Thief

# Thief Performance Analysis: Intel Core Ultra 9 290K Plus + NVIDIA GeForce RTX 2060

This combo ranks 2251st out of 2851 tested configurations in Thief, placing it in the lower third of the database — a position driven almost entirely by the GPU's modest standing rather than the CPU's top-tier compute strength. The measured frame rates across resolutions confirm that the RTX 2060, which sits at the 58th percentile among all GPUs, is the decisive bottleneck at every quality setting tested.

Resolution Scaling

The frame rate drop from 1080p to 4K is steep and consistent across all four quality presets, with the Low setting showing the largest absolute decline. At Low, the combo delivers 142 FPS at 1920x1080, drops to 97 FPS at 2560x1440, and falls further to 57 FPS at 3840x2160 — a total reduction of 85 FPS, or roughly 60% of the 1080p figure. This pattern indicates that even at the lowest settings, the GPU's pixel throughput becomes the limiting factor as resolution climbs; the CPU, with 24 cores and a boost clock of 5.80 GHz, is never the constraint here.

The Medium preset shows a similar trajectory: 108 FPS at 1080p, 74 FPS at 1440p, and 43 FPS at 4K. The gap between 1080p and 1440p is 34 FPS, while the gap from 1440p to 4K is 31 FPS — nearly linear scaling, which suggests the rendering workload scales predictably with pixel count and that the RTX 2060's 6 GB of GDDR6 memory and 192-bit bus are being taxed evenly across these resolutions.

High settings amplify the resolution penalty. At 1080p, the combo reaches 82 FPS; at 1440p, it drops to 56 FPS; at 4K, it manages only 33 FPS. The percentage loss from 1080p to 4K is about 60%, mirroring the Low preset's relative decline. Ultra settings follow the same shape: 78 FPS at 1080p, 54 FPS at 1440p, and 31 FPS at 4K. The Ultra-to-High difference at 4K is just 2 FPS (31 vs 33), which indicates that the GPU is fully saturated at that resolution regardless of whether the preset is High or Ultra — additional quality settings cost almost nothing because the renderer is already bound by pixel fill and memory bandwidth.

The 4K results are the clearest evidence of GPU-limited scaling. Every preset at 3840x2160 lands between 31 and 57 FPS, with the spread between Low and Ultra being only 26 FPS. At 1080p, the spread between Low and Ultra is 64 FPS (142 vs 78), showing that the GPU has headroom to absorb quality increases at lower pixel counts but exhausts that headroom quickly as resolution rises. The data supports a firm conclusion: resolution scaling in Thief is governed by the RTX 2060's capabilities, and the Core Ultra 9 290K Plus contributes no measurable bottleneck at any tested resolution.

How This Combo Ranks

With a rank of 2251 out of 2851 combos, this pairing sits in the bottom 21% of the database for Thief. That placement is notable because the CPU alone is exceptional: the Core Ultra 9 290K Plus scores 84,003 on average across all benchmarks, placing it in the 96th percentile of all CPUs and slightly ahead of the Intel Core Ultra 9 285K (83,807, 0.2% delta) and the AMD EPYC 4584PX (83,090, 1.1% delta). The GPU, by contrast, manages an average benchmark score of 15,290 — the 58th percentile — and sits within 0.8% of the AMD Radeon RX 7600 (15,171) and within 0.6% of the NVIDIA GeForce RTX 3050 OEM (15,199).

The rank in Thief reflects the game's reliance on the GPU. A combo with a top-4% CPU and a mid-range GPU cannot overcome the graphics card's ceiling in a title that scales primarily with pixel and shader work. The delta between the CPU's percentile (96) and the combo's in-game rank (2251 out of 2851) is stark, quantifying how much the RTX 2060 holds back the system. The nearest GPU rival, the NVIDIA GeForce GTX 580, scores 15,283 — essentially tied with the RTX 2060's 15,290 — and would likely produce nearly identical Thief results, reinforcing that the combo's rank is a function of GPU class rather than any specific architectural advantage.

