Thief

# Thief: Intel Core i5-12400 + NVIDIA GeForce RTX 5090 D — Benchmark Analysis

This combination pairs a six-core Alder Lake processor with a flagship Blackwell GPU, producing frame rates that range from 87 FPS at 4K Ultra to 170 FPS at 1080p Low in Thief. The measured data shows a system that is overwhelmingly GPU-limited at lower resolutions, with the CPU's 4.40 GHz boost clock and 18 MB of shared L3 cache proving sufficient to feed the RTX 5090 D across all tested settings.

Resolution Scaling

The frame rate progression from 1080p to 4K reveals a classic GPU-bound scaling pattern. At High settings, the average FPS drops from 139 at 1920x1080 to 123 at 2560x1440, then falls further to 92 at 3840x2160. That represents a 33.8% reduction from 1080p to 4K, which is substantial but not as severe as what would be expected if the GPU were the sole bottleneck across the entire range.

Looking at Low settings, the scaling is more aggressive: 170 FPS at 1080p drops to 150 FPS at 1440p (an 11.8% decrease), and then to 126 FPS at 4K (a 16% decrease from 1440p). The 1080p to 1440p gap at Low settings is notably smaller than the 1440p to 4K gap, indicating that at 1080p Low, the CPU is beginning to impose a ceiling. The i5-12400's single-thread score of 910 in 3dmark and 2257 in Cinebench R23 suggests it can sustain roughly 170 FPS in this title, beyond which the RTX 5090 D would need more processor headroom.

At Medium settings, which include the only min/max FPS measurements in the dataset, the scaling is smooth: 154 FPS at 1080p, 136 FPS at 1440p, and 114 FPS at 4K. The min FPS values at 4K Medium (96) versus 1440p Medium (115) show a 19 FPS difference, while the max values differ by 25 FPS (131 vs 156). This consistency suggests the frame pacing remains stable as resolution increases, with no sudden stutters appearing at higher pixel counts.

Ultra settings produce the tightest resolution spread: 137 FPS at 1080p, 121 FPS at 1440p, and 87 FPS at 4K. The 1080p to 1440p delta of 16 FPS is smaller than the 1440p to 4K delta of 34 FPS, which aligns with the GPU's compute load growing super-linearly with pixel count. The 4K Ultra result of 87 FPS sits just below the 90 FPS threshold that many users consider the minimum for smooth gameplay, but it remains playable.

The data indicates that at 1080p, the CPU is the limiting component in Low and possibly Medium settings, while at 4K, the RTX 5090 D's raw compute — 104.8 TFLOPS FP32 and 1,636.8 GTexel/s texture rate — becomes the dominant factor. The pixel rate of 423.6 GPixel/s is more than sufficient for 4K, suggesting the game engine's draw call overhead, not pixel throughput, is what caps performance at the highest resolution.

GPU Role

The NVIDIA GeForce RTX 5090 D carries 32 GB of GDDR7 memory on a 512-bit bus, yielding 1.79 TB/s of bandwidth. In Thief, a 2014 title, this memory configuration is vastly overprovisioned; the game's textures and geometry from that era would fit comfortably within even a small fraction of this capacity. The benchmark results confirm that memory capacity is not a limiting factor at any resolution or setting.

The GPU's clock behavior shows a 2017 MHz base and 2407 MHz boost. The measured FPS at 4K Ultra (87) versus 4K High (92) suggests that the boost clock is being sustained under load, as the 5 FPS difference between these settings is modest relative to the graphical complexity increase. The GPU's 92nd percentile ranking among all GPUs, with an average benchmark score of 77712, places it ahead of the AMD Radeon RX 6850M XT by 1.6% and behind the NVIDIA Tesla P100 PCIe 16 GB by 2.4% in synthetic tests, though those comparisons are for compute workloads rather than gaming.

At 1080p Low, the RTX 5090 D produces 170 FPS, which is likely near the CPU's practical limit for this game. The GPU's 32 GB frame buffer and 512-bit memory interface are irrelevant at this resolution, but the architecture's geometry processing and rasterization throughput still matter. The 176 ROPs and 21760 shading units deliver the pixel fill and shader work needed to maintain high frame rates even when the CPU is the primary constraint.

The transition from Medium to Ultra at 4K costs 27 FPS (114 to 87), which is a 23.7% performance penalty. This suggests that Ultra settings introduce effects that heavily tax the GPU's shading units and texture units — the 680 TMUs and 680 tensor cores are engaged more aggressively, though Thief predates widespread tensor core usage. The 1.79 TB/s bandwidth ensures that texture streaming at 4K Ultra does not cause frame drops, as evidenced by the min FPS of 96 at 4K Medium being only 18 FPS below the average.

