Thief

The AMD Ryzen 7 5700G paired with the NVIDIA GeForce RTX 5090 D produces a striking set of results in Thief, a 2014 stealth title that remains a useful benchmark for older game engines. The data shows a system that delivers very high frame rates across all tested resolutions and settings, yet the combination ranks 461st out of 2851 tested combos in this game, which suggests that many other pairings achieve even higher performance. This review examines the measured FPS data, the roles of the CPU and GPU, and what the scaling patterns imply about the limiting component in this specific configuration.

FAQ

Q: What is the average FPS at 1920x1080 with Ultra settings?

A: The measured average FPS at 1920x1080 with Ultra settings is 130 frames per second. This is the lowest average at 1080p among the four settings tested, but it remains well above the refresh rate of most monitors.

Q: How does the average FPS change from 1080p to 4K at High settings?

A: At High settings, the average FPS drops from 132 at 1920x1080 to 116 at 2560x1440, and then to 92 at 3840x2160. This represents a decline of approximately 30% from 1080p to 4K, indicating that the GPU becomes more stressed at higher resolutions.

Q: What is the best average FPS recorded in the measured data?

A: The highest average FPS recorded is 161, achieved at 1920x1080 with Low settings. The next best is 146 at 1080p Medium, followed by 142 at 1440p Low, showing that lower settings consistently produce higher frame rates.

Q: Does the combo rank highly among all tested combinations in this game?

A: The combo ranks 461st out of 2851 tested combinations in Thief. This places it in the top 16% of all tested combos, which is a strong position, though not at the very top of the leaderboard.

Q: What are the minimum and maximum FPS values recorded for Medium settings at 2560x1440?

A: At 2560x1440 with Medium settings, the system recorded a minimum of 109 FPS and a maximum of 148 FPS, with an average of 129 FPS. This indicates a relatively stable performance window across the test run.

Q: What is the GPU's average benchmark score compared to its nearest rivals?

A: The RTX 5090 D has an average benchmark score of 77712, which is 1.1% higher than the AMD Radeon RX 6650M XT’s score of 76904 and 1.6% lower than the AMD Radeon RX 6850M XT’s score of 78940. The data shows it sits between these two rivals in raw compute performance.

CPU Role

The AMD Ryzen 7 5700G is an 8-core, 16-thread processor based on the Zen 3 architecture, with a base clock of 3.80 GHz and a boost clock of 4.60 GHz. It features a 16 MB L3 cache and supports dual-channel DDR4 memory with a bandwidth of 51.2 GB/s. In synthetic benchmarks, this CPU achieves a Cinebench R23 multi-core score of 12830.5 and a single-core score of 1494, while its Geekbench multi-core score is 9950 and single-core score is 1894. The CPU's percentile rank among all tested processors is 79, meaning it outperforms 79% of the CPUs in the database.

In the context of Thief, a game from 2014, the CPU's role is less demanding than in modern titles. The game's engine was designed for the hardware of its era, so even a mid-range processor from 2021 like the 5700G is likely sufficient to avoid bottlenecking the GPU. The measured FPS data supports this interpretation: at 1080p, the average FPS ranges from 130 at Ultra to 161 at Low, which are high numbers that suggest the CPU is not a limiting factor at this resolution. If the CPU were the bottleneck, we would expect to see smaller differences between resolution settings, as the CPU workload does not scale with resolution in the same way the GPU workload does.

The 5700G's single-thread performance, measured at 899 in 3DMark single-thread and 3283 in PassMark single-thread, is relevant for a game like Thief that relies heavily on a few primary threads for game logic and physics. The data shows that even at 4K with Ultra settings, the average FPS is 87, which is well above playable thresholds. This suggests that the CPU is delivering enough single-thread throughput to keep the game running smoothly, even when the GPU is under maximum load. The 16 threads are likely overkill for this title, but the high single-core boost clock of 4.60 GHz ensures that the primary game thread has ample processing headroom.

How This Combo Ranks

The combination of the Ryzen 7 5700G and the RTX 5090 D achieves a rank of 461 out of 2851 tested combos in Thief. This rank places the pairing in the 84th percentile of all tested combinations, which is a strong showing for a CPU that is not typically associated with top-tier gaming performance. The data suggests that the GPU is doing most of the heavy lifting, as the RTX 5090 D is a flagship-class card, but the CPU is not holding it back to a significant degree.

To contextualize this rank, consider that the CPU's average benchmark score of 27051 is nearly identical to its nearest rivals, with the Intel Core i9-9900K scoring 27097 (a delta of -0.2%) and the AMD Ryzen AI 7 445 scoring 26936 (a delta of 0.4%). This means the 5700G is performance-equivalent to these CPUs in general benchmarks. In Thief, the combo rank of 461 suggests that the overall system is competitive with many other pairings, but there are likely 460 combinations that achieve higher frame rates. This could be due to CPUs with higher single-thread performance or GPUs with better optimization for this specific game engine.

The rank also indicates that the RTX 5090 D is not being fully utilized at all settings. At 1080p Low, the average FPS is 161, which is high but not exceptionally so for a card of this class. The GPU's nearest rivals in the benchmark database include the AMD Radeon RX 6850M XT, which has a 1.6% higher average score, yet the RTX 5090 D is a desktop card with much higher specifications. This discrepancy suggests that Thief may not be able to fully leverage the GPU's capabilities due to engine limitations or CPU overhead, which would explain the mid-tier rank despite the powerful hardware.

