6 GB GDDR6 frame buffer and its 192‑bit memory interface become the primary gates.

Crucially, the gap between the four setting presets diminishes as resolution rises. At 1080p, the difference between Low and Ultra is 61 FPS (from 120 to 66). At 4K, that spread narrows to 21 FPS (from 48 to 27). This convergence indicates that at 4K, the GPU's average throughput ceiling is reached even at Low settings, leaving only a modest headroom for graphical detail. Conversely, at 1080p, the graphics settings have a substantial effect on frame rates, confirming that the GPU has enough bandwidth there to allow settings to matter.

Data for the Medium preset includes minimum and maximum FPS measurements, providing a sense of variability. At 1080p Medium, the average is 91 FPS, with min/max of 77 and 105; at 1440p Medium, 62 average (53–72); at 4K, 37 average (31–42). This consistent 30–40% variance between the lowest and highest FPS readings is a consequence of the game’s scenes having varying amounts of geometry, lighting, and particle effects. The overall curve, though, runs along the expected lines—the more pixels, the more work for the GPU, and the worse the FPS.

GPU Role

The GPU in this combo is a NVIDIA GeForce GTX 1660 SUPER, built on the Turing architecture on a 12 nm TSMC process with a chip of TU116. It includes 1,408 shading units, 88 texture mapping units, and 48 ROPs. Its base clock is 1530 MHz, boost clock is 1785 MHz, and the memory runs at 1750 MHz with a 14 Gbps effective rate across a 192‑bit bus, achieving a bandwidth of 336.0 GB/s. The GPU has 6 GB of GDDR6 VRAM. These specifications put it in the mid-range section of the GeForce family (its series is listed as "GeForce 10", though the generation is GeForce 16).

In Thief, the GPU's performance characteristics are the primary variable when resolution is increased. The 6 GB frame buffer is sufficient for the game’s maps at 1080p and 1440p, but at 4K the large texture needs begin to test the VRAM capacity, and the increased memory pressure shows up in the measured FPS drop to sub‑30 averages. The effective memory bandwidth of 336 GB/s is likely not a hard bottleneck below 4K but becomes a limiting factor at the highest resolution, where access patterns become very heavy.

The benchmark scores for the GPU place it in a specific context. In the 3DMark Steel Nomad DX12 test, it scores 1,278, a moderate value. Its passmark G3D score is 12,699, and its Geekbench OpenCL is 5,‑949. The GPU's percentile versus all GPUs is 53%, meaning it sits exactly at the median of all tested GPUs. Looking at the nearest rivals provided, the GTX 1660 SUPER is extremely close to the NVIDIA GeForce RTX 3050 Ti Mobile (deltaPct of +0.4%—meaning the 1660 Super is 0.4% higher average score), the AMD Radeon RX 580 (+0.4%), and the AMD Radeon 740M (+0.9%). This suggests the GTX 1660 SUPER is more or less equal to these cards in overall compute ability, though the RTX 3050 Ti Mobile is a laptop chip, and the RX 580 is an older design. Interestingly, the Tesla M2090 is 0.7% lower in score, which is a compute-oriented card from NVIDIA.

When the game is paired with a CPU that has 12 cores and 20 threads running at up to 5.0 GHz, the GPU becomes the limiting piece in every resolution. The fact that at 1080p Low the FPS hits 120 while the CPU is far from maxed out suggests the CPU is not causing any frame stalls. The gaming data shows no CPU limiting at even the highest frame counts. It is purely the GPU that determines whether the game stays fluid, and at 4K Ultra the average FPS is a mere 27, clearly too low for smooth play.

How This Combo Ranks

Among all tested combinations for Thief, this particular CPU plus GPU yields a relative rank of 3179 out of a total of 3720 combos. That places the combo in the lower 84.6% of the set—only 3179 of 3720 other combinations are worse, meaning a large fraction of other tested setups perform better. The rank reflects the fact that the GTX 1660 SUPER is a mid-range GPU from 2019, paired with a strong 12th‑gen Intel processor. The CPU effortlessly feeds the GPU enough data to stay busy, but the graphics card's modest compute capability limits overall performance. In a game with heavy shader work and a fairly high baseline GPU requirement, this pairing ends up near the low end of results.

The combo rank data includes the value `comboRankInGame.rank = 3179` and `ofCombos = 3720`. That means that around 14.6% of tested combinations deliver lower frame averages than this one, or put another way, this system outperforms about 1 out of 7 of the alternatives. While that is is not flattering, it is important to note that the numbers are measured and thus reflect actual in‑game behavior. The CPU's strength does little to help in this case; because the frame rate is essentially dictated by the GPU, the ranking is essentially determined by graphics capabilities alone.

Measured FPS Breakdown

Detailed measured FPS values for each resolution and setting are given below. The table lists the exact numbers as recorded in the benchmark.

