Thief

# Thief Performance Analysis: AMD Ryzen 7 5700X3D + NVIDIA GeForce RTX 4090

The combination of AMD's Ryzen 7 5700X3D and NVIDIA's GeForce RTX 4090 produces a fascinating dataset in Thief, a 2014 stealth title that remains demanding at modern resolutions. The measured FPS across three resolutions and four quality presets reveals a system that scales predictably with resolution but shows surprising behavior at 1440p Ultra, where the average FPS nearly matches the High preset. This suggests the CPU's 96 MB shared L3 cache and 8-core/16-thread configuration may be interacting with the game engine in ways that reward certain settings combinations, while the RTX 4090's 24 GB of GDDR6X memory and 1.01 TB/s bandwidth provide ample headroom for texture-heavy scenes. The data shows a system ranked 487 out of 1642 tested combos, placing it comfortably in the upper third of all configurations while still leaving room for optimization.

Resolution Scaling

The FPS drop from 1080p to 4K tells a clear story about where the bottleneck shifts. At Low settings, the average FPS falls from 144.9 at 1920x1080 to 109.2 at 3840x2160 — a reduction of 35.7 FPS, or roughly 24.6%. At High settings, the drop is steeper in percentage terms: from 120.4 at 1080p down to 86.4 at 4K, a 34 FPS decline that represents about 28.2% of the 1080p figure. This pattern indicates that the RTX 4090 is increasingly becoming the limiting component as resolution climbs, since pixel throughput demands grow faster than the CPU's frame generation capacity at lower resolutions.

However, the scaling is not perfectly uniform across presets. Medium settings show a drop from 129.8 at 1080p to 99.9 at 4K (29.9 FPS, or 23%), while Ultra settings decline from 118.2 to 80.2 (38 FPS, or 32.1%). The fact that Ultra loses a larger absolute FPS count than High when moving from 1080p to 4K suggests that the additional GPU workload from Ultra effects — likely related to shadow quality or post-processing — amplifies the resolution penalty. The 1440p data points reinforce this: at 2560x1440, High averages 103.1 FPS, while Ultra averages 103 FPS. This near-identical performance at 1440p implies the GPU is not yet saturated at this resolution with these settings, and the CPU's 3.00 GHz base clock with 4.10 GHz boost may be holding back the RTX 4090 from fully expressing its capability.

GPU Role

The NVIDIA GeForce RTX 4090 brings formidable specifications to Thief: 24 GB of GDDR6X memory on a 384-bit bus, delivering 1.01 TB/s of bandwidth. In a 2014 game, this memory configuration is almost certainly overkill for texture storage, yet the data shows measurable differences between presets at every resolution. At 4K, the spread from Low (109.2 FPS) to Ultra (80.2 FPS) spans 29 FPS, indicating that GPU compute load — not memory capacity — is the primary differentiator between quality levels.

The GPU's Ada Lovelace architecture, built on a 5 nm process with 76,300 million transistors, provides 16,384 shading units and 512 tensor cores. The 82.58 TFLOPS of FP32 performance is more than sufficient for Thief's DirectX 12 Ultimate rendering path, yet the game's age means it likely doesn't utilize ray tracing cores (128 RT cores) or the full feature set. The boost clock of 2520 MHz versus a base of 2235 MHz suggests thermal headroom is available, but the measured FPS at 1080p Low (144.9) versus 4K Ultra (80.2) shows the GPU scaling with resolution as expected. Interestingly, the GPU's average benchmark score of 66473 places it near the NVIDIA Tesla T4 (66733, deltaPct -0.4) and AMD Radeon Pro Vega 56 (67097, deltaPct -0.9) in synthetic tests, yet in Thief the RTX 4090's real-world performance clearly exceeds what those workstation-oriented rivals would deliver.

CPU Role

The AMD Ryzen 7 5700X3D uses Zen 3 architecture on a 7 nm process, with 8 cores and 16 threads running between 3.00 GHz base and 4.10 GHz boost. The defining feature is the 96 MB shared L3 cache, which likely benefits Thief's level streaming and object culling. The CPU's percentile rank of 82 against all CPUs, with an average benchmark score of 24673, places it just 0.2% ahead of the Intel Core i7-1360P and 0.3% ahead of the AMD Ryzen 5 7600X3D in synthetic averages. In gaming, the large L3 cache can reduce memory latency, which may explain why the 1440p Ultra result (103 FPS) matches the High preset (103.1 FPS) — the CPU may be delivering consistent frame times regardless of preset at this resolution.

The single-thread performance is modest: a 3DMark single-thread score of 804 and Cinebench R23 single-core score of 3157. Thief, released in 2014, likely relies heavily on one or two primary threads for game logic, which could create a ceiling. The 1080p Low result of 144.9 FPS versus 4K Low at 109.2 FPS suggests that even at 1080p, the CPU cannot push beyond roughly 145 FPS in this title, regardless of GPU capability. The passmark single-thread score of 2970 and multithread score of 26318 indicate balanced performance, but the 3DMark 2-thread score of 1575 relative to the 16-thread score of 6764 highlights that Thief's engine may not scale well across many cores. The CPU's 51.2 GB/s memory bandwidth with DDR4 support could also be a limiting factor for a game that streams large environments.

FAQ

Q: Why does the 1440p Ultra FPS (103) nearly match the 1440p High FPS (103.1)?