Measured FPS Breakdown

The measured FPS data provides a comprehensive look at how this combo performs across three resolutions and four settings levels. At 1920x1080, the average FPS ranges from 161 at Low to 130 at Ultra. The Medium setting records 146 FPS with a minimum of 124 and a maximum of 168, while High delivers 132 FPS. These numbers show that the system can easily handle 1080p gaming with maximum refresh rates, as even the lowest average (130 at Ultra) is above 120 FPS.

Moving to 2560x1440, the average FPS decreases slightly but remains high. Low settings produce 142 FPS, Medium delivers 129 FPS (with a minimum of 109 and maximum of 148), High yields 116 FPS, and Ultra drops to 114 FPS. The difference between High and Ultra at 1440p is only 2 FPS, which suggests that the Ultra preset may not be significantly more demanding than High, or that the GPU is reaching a performance plateau at this resolution. The minimum FPS of 109 at Medium indicates that there are occasional dips, but they stay well above 60 FPS.

At 3840x2160, the average FPS drops further. Low settings produce 119 FPS, Medium delivers 108 FPS (with a minimum of 92 and maximum of 124), High yields 92 FPS, and Ultra drops to 87 FPS. The pattern here is consistent: higher settings reduce performance, and 4K resolution puts more strain on the GPU. The minimum FPS of 92 at Medium is still very playable, but the Ultra average of 87 FPS is the lowest recorded in the entire dataset. This indicates that 4K Ultra is the most demanding configuration, but the system still maintains a smooth experience.

The data also reveals that the difference between Low and Ultra at each resolution is relatively consistent. At 1080p, the gap is 31 FPS (161 to 130), at 1440p it is 28 FPS (142 to 114), and at 4K it is 32 FPS (119 to 87). This consistency suggests that the settings scaling is uniform across resolutions, with the GPU handling the increased workload proportionally.

GPU Role

The NVIDIA GeForce RTX 5090 D is a flagship GPU with 32 GB of GDDR7 memory on a 512-bit bus, providing a memory bandwidth of 1.79 TB/s. Its base clock is 2017 MHz with a boost clock of 2407 MHz, and it features 21760 shading units, 680 texture mapping units, and 176 ROPs. The GPU is built on the Blackwell 2.0 architecture with a 5 nm process node and contains 92,200 million transistors on a 750 mm² die. It supports PCIe 5.0 x16 and has a TDP of 575 W.

In benchmark tests, the RTX 5090 D scores 44065 in PassMark G3D and 14326 in 3DMark Steel Nomad DX12. Its Geekbench OpenCL score is 310674 and Vulkan score is 376915. The GPU's percentile rank is 92, meaning it outperforms 92% of all GPUs in the database. Its nearest rivals include the AMD Radeon RX 6850M XT, which has a 1.6% higher average score, and the NVIDIA Tesla P100 PCIe 16 GB, which has a 2.4% higher average score. This is surprising given that the RTX 5090 D is a newer and more powerful card, but the benchmark scores reflect synthetic tests that may not align perfectly with gaming performance.

In Thief, the GPU's role is dominant at higher resolutions. The FPS data shows that at 4K, the average FPS drops by approximately 30% compared to 1080p at the same settings. This is a classic GPU-bound scenario, where the rendering resolution increases the workload on the GPU. The RTX 5090 D's massive memory bandwidth of 1.79 TB/s and 32 GB of VRAM are more than sufficient for this 2014 game, which likely uses less than 4 GB of VRAM even at 4K. The GPU's high pixel rate of 423.6 GPixel/s and texture rate of 1,636.8 GTexel/s ensure that it can handle the game's rendering demands with ease, as evidenced by the 87 FPS average at 4K Ultra.

Settings Recommendations

Based on the measured FPS data, the optimal settings depend on the user's target resolution and refresh rate. At 1920x1080, all settings produce average FPS above 130, so the Ultra preset is recommended for the best visual quality without sacrificing performance. The difference between High and Ultra is only 2 FPS (132 vs 130), so choosing Ultra provides a marginal visual improvement at negligible performance cost.

At 2560x1440, the Medium settings offer a balanced experience with an average of 129 FPS and a minimum of 109 FPS. The Ultra preset delivers 114 FPS, which is still smooth, but the 15 FPS difference between Medium and Ultra may not justify the potential visual improvements. For users with 144 Hz monitors, High settings at 116 FPS are also viable, but Medium provides a more comfortable margin above the 100 FPS threshold.

At 3840x2160, the data shows that Medium settings produce an average of 108 FPS with a minimum of 92 FPS, which is the best balance of quality and performance. Low settings increase the average to 119 FPS, but the visual downgrade may be significant. High settings drop to 92 FPS, which is still playable, but the minimum FPS is not recorded for this setting, so stability is uncertain. Ultra at 4K delivers 87 FPS, which is the lowest measured average, but remains above 60 FPS for a playable experience. Therefore, Medium is the recommended preset for 4K gaming, as it maintains a high frame rate with a reasonable minimum FPS.

Resolution Scaling

The resolution scaling data reveals important insights about the limiting component in this system. At High settings, the average FPS drops from 132 at 1080p to 116 at 1440p (a 12% decrease) and then to 92 at 4K (a 21% decrease from 1440p). This pattern shows that the FPS penalty increases as resolution rises, which is typical of a GPU-bound scenario. The GPU is being tasked with rendering more pixels, and its performance scales accordingly.

At Low settings, the FPS drