Thief

The AMD Ryzen 5 5600G and NVIDIA GeForce RTX 5090 D pairing in Thief is a study in extreme imbalance, where a 6-core, 12-thread Zen 3 CPU with a base clock of 3.90 GHz and a boost of 4.40 GHz anchors a GPU that is fundamentally overqualified for the task. The data shows the CPU is the clear limiting factor in nearly every measured configuration, as its 16 MB of L3 cache and dual-channel DDR4 memory bandwidth of 51.2 GB/s are modest by modern standards. The RTX 5090 D, by contrast, is a Blackwell 2.0 behemoth with 21,760 shading units, 170 RT cores, and a boost clock of 2407 MHz, yet it cannot stretch its legs because the Ryzen 5 5600G struggles to feed it. The result is a benchmark set where frame rates scale more with resolution than with CPU headroom, a classic sign of a bottleneck, though the absolute numbers remain playable across the board.

CPU Role — cores, clocks, and how they relate to this game's results

The Ryzen 5 5600G is a 6-core, 12-thread processor built on the Zen 3 architecture, codenamed Cezanne, with a base clock of 3.90 GHz and a boost clock of 4.40 GHz. Its 16 MB of L3 cache and 65 W TDP are typical for a mainstream desktop chip, but its benchmark scores tell a more nuanced story. In single-threaded tests, it scores 872 in 3dmark_single_thread and 1409 in Cinebench R23 single-core, placing it in the 74th percentile of all CPUs. However, its multi-threaded results — 10,551 in Cinebench R23 multicore and 19,717 in Passmark multithread — are solid but unremarkable, sitting just 0.1% above the Intel Core i3-1220P and 0.4% below the Intel Core Ultra 7 155U in average benchmark score.

In Thief, a 2014 stealth title, the CPU’s role is to manage game logic, AI, and draw calls, and the data indicates it does so adequately but without headroom. The measured FPS at 1080p Low is 156 avg, which is the highest score in the entire dataset, but this is also the configuration where the GPU is most idle, suggesting the CPU is already near its ceiling. As resolution increases, the FPS drops, but the drop is not as steep as it would be if the GPU were the bottleneck — from 1080p Low to 4K Low, the average falls from 156 to 115, a 26% reduction. This modest scaling implies that the CPU is holding the frame rate back at lower resolutions, preventing the RTX 5090 D from fully expressing its power. The CPU’s 3dmark_max_threads score of 5455 and passmark_physics score of 900 are consistent with a chip that can handle the game’s threaded workloads, but its 51.2 GB/s memory bandwidth and Gen 3 PCIe lanes may introduce latency that the GPU cannot overcome.

GPU Role — VRAM, clocks, and how they relate to this game's results

The NVIDIA GeForce RTX 5090 D is a flagship GPU with 32 GB of GDDR7 memory on a 512-bit bus, delivering 1.79 TB/s of bandwidth. Its boost clock of 2407 MHz and 104.8 TFLOPS of FP32 compute are staggering, and its 92nd percentile ranking among all GPUs confirms its elite status. However, its nearest rivals in average benchmark score are the AMD Radeon RX 6650M XT (1.1% slower) and the RX 6850M XT (1.6% faster), which are mobile-class GPUs, highlighting that the RTX 5090 D’s raw compute is not always reflected in synthetic tests. In Thief, the GPU’s 32 GB VRAM is far beyond what the game requires, and its 176 ROPs and 680 TMUs are more than sufficient for a 2014 title’s modest texture and shading demands.