A: The data shows only a 0.1 FPS difference, suggesting the CPU or memory subsystem becomes the limiting factor at this resolution and settings combination, preventing the RTX 4090 from showing a larger gap.

Q: Is the RTX 4090's 24 GB VRAM necessary for Thief?

A: The measured results show no evidence of VRAM limitations — the GPU's 1.01 TB/s bandwidth and 24 GB capacity exceed any demands Thief can place on them, but the differences between presets at 4K (80.2 to 109.2 FPS) indicate compute load matters more than memory capacity.

Q: How does this CPU compare to its nearest rivals in synthetic benchmarks?

A: The Ryzen 7 5700X3D's average benchmark score of 24673 is 0.2% higher than the Intel Core i7-1360P (24627), 0.3% higher than the AMD Ryzen 5 7600X3D (24605), and 0.5% higher than the Intel Core i5-11600 (24559), but 0.6% lower than the AMD Ryzen AI Max 385 (24811).

Q: What is the fastest measured FPS in this dataset?

A: The highest average FPS recorded is 144.9 at 1920x1080 with Low settings, followed by 129.8 at Medium and 120.4 at High, indicating the CPU's ceiling around 145 FPS in this game.

Q: Does the 5700X3D's 96 MB L3 cache provide a measurable benefit in Thief?

A: The data cannot isolate cache effects directly, but the consistent performance at 1440p across High and Ultra presets (103.1 vs 103 FPS) hints that the CPU's cache hierarchy may be mitigating the performance difference those settings would normally create.

Q: How does this combo rank overall?

A: The combination ranks 487 out of 1642 tested combos in Thief, placing it in the top 29.7% of all configurations tested.

How This Combo Ranks

The rank of 487 out of 1642 combos places this system in the upper third of tested configurations, but the position warrants scrutiny. Given that the RTX 4090 is among the most powerful consumer GPUs ever made, and the 5700X3D is a strong gaming CPU with an 82nd percentile ranking against all CPUs, a rank of 487 suggests that Thief's engine does not fully reward this hardware combination. The RTX 4090's percentile of 91 against all GPUs, with an average benchmark score of 66473, indicates it outperforms 91% of GPUs in synthetic tests — yet the game's measured FPS at 4K Ultra (80.2) is not dramatically higher than what mid-range GPUs might achieve in this older title.

The gap between the CPU's synthetic performance (average benchmark score 24673, ranking near the Intel Core i7-1360P and AMD Ryzen 5 7600X3D) and the in-game results implies that Thief's engine is not CPU-bound in a way that leverages the 5700X3D's strengths. The 16 threads and 96 MB L3 cache are likely underutilized by a 2014 game engine, while the RTX 4090's raw power is similarly constrained. The 1440p results — where High and Ultra both average around 103 FPS — suggest a plateau where neither component can push further, likely due to engine-level limitations rather than hardware capability. For a combo ranked 487th, the data implies that many systems with less powerful GPUs but stronger single-thread performance could achieve similar results in Thief.

Settings Recommendations

The measured data provides clear guidance for maximizing the experience. At 1920x1080, Low settings deliver 144.9 FPS, but Medium (129.8) and High (120.4) both remain above 120 FPS, making High the preferable choice for visual fidelity without sacrificing smoothness. The jump to Ultra at 1080p yields 118.2 FPS — still playable, but the 2.2 FPS difference from High suggests the extra GPU load is not worth the marginal visual gain at this resolution.

At 2560x1440, the analysis becomes more nuanced. High averages 103.1 FPS, while Ultra averages 103 FPS — a negligible difference of 0.1 FPS. This makes Ultra the optimal choice at 1440p, as the data shows no performance penalty for the highest preset. Medium (113.4) and Low (128.9) offer higher frame rates, but the gap between High and Ultra is so small that there is no reason to sacrifice quality. At 3840x2160, High (86.4 FPS) provides the best balance between visual quality and performance, with Ultra (80.2) dropping below 85 FPS and Medium (99.9) offering only a modest gain over High. For players targeting 60 FPS as a minimum, all 4K presets exceed that threshold, but High appears to be the sweet spot where quality and frame rate align most favorably.

Measured FPS Breakdown

The complete dataset reveals consistent scaling across the twelve measured configurations. At 1920x1080, the FPS ranges from 144.9 at Low to 118.2 at Ultra, with Medium at 129.8 and High at 120.4. The spread of 26.7 FPS between Low and Ultra at 1080p indicates the GPU has significant headroom at this resolution, with the CPU likely capping performance near 145 FPS.

At 2560x1440, Low produces 128.9 FPS, Medium 113.4, High 103.1, and Ultra 103. The compression of the high-end presets — just 0.1 FPS difference between High and Ultra — is the most notable anomaly in the entire dataset. This suggests that at 1440p, the system reaches a performance plateau where additional GPU workload from Ultra settings does not translate into lower frame rates, possibly due to the CPU delivering frames at a consistent rate regardless of GPU load.

At 3840x2160, the results are 109.2 FPS at Low, 99.9 at Medium, 86.4 at High, and 80.2 at Ultra. The 29 FPS difference between Low and Ultra at 4K is the largest absolute spread across all resolutions, confirming that the GPU becomes the primary bottleneck at this resolution. The progression from Low to Ultra at 4K shows a nearly linear decline, with each preset step costing between 6.3 and 13.5 FPS, whereas at 1080p the steps are 14.4, 9.4, and 2.2 FPS respectively — a much less predictable pattern that points to CPU limitations at lower resolutions.