For context, the CPU's nearest rivals are all high-core-count server or flagship desktop parts, with the AMD EPYC 7F72 scoring 85,072 (1.3% higher). None of those rivals would change the Thief outcome because the game never approaches CPU limits here. The data makes clear that upgrading the GPU would move the combo rank far more than any CPU swap could.

Settings Recommendations

The measured rows point to Medium as the optimal balance for this combo. At 1920x1080, Medium delivers 108 FPS average with a minimum of 91 FPS and a maximum of 124 FPS — a range that stays comfortably above 60 FPS even in dips. High at 1080p drops to 82 FPS, which is still playable but sacrifices 26 FPS for quality gains that may be marginal in Thief's stealth-focused scenes. Ultra at 1080p (78 FPS) costs another 4 FPS over High, offering little reason to choose it over Medium unless visual fidelity is paramount.

At 2560x1440, Medium produces 74 FPS average with a minimum of 63 FPS and a maximum of 85 FPS. That minimum is just above the 60 FPS threshold, making Medium the highest preset that reliably holds playable frame rates at this resolution. High at 1440p (56 FPS) drops below 60, and Ultra (54 FPS) is only marginally worse. For 1440p users, Medium is the clear recommendation: it provides the best experience per the measured rows, with no preset above it maintaining a 60 FPS minimum.

At 3840x2160, no preset achieves a 60 FPS average. Low is the only option above 50 FPS, averaging 57 FPS, while Medium drops to 43 FPS, High to 33 FPS, and Ultra to 31 FPS. The gap between Low and Medium at 4K is 14 FPS, which is substantial, but the Low preset's 57 FPS average is the only playable result for 4K users who prioritize smoothness over visuals. The Medium preset at 4K does report a minimum of 37 FPS and a maximum of 50 FPS, indicating noticeable frame variability that could detract from the stealth gameplay experience.

The recommendation hierarchy is straightforward: Medium at 1080p or 1440p, Low at 4K. High and Ultra should be reserved for 1080p only, where the frame rate penalty over Medium is 26 and 30 FPS respectively, and even then the gains are not supported by the data as worth the trade-off.

Measured FPS Breakdown

At 1920x1080, the combo produces the following averages: Low at 142 FPS, Medium at 108 FPS (minimum 91, maximum 124), High at 82 FPS, and Ultra at 78 FPS. The differential between Low and Medium is 34 FPS, between Medium and High is 26 FPS, and between High and Ultra is 4 FPS. This compression at the top end suggests diminishing returns from Medium upward — the RTX 2060's 6 GB memory and 201.6 GTexel/s texture rate are sufficient for Low and Medium, but High and Ultra begin to strain its 1920 shading units.

At 2560x1440, the averages are: Low at 97 FPS, Medium at 74 FPS (minimum 63, maximum 85), High at 56 FPS, and Ultra at 54 FPS. The step from Low to Medium is 23 FPS, from Medium to High is 18 FPS, and from High to Ultra is 2 FPS. The near-identical High and Ultra results at 1440p (56 vs 54) mirror the pattern seen at 4K, indicating that the GPU's compute resources are fully engaged by the High preset and Ultra adds little overhead.

At 3840x2160, the averages are: Low at 57 FPS, Medium at 43 FPS (minimum 37, maximum 50), High at 33 FPS, and Ultra at 31 FPS. The Low-to-Medium step is 14 FPS, Medium-to-High is 10 FPS, and High-to-Ultra is 2 FPS. The 4K results are uniformly below 60 FPS, with only Low approaching playability. The minimum frame rate at Medium (37 FPS) is notably lower than the average, suggesting occasional stutter in demanding scenes.

Across all resolutions, the pattern is consistent: Low offers the highest headroom, Medium provides a balanced experience, and High/Ultra converge as resolution increases. The measured data includes no minimum or maximum values for Low, High, or Ultra presets, so the analysis of frame stability relies on the Medium preset's reported range, which shows a 27 FPS spread at 1080p (91 to 124), a 22 FPS spread at 1440p (63 to 85), and a 13 FPS spread at 4K (37 to 50). This narrowing spread at higher resolutions further confirms GPU saturation.

