Thief

Thief, a 2014 stealth title, presents an interesting case study in hardware scaling when paired with the AMD Ryzen 5 3600 and the NVIDIA GeForce RTX 5090 D. The measured FPS data reveals a combination that is heavily constrained by the aging CPU, yet still capable of delivering high frame rates due to the sheer power of the modern GPU. The benchmark results indicate that this pairing sits at the 1214th rank out of 2851 tested combos, placing it in the upper-middle tier of performance for this specific title. The data suggests a system where the GPU is often waiting on the processor, a dynamic that becomes clearer as resolution and settings are adjusted.

Resolution Scaling

The FPS progression from 1080p to 4K tells a clear story about where the bottleneck lies in this system. At 1920x1080 with Low settings, the combo achieves an average of 126 FPS. Moving up to 2560x1440 at Low settings, the average drops to 111 FPS, a 11.9% decrease. Further scaling to 3840x2160 at Low settings yields 93 FPS, which is a 16.2% drop from the 1440p figure. This pattern of diminishing returns as resolution increases is a classic indicator of a CPU-bound scenario at lower resolutions. The GPU has ample headroom to render frames quickly at 1080p, but the CPU cannot feed it data fast enough to push the frame rate much higher.

The data becomes even more telling when examining the High settings preset across resolutions. At 1080p High, the average FPS is 103. At 1440p High, it falls to 91 FPS, a drop of 11.7%. At 4K High, the average is 76 FPS, which is a 16.5% reduction from the 1440p result. The consistency of these percentage drops—roughly 12% between 1080p and 1440p, and 16% between 1440p and 4K—suggests that while the GPU is becoming more involved at higher resolutions, the CPU is still limiting the absolute maximum frame rate. If the system were purely GPU-bound, the drop from 1080p to 4K would typically be much more severe, often exceeding 40-50%, because the pixel count quadruples.

However, the scaling is not entirely flat, which implies the GPU is not completely idle. The difference between the Low and Ultra presets at each resolution shows the GPU's workload is increasing. At 4K, the spread between Low (93 FPS) and Ultra (75 FPS) is 18 FPS, or a 19.4% performance penalty for maxing out settings. At 1080p, the spread between Low (126 FPS) and Ultra (101 FPS) is 25 FPS, a 19.8% penalty. The similar relative penalties across resolutions are curious. It suggests that while the absolute frame rate is capped by the CPU, the incremental cost of higher settings is a relatively constant burden on the GPU. This points to a system where the CPU sets a performance ceiling, but the GPU still dictates how close to that ceiling you can get based on the graphical load.

GPU Role

The NVIDIA GeForce RTX 5090 D is a monstrously powerful piece of hardware, but its role in this pairing is largely to provide a high baseline that the CPU cannot fully utilize. The GPU features 32 GB of GDDR7 memory on a 512-bit bus, providing a massive 1.79 TB/s of bandwidth. This memory configuration is overkill for Thief, a game from 2014 that likely does not require anywhere near that capacity or bandwidth. The data shows the GPU's 3DMark Steel Nomad DX12 score of 14326 and a Passmark G3D score of 44065, placing it in the 92nd percentile of all GPUs. Yet, in this game, it is often held back.

The GPU's clocks are listed as a base of 2017 MHz and a boost of 2407 MHz. These high clock speeds, combined with 21760 shading units, should theoretically allow for massive frame rates in an older title. The fact that the system only manages 126 FPS at 1080p Low reflects the CPU limitation. The RTX 5090 D's performance in this scenario is akin to a high-performance sports car stuck in city traffic; the potential is there, but the environment prevents it from being realized. The benchmark results show that the GPU is not the limiting factor, as its nearest rivals in the database—like the AMD Radeon RX 6850M XT with a 1.6% lower average score—would likely perform similarly in this scenario because the game engine is waiting on the processor.

Interestingly, the GPU's Passmark DirectX 9 score of 434 is notably higher than its DirectX 11 score of 371 and DirectX 12 score of 219. Thief, released in 2014, likely uses an API like DirectX 11. The high DirectX 9 score indicates legacy performance is strong, but the game's specific engine and API usage may not be able to leverage the newer architectural features of the Blackwell 2.0 design effectively. This is a case where the GPU's raw compute power, as evidenced by its 104.8 TFLOPS FP32 rating, cannot be fully translated into on-screen frames due to the software and CPU constraints. The GPU is a future-proofing monster, but for Thief, it is a component operating well below its potential.

How This Combo Ranks

The combo of the AMD Ryzen 5 3600 and the NVIDIA GeForce RTX 5090 D ranks 1214 out of 2851 tested combinations in Thief. This places it in the 57.4th percentile of all tested systems, meaning it outperforms a slight majority of other setups. This ranking is surprisingly low for a system containing one of the most powerful consumer GPUs available. The data implies that many other combinations, likely those with newer and faster CPUs paired with mid-range GPUs, achieve higher frame rates in this specific game.

The ranking suggests that the CPU is the primary drag on performance. The Ryzen 5 3600, while a capable processor, is from an older generation. Its 3DMark single-thread score of 696 and multi-thread score of 4603 are respectable, but they are not exceptional by modern standards. When paired with a GPU that is in the 92nd percentile, the mismatch is stark. The GPU is performing at a level that is significantly above the CPU's ability to keep up. This is a classic case of a bottleneck where the system's overall performance is defined by its weakest link, which is the processor.

The data shows that the RTX 5090 D's nearest rivals in the GPU benchmark database are mobile and data-center parts, not desktop gaming GPUs. This further highlights the performance class of the GPU. Yet, in Thief, this high-tier GPU is being held to the level of a mid-range system. The 1214th rank is a direct consequence of this imbalance. If the same GPU were paired with a top-tier, newer processor, the rank would likely be in the top 100 or even top 50. The current ranking confirms how important CPU-GPU balance is, even for older games that are not typically considered CPU-intensive.

Measured FPS Breakdown

The measured FPS data provides a complete picture of this combination's performance across different settings and resolutions. At 1920x1080, the average frame rates are as follows: Low settings produce 126 FPS, Medium produces 114 FPS, High produces 103 FPS, and Ultra produces 101 FPS. The Medium setting also has a recorded minimum of 97 FPS and a maximum of 131 FPS, indicating some frame time variance. The step from Medium to High is a 9.6% drop, while the step from High to Ultra is only a 1.9% drop. This suggests that the High to Ultra transition is not particularly demanding, but the Medium to High step has a more significant impact on performance.

At 2560x1440, the average frame rates are: Low at 111 FPS, Medium at 100 FPS, High at 91 FPS, and Ultra at 89 FPS. The Medium preset at 1440p has a minimum of 85 FPS and a maximum of 115 FPS. The scaling here is interesting: the drop from Low to Medium is 9.9%, from Medium to High is 9.0%, and from High to Ultra is 2.2%. The pattern is similar to 1080p, with the largest performance hit occurring when moving from Low to Medium, and a relatively small penalty for the final step to Ultra. This consistency across resolutions reinforces the idea that the CPU is setting a performance ceiling, and the settings are just adjusting the GPU load within that ceiling.

At 3840x2160, the average frame rates are: Low at 93 FPS, Medium at 84 FPS, High at 76 FPS, and Ultra at 75 FPS. The Medium preset at 4K has a minimum of 71 FPS and a maximum of 97 FPS. The drop from Low to Medium is 9.7%, from Medium to High is 9.5%, and from High to Ultra is 1.3%. At 4K, the system still maintains an average above 60 FPS on all presets, which is a positive result. However, the frame rate is clearly capped, as even at Low settings, the system cannot break past 100 FPS. The maximum FPS recorded on the Medium preset at 4K is 97, which is lower than the average at 1080p Low. This shows that the resolution is starting to tax the GPU more, but the CPU is still preventing the frame rate from rising further.

CPU Role

The AMD Ryzen 5 3600 is a 6-core, 12-thread processor based on the Zen 2 architecture. It has a base clock of 3.60 GHz and a boost clock of 4.20 GHz. Its benchmark scores are modest by today's standards: a Cinebench R23 multi-core score of 15045 and a single-core score of 2124. The Passmark single-thread score is 2562. These numbers, while solid for a mid-range CPU from 2019, are the primary constraint in this system. The 3DMark 2-thread score of 1361 is particularly relevant for an older game like Thief, which may not be well-optimized for many cores.

The CPU's 32 MB of shared L3 cache is a significant asset, but it cannot compensate for the lower single-thread performance when compared to newer processors. The data indicates that the Ryzen 5 3600 is the limiting factor in this pairing. The fact that the average FPS at 1080p Low (126 FPS) is only 35.5% higher than at 4K Low (93 FPS) shows that the CPU is hitting its limit even at low resolutions. If the GPU were the bottleneck, we would expect a much larger difference, closer to a 100% or more increase in FPS when dropping from 4K to 1080p.

The CPU's nearest rivals in the database include the Intel Core i7-1260P, which has a 0.2% higher average score, and the Intel Core i5-11400, which has a 0.2% lower score. These are all mid-range parts. The fact that the Ryzen 5 3600 is grouped with these processors rather than top-tier chips confirms its performance class. In Thief, this CPU is likely unable to feed the GPU with enough draw calls and game logic updates to keep it fully occupied. The result is that the system's frame rate is capped at around 126 FPS, regardless of how powerful the GPU is. The CPU's 65W TDP and 7nm process node are irrelevant here; what matters is the raw execution speed, and it falls short of what the RTX 5090 D needs.

FAQ

Q: Why does this high-end GPU not produce higher frame rates in Thief?

A: The benchmark data indicates the AMD Ryzen 5 3600 CPU is the limiting factor. The frame rate at 1080p Low is 126 FPS, which is only slightly higher than at 4K Low (93 FPS), showing that the CPU reaches its performance ceiling early and prevents the GPU from working to its full potential.

Q: Is the RTX 5090 D's 32 GB of VRAM useful for this game?

A: The data does not suggest the VRAM capacity is a factor. Thief is a 2014 title, and the GPU's massive memory pool of 32 GB is unlikely to be fully utilized. The performance is instead constrained by the CPU, as evidenced by the relatively small FPS differences between low and ultra settings at each resolution.

Q: What is the best resolution to play this game with this hardware combo?

A: According to the measured FPS, the system can maintain an average above 60 FPS at 4K on all presets, with Ultra at 75 FPS. This makes 4K a viable option. However, for higher refresh rates, 1440p offers averages between 89 and 111 FPS, which is a better fit for a 144Hz monitor.

Q: How does this combo's performance compare to other systems in the database?

A: The system ranks 1214th out of 2851 tested combos in Thief. This places it in the upper half of tested systems, but it is relatively low given the GPU's power. This indicates that many other CPU-GPU pairings achieve better results, likely due to more balanced processor choices.

Q: Is the GPU bottlenecked at 1080p or 4K?

A: The data suggests the CPU is the bottleneck at all resolutions. The FPS difference between 1080p Low (126 FPS) and 4K Low (93 FPS) is only 26.2%. If the GPU were the main constraint, the difference would be significantly larger due to the 4x pixel count increase at 4K.

Q: What is the performance impact of the High vs. Ultra settings?

A: The performance impact of moving from High to Ultra is very small. At 4K, the average FPS drops from 76 to 75, a 1.3% decrease. At 1080p, the drop is from 103 to 101, a 1.9% decrease. This suggests that Ultra settings are nearly free in terms of performance cost for this GPU.

Settings Recommendations

Based on the measured FPS data, the Ultra preset is the most logical choice for this hardware combination. The performance cost of Ultra compared to High is minimal, with a difference of only 1 FPS at 4K and 2 FPS at 1080p. Given that the system maintains an average of 75 FPS at 4K Ultra, which is well above the 60 FPS standard for smooth gameplay, there is no reason to sacrifice visual quality by dropping to a lower preset. The data shows that the GPU has enough headroom to handle the extra graphical load from Ultra settings without a significant frame rate penalty.

For users with a 1440p monitor, the Ultra preset is also recommended, as it delivers an average of 89 FPS. This is a good result for a high-refresh-rate display. The Medium preset at 1440p offers 100 FPS, but the 11 FPS gain from dropping from Ultra to Medium is not worth the reduction in visual fidelity, especially since the game is not competitive and frame rates above 90 are already smooth. The data shows that the High preset is a poor middle ground, as it offers only a 2 FPS advantage over Ultra at 1440p, which is negligible.

At 1080p, the Ultra preset still delivers an average of 101 FPS. While the Low preset offers 126 FPS, this 25 FPS difference is unlikely to be noticeable on a standard 60Hz or even 144Hz monitor, especially considering the massive visual downgrade. The game is a stealth title, where image quality and atmosphere are important. Therefore, the Ultra preset is the best choice across all resolutions, as it provides the best visual experience while maintaining frame rates that are well above playable thresholds. The only scenario where a lower preset might be considered is if the user has a very high refresh rate monitor (240Hz or more) and prioritizes frame rate over visuals, but the data suggests the CPU would still limit the maximum FPS to around 126.