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

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

| 1920×1080 | High | 69 | – | – |

| 1920×1080 | Low | 120 | – | – |

| 1920×1080 | Medium | 91 | 77 | 105 |

| 1920×1080 | Ultra | 66 | – | – |

| 2560×1440 | High | 47 | – | – |

| 2560×1440 | Low | 82 | – | – |

| 2560×1440 | Medium | 62 | 53 | 72 |

| 2560×1440 | Ultra | 45 | – | – |

| 3840×2160 | High | 28 | – | – |

| 3840×2160 | Low | 48 | – | – |

| 3840×2160 | Medium | 37 | 31 | 42 |

| 3840×2160 | Ultra | 27 | – | – |

The averages show a clean monotonic step. At each preset, the frame rate climbs as settings are lowered. The largest jump occurs when going from High to Low at 4K—a difference of 20 FPS (from 28 to 48). At 1080p the difference is 54 FPS (120 vs 66). The presence of minimum and maximum values at only the Medium preset underscores the notion that quality preset also has an effect on frame time variation; the 1080p Medium range of 77–105 (a span of 28 FPS) is the widest of the three measured sets, indicating more noticeable stuttering at that combination than at 1440p or 4K, where the ranges are 19 and 11 FPS respectively.

The data can be used to estimate effective refresh rates: 1080p High is sustained at 69 FPS, aligning close to a 60 Hz monitor's refresh with occasional dips and peaks; 1440p, Medium at 62 FPS also lands in that comfortable range. For 4K users, a 30 FPS cap is roughly attainable only by using High or lower settings, while Ultra drops below that to 27 FPS, a place.

FAQ

Q: What average frame rate does the combo achieve at 1080p Ultra?

A: The measured average is 66 FPS.

Q: Is there any resolution where the framerate exceeds 100 FPS?

A: Yes, 1080p Low yields an average of 120 FPS. No other resolution exceeds 100 FPS (the next highest is 82 at 1440p Low).

Q: (No question)

Q: How does this GPU compare to a similar card?

A: The GTX 1660 SUPER’s nearest rival in average performance is the AMD Radeon RX 580, with a delta of 0.4% (the 1660 SUPER scores 12986 in its average benchmark, while the RX 580 scores 12928). The NVIDIA RTX 3050 Ti Mobile is also at 12940, a delta of 0.4%, so both are within 0.4% of the GTX.

Q: What is the CPU’s performance percentile among all CPUs tested?

A: The Core i7-12700KF ranks in the 85th percentile of all CPUs.

Q: What is the lowest a FPS measured for this combo, and at which setting?

A: The lowest average FPS is 27, recorded at 4K Ultra (3840×2160).

Q: How many total combinations does the ranking reference?

A: The page's combo rank is 3179 out of 3720 total tested combinations.

CPU Role

The CPU side of the equation is an Intel Core i7‑12700KF, a 12-core (20-thread) Alder Lake desktop processor with a base clock of 3.60 GHz and a boost of 5.00 GHz. It has 25 MB of shared L3 cache, 80 KB of L1 per core, and 1.25 MB of L2 per core, and supports both DDR4 and DDR5 memory in a dual-channel configuration. The chip is built on Intel’s 10 nm process and uses the Intel Socket 1700, with a TDP of 125 W. It was released on November 3, 2021.

The CPU’s benchmark numbers showcase its strong single‑threaded performance: Cinebench R20. Scores: R15 multicore 2906, single‑core 410; R20 multicore 12111, single‑core ; R23 multicore 28838, single‑core 4071. Geekbench 5 scores are 14367 multi and 2255 single. The 3DMark thread tests show a linear scaling from 4 threads (4011 points) to 20 threads (9983 points). In PassMark the overall multi‑thread result is 34092 and the single‑thread result is 3984. These scores imply a processor that can easily feed any current GPU, and the 12-core cluster ensures plenty of throughput for background tasks.

In the context of Thief, the CPU’s role is minimal unless the resolution is low enough that the GPU can push frame rates high. The data shows at 1080p Low the average FPS is 120, which is just below the threshold where a CPU bottleneck might appear. The i7‑12700K’s single‑core passmark score of #394 is roughly 1200 points above the default for many games, but in this title the engine's thread scaling is not challenged. The benchmarks nearest rivals (Intel Core i7‑13700T with an average score of 35403, 0.1% lower; Intel Core i5‑13600T with 35305, 0.2% higher) place it identically; the practical outcome is that the processor is a non‑factor in balancing this combo. Therefore, the measured FPS is entirely dictated by the GPU’s basic fill rate, memory bandwidth, and raster power.

The belief that adding more CPU cores helps in Thief is not proven by the dataset: the combo's ranking in the game is low because the GPU is the weak link. The high‑end processor only ensures there are no frame dips or CPU logs. Based purely on the FPS data available, at no resolution<|eot_id|>