Settings Recommendations

Based on the measured rows, the Medium preset offers the best balance of frame rate and visual fidelity. At 1080p Medium, the system delivers 154 FPS average with a min of 131 and max of 177, which provides a 23 FPS headroom above the average — a stable experience. At 1440p Medium, the 136 FPS average with 115 min and 156 max maintains that consistency, making it the recommended choice for high-refresh-rate monitors.

For 4K gaming, Medium is again the sensible pick. The 114 FPS average at 3840x2160 Medium is 22 FPS higher than High's 92 and 27 FPS higher than Ultra's 87. The min FPS of 96 at 4K Medium ensures that even the heaviest scenes remain above 90 FPS, whereas High and Ultra lack measured min values, suggesting the average could dip lower in demanding moments.

Ultra settings deliver only marginal gains over High at 1080p (137 vs 139 FPS, actually 2 FPS slower) and 1440p (121 vs 123 FPS, also 2 FPS slower). This counterintuitive result indicates that Ultra settings in Thief may introduce CPU-side overhead or inefficient rendering paths that offset any visual improvements. At 4K, Ultra's 87 FPS is 5 FPS below High, a small absolute difference but one that pushes the experience below the 90 FPS comfort threshold.

Low settings are only recommended for competitive play or extremely high refresh rates. The 170 FPS at 1080p Low is the highest measured result, but the jump from Low to Medium at 1080p (170 to 154 FPS) costs only 16 FPS while likely providing substantially better visual quality. The same pattern holds at 1440p (150 to 136 FPS) and 4K (126 to 114 FPS), where the 12 FPS difference between Low and Medium is a reasonable trade for improved fidelity.

Measured FPS Breakdown

At 1920x1080, the full settings range was tested. Low produces 170 FPS average, Medium 154 FPS (min 131, max 177), High 139 FPS, and Ultra 137 FPS. The gap between Low and Medium is 16 FPS, while the gap between Medium and High is 15 FPS. High and Ultra are nearly identical, differing by just 2 FPS. The min-to-max spread at Medium (46 FPS) shows significant frame variability, indicating that some scenes are much lighter than others.

At 2560x1440, the results are 150 FPS on Low, 136 FPS on Medium (min 115, max 156), 123 FPS on High, and 121 FPS on Ultra. The Low-to-Medium delta narrows to 14 FPS, while Medium-to-High costs 13 FPS. The Medium min-to-max spread is 41 FPS, slightly tighter than at 1080p. High and Ultra remain within 2 FPS of each other, reinforcing the pattern that Ultra provides no measurable benefit over High at this resolution.

At 3840x2160, the scaling becomes steeper. Low yields 126 FPS, Medium 114 FPS (min 96, max 131), High 92 FPS, and Ultra 87 FPS. The Low-to-Medium gap is 12 FPS, but Medium-to-High jumps to 22 FPS — the largest single-setting delta across all resolutions. High-to-Ultra costs 5 FPS. The Medium min FPS of 96 is still above the 90 FPS threshold, but the 35 FPS min-to-max spread suggests that 4K scenes vary considerably in complexity.

Comparing across resolutions at Medium settings: 1080p to 1440p costs 18 FPS, 1440p to 4K costs 22 FPS, and 1080p to 4K costs 40 FPS total (154 to 114). The scaling is nearly linear, with each resolution step costing roughly 12-14% of the previous frame rate. At Ultra settings, the same comparison shows 16 FPS lost from 1080p to 1440p and 34 FPS lost from 1440p to 4K — a nonlinear increase that reflects the GPU's compute limits.

FAQ

Q: Is the RTX 5090 D overkill for Thief?

A: At 1080p Low, the 170 FPS average is likely CPU-limited, meaning the i5-12400 cannot push higher frame rates. At 4K Ultra, the 87 FPS average shows the GPU is fully utilized, but the 32 GB memory and 1.79 TB/s bandwidth are far beyond what this 2014 game needs.

Q: Why does Ultra run slower than High at 1080p and 1440p?

A: At 1080p, Ultra averages 137 FPS versus High's 139 FPS; at 1440p, Ultra averages 121 FPS versus High's 123 FPS. This 2 FPS difference suggests Ultra settings introduce CPU overhead or inefficient rendering paths that slightly degrade performance without visual payoff.

Q: What is the best resolution for this combo?