GPU Role

The NVIDIA GeForce RTX 2060 is the defining component in this combo's Thief performance. Its 6 GB of GDDR6 memory on a 192-bit bus provides 336.0 GB/s of bandwidth, which is adequate for 1080p and 1440p at Medium settings but becomes a limiting factor at 4K, where the memory bus is forced to feed more pixels per frame. The GPU's boost clock of 1680 MHz, with a base of 1365 MHz, drives 1920 shading units and 48 ROPs, yielding a pixel rate of 80.64 GPixel/s and a texture rate of 201.6 GTexel/s. These figures explain the steep drop from 1080p to 4K: the ROP count and memory bandwidth are simply not sufficient for 3840x2160 at high quality settings.

The RTX 2060's benchmark profile places it at the 58th percentile of all GPUs, with an average score of 15,290. Its closest rivals — the GTX 580 (15,283), the Radeon 680M (15,270), and the RTX 3050 OEM (15,199) — are all within 0.8% of its score, meaning that in Thief, swapping to any of these cards would produce nearly identical frame rates. The GPU's DirectX 12 Ultimate support and Vulkan 1.4 compatibility are relevant to modern titles, but Thief, released in 2014, does not stress these APIs heavily; the performance ceiling is set by raw throughput rather than feature support.

The GPU's 30 RT cores and 240 tensor cores are present on the silicon but unused in Thief, which has no ray tracing or DLSS implementation in the measured data. The 12 nm process node and 10,800 million transistors on a 445 mm² die result in a 160 W TDP, but the power draw is not a factor in the frame rate analysis. The 6 GB memory capacity is the key constraint: at 4K Ultra, the game averages 31 FPS, and the lack of headroom suggests that memory pressure, not just compute, is limiting performance. The 1750 MHz memory clock (14 Gbps effective) provides the 336.0 GB/s bandwidth, which is modest by modern standards and explains why the combo cannot sustain 60 FPS at 4K on any preset.

The GPU's end-of-life production status and launch MSRP of 349 USD are historical facts, but the measured performance in Thief reflects its mid-range positioning. The data shows that the RTX 2060 is the bottleneck in every scenario tested, with the CPU never contributing to frame rate variance.

FAQ

Q: Why does this combo rank so low in Thief despite having a top-tier CPU?

A: The combo ranks 2251st out of 2851 tested configurations. The Intel Core Ultra 9 290K Plus scores in the 96th percentile of all CPUs, but the NVIDIA GeForce RTX 2060 is in the 58th percentile of GPUs. Thief's performance is GPU-bound, so the CPU's strength does not translate into a high in-game rank.

Q: What is the best preset to use at 2560x1440?

A: Medium is the recommended preset at 1440p. It averages 74 FPS with a minimum of 63 FPS and a maximum of 85 FPS. The High preset drops to 56 FPS and Ultra to 54 FPS, both below the 60 FPS threshold that Medium maintains.

Q: Can this combo run Thief at 4K with smooth frame rates?

A: No preset achieves 60 FPS at 3840x2160. Low averages 57 FPS, Medium 43 FPS, High 33 FPS, and Ultra 31 FPS. Only Low approaches playability, and even it falls short of a consistent 60 FPS.

Q: How does the RTX 2060 compare to its nearest rivals in terms of benchmark scores?

A: The RTX 2060 averages 15,290 across benchmarks. The NVIDIA GeForce GTX 580 scores 15,283 (0% delta), the AMD Radeon 680M scores 15,270 (0.1% delta), and the NVIDIA GeForce RTX 3050 OEM scores 15,199 (0.6% delta). All are effectively tied, so Thief performance would be similar with any of these GPUs.

Q: What is the frame rate difference between High and Ultra settings at 1080p?

A: At 1920x1080, High averages 82 FPS and Ultra averages 78 FPS, a difference of just 4 FPS. The minimal gap suggests the GPU is nearing its limits at these presets, with diminishing returns from further quality increases.

Q: Does the CPU bottleneck the GPU in this game?

A: The data shows no evidence of CPU bottlenecking. The Core Ultra 9 290K Plus has 24 cores, a 5.80 GHz boost clock, and scores 84,003 average across benchmarks. Frame rates scale directly with resolution and settings in a pattern consistent with GPU limitation, not CPU limitation.