The GPU’s measured performance in Thief is paradoxically low relative to its capabilities. At 4K Ultra, it averages 87 FPS, which is impressive for any card but modest for a GPU with 104.8 TFLOPS. The data suggests the GPU is being starved by the CPU at lower resolutions — at 1080p Low, the RTX 5090 D averages 156 FPS, but this is only marginally higher than the 1440p Low score of 138 FPS, a 13% gain for a 56% reduction in pixel count. This indicates that the GPU is not the limiting factor at 1080p; instead, the CPU is preventing the GPU from pushing higher frame rates. At 4K, the GPU begins to take over as the bottleneck, with the 4K Ultra score of 87 FPS being 31% lower than the 1080p Ultra score of 126 FPS, a steep drop that reflects the increased pixel workload. The GPU’s 5 nm process and 92,200 million transistors are irrelevant to the game’s performance, as the bottleneck is architectural rather than computational.

Resolution Scaling — how FPS drops from 1080p to 4K and what it says about the limiting component

The FPS scaling across resolutions provides the clearest evidence of the CPU bottleneck. From 1080p Low (156 avg) to 1440p Low (138 avg), the drop is 11.5%, and from 1440p Low to 4K Low (115 avg), the drop is 16.7%. This is a shallow curve for a GPU with 1.79 TB/s of bandwidth, which would typically show a much steeper decline if it were the limiting factor. In contrast, the Ultra settings tell a different story: 1080p Ultra (126 avg) to 1440p Ultra (111 avg) is a 11.9% drop, but 1440p Ultra to 4K Ultra (87 avg) is a 21.6% drop. The steeper decline at Ultra settings across higher resolutions suggests that the GPU is becoming more engaged as the pixel count increases, but even at 4K Ultra, the 87 FPS average is well below what the RTX 5090 D could achieve in a CPU-unconstrained scenario.

The data reveals that at 1080p and 1440p, the CPU is the primary limiter. The Ryzen 5 5600G’s 6 cores and 12 threads, with a passmark_single_thread score of 3177, are sufficient for the game’s logic, but the frame rates plateau around 126-156 FPS regardless of settings, indicating that the CPU cannot push beyond this threshold. At 4K, the GPU begins to dominate, but the 4K Low score of 115 FPS is only 26% lower than the 1080p Low score, while the 4K Ultra score of 87 FPS is 31% lower than the 1080p Ultra score. This asymmetry suggests that at Low settings, the CPU is still the bottleneck even at 4K, while at Ultra settings, the GPU’s increased workload starts to matter. The takeaway is that this combo is CPU-bound in almost all scenarios, with the GPU only becoming relevant at the highest resolution and settings.

Measured FPS Breakdown — resolution by resolution, settings by settings, exact numbers

The measured FPS data provides a complete picture of performance across three resolutions and four settings levels. At 1920x1080, the Low preset delivers the highest average at 156 FPS, followed by Medium at 141 FPS (with a min of 120 and max of 162), High at 128 FPS, and Ultra at 126 FPS. The narrow gap between High and Ultra (only 2 FPS) suggests that the CPU is the limiting factor at these settings, as the GPU’s extra work on Ultra does not translate to a significant frame rate penalty. The Medium preset’s min FPS of 120 is notably higher than the Ultra’s average, indicating smoother frame pacing.

At 2560x1440, the pattern repeats with Low at 138 FPS, Medium at 124 FPS (min 106, max 143), High at 113 FPS, and Ultra at 111 FPS. The drop from 1080p to 1440p is consistent across settings, averaging around 11-13%, which is modest and again points to a CPU bottleneck. The Medium preset’s min FPS of 106 is the lowest recorded across all configurations, suggesting occasional frame dips that may be tied to CPU load spikes. At 3840x2160, the scores are 115 FPS for Low, 104 FPS for Medium (min 89, max 120), 92 FPS for High, and 87 FPS for Ultra. The 4K Ultra score of 87 FPS is the lowest in the dataset, but it is still well above playable thresholds, and the 4K Low score of 115 FPS demonstrates that even at 4K, the CPU can sustain high frame rates when the GPU is not heavily taxed.

The data shows a clear hierarchy: Low > Medium > High > Ultra at every resolution, with the gaps between High and Ultra being minimal (2 FPS at 1080p, 2 FPS at 1440p, and 5 FPS at 4K). The Medium preset offers the best frame time consistency, as evidenced by its min/max values being the only ones recorded, but the overall spread across settings is narrow, reinforcing the CPU-bound nature of this combo.

How This Combo Ranks — use comboRankInGame vs other tested combos

In the global ranking of tested CPU-GPU combinations for Thief, this pairing of the AMD Ryzen 5 5600G and NVIDIA GeForce RTX 5090 D ranks 550th out of 2,851 combos, placing it in the top 19.3% of all tested systems. This rank is surprisingly low for a GPU that sits in the 92nd percentile of all GPUs, but it is inflated by the CPU’s modest 74th percentile ranking. The combo’s performance is heavily influenced by the CPU’s limitations, as the RTX 5090 D’s capabilities are only partially realized. The nearest rival CPUs to the Ryzen 5 5600G — the Intel Core i3-1220P (0.1% faster), Intel Core Ultra 7 256V (0.1% slower), and AMD Ryzen 5 7530U (0.2% slower) — all have similar average benchmark scores, suggesting that swapping the CPU would not yield significant gains.

The rank of 550 out of 2,851 indicates that there are 549 combos that outperform this pairing, many of which likely feature stronger CPUs paired with mid-range GPUs. This highlights the mismatch: the RTX 5090 D is a top-tier GPU that is being held back by a mid-range CPU. In Thief, where the CPU is the primary bottleneck, the combo’s rank is a direct reflection of the CPU’s performance, not the GPU’s. If the GPU were paired with a higher-end CPU, the rank would likely be in the top 100, but as configured, it sits in the upper-middle tier. The data suggests that this combo is best suited for users who prioritize GPU features like ray tracing or VRAM capacity over raw frame rates in CPU-bound titles like Thief.

Settings Recommendations — which preset gives the best experience per the measured rows

Given the measured FPS data, the Medium preset emerges as the optimal choice for balancing performance and visual quality in Thief with this combo. At 1080p, Medium delivers 141 FPS avg with a min of 120 and max of 162, which is 26% faster than Ultra (126 FPS) and only 10% slower than Low (156 FPS). The Medium preset’s min FPS of 120 is the highest minimum recorded at any resolution, indicating stable frame delivery that avoids the dips that might occur at higher settings. At 1440p, Medium averages 124 FPS (min 106, max 143), which is 11.7% higher than Ultra’s 111 FPS and 10.1% lower than Low’s 138 FPS. This makes Medium the best middle ground, as it provides a smooth experience without sacrificing too much visual fidelity.

At 4K, the Medium preset averages 104 FPS (min 89, max 120), which is 19.5% faster than Ultra’s 87 FPS and 9.5% slower than Low’s 115 FPS. The 4K Medium score of 104 FPS is well above the 60 FPS threshold for smooth gameplay, and the min FPS of 89 ensures that even the most demanding scenes remain fluid. The High preset, while visually superior to Medium, offers only marginal gains over Ultra (e.g., 92 FPS vs 87 FPS at 4K), making it a poor choice for users seeking the best experience per the measured rows. The Low preset, while delivering the highest frame rates, sacrifices too much visual quality for a game that can run at Medium with ease.

The data indicates that Ultra is not worth the performance cost, as it only provides a 2-5 FPS improvement over High at all resolutions, while Medium offers a significant frame rate boost with minimal visual degradation. For users with 1080p or 1440p displays, Medium is the clear recommendation, as it delivers frame rates between 124 and 141 FPS with excellent frame pacing. For 4K users, Medium is also recommended, as the 104 FPS average is more than sufficient for a stealth game, and the min FPS of 89 ensures no noticeable stutter. In all cases, the CPU bottleneck means that higher settings do not proportionally reduce performance, so users should not fear pushing to Medium, but Ultra remains a luxury that the CPU cannot fully support.