A: The data shows 1440p Medium provides 136 FPS average with a 115 min, which is smooth and visually detailed. 4K Medium at 114 FPS is also viable if the display supports it, but 1440p offers more headroom above the 90 FPS comfort threshold.

Q: How does the i5-12400's single-thread performance affect Thief?

A: The CPU's 3dmark single-thread score of 910 and Cinebench R23 single-core score of 2257 correlate with the 170 FPS ceiling observed at 1080p Low. Below this threshold, the GPU is the limiter; above it, the CPU cannot keep up.

Q: Are the min FPS values reliable for judging frame pacing?

A: Only Medium settings include min/max measurements. At 4K Medium, the min is 96 FPS (84% of the 114 average), and at 1080p Medium, the min is 131 FPS (85% of the 154 average). These ratios indicate consistent frame delivery without severe stutters.

Q: Does the RTX 5090 D's 92nd percentile GPU ranking reflect its Thief performance?

A: The synthetic average benchmark score of 77712 places it ahead of the RX 6850M XT by 1.6%, but in Thief, the GPU's real-world results are more influenced by the CPU and game engine than by raw compute scores.

CPU Role

The Intel Core i5-12400 provides six cores and twelve threads with a base clock of 2.50 GHz and boost clock of 4.40 GHz. Its Cinebench R23 multicore score of 15989 and single-core score of 2257 indicate strong per-thread performance for a 2022 processor. In Thief, the game's 2014 engine likely uses fewer threads, making the single-core capabilities more relevant than the multicore throughput.

The 18 MB shared L3 cache and 80 KB per-core L1 cache provide adequate working set for the game's data. The CPU's 3dmark scores show scaling from 910 (single-thread) to 5890 (max threads), a 6.5x improvement that suggests good multi-threading efficiency, though Thief may not exploit all twelve threads. The Passmark single-thread score of 3456 and multithread score of 18747 reinforce that this CPU has balanced capabilities.

At 1080p Low, the 170 FPS result is likely near the CPU's limit. The 3dmark 2-thread score of 1692 and 4-thread score of 3012 suggest that the game's main thread and a few helper threads are well-served by the Alder Lake architecture's 10 nm process and 4.40 GHz boost. The CPU's 72nd percentile ranking among all CPUs, with an average benchmark score of 18683, places it slightly below the AMD EPYC 7643 (by 0.1%) and Intel Core i7-1355U (by 0.3%), but ahead of the Core i3-14100F by 0.9%.

The transition from 1080p Low (170 FPS) to 1080p Medium (154 FPS) costs 16 FPS, which is a 9.4% reduction. This drop is smaller than the CPU's single-thread headroom would suggest, indicating that the GPU begins to take over as the limiter even at Medium settings. At 4K Low (126 FPS), the CPU is no longer the bottleneck, as the GPU's workload dominates.

The i5-12400's 65 W TDP and dual-channel memory support (DDR4 or DDR5) allow it to sustain boost clocks without thermal throttling, which is critical for maintaining the 170 FPS ceiling. The CPU's PCIe Gen 5 interface, with 16 lanes, provides ample bandwidth to the RTX 5090 D's PCIe 5.0 x16 bus interface, ensuring no data transfer bottleneck.

How This Combo Ranks

The combination of Intel Core i5-12400 and NVIDIA GeForce RTX 5090 D ranks 321st out of 2851 tested combos in Thief, placing it in the top 11.3% of all configurations. This high ranking is primarily driven by the GPU's performance, as the CPU's 72nd percentile ranking is below the GPU's 92nd percentile — the GPU is the stronger component in this pairing.

The combo rank of 321 out of 2851 means that approximately 2530 other combinations score higher in Thief. Given that the RTX 5090 D is among the fastest GPUs available (92nd percentile), the combos ranking above this one likely pair it with faster CPUs or use configurations that avoid the CPU bottleneck observed at 1080p. The i5-12400's nearest rivals — the AMD EPYC 7643, Intel Core i7-1355U, AMD Ryzen AI 5 330, and Intel Core i3-14100F — all score within 0.9% of the i5-12400's average benchmark score, suggesting that swapping to any of these CPUs would not dramatically alter the combo's Thief ranking.

The data indicates that this combo's performance is strong at 4K, where the GPU dominates and the CPU's limitations are masked. At 1080p, the combo's ranking would likely be lower due to the CPU ceiling, but the overall rank of 321 reflects performance across all tested scenarios. The 2851 total combos include both weaker and stronger pairings, and this configuration sits comfortably in the upper decile, making it a capable system for Thief despite the age of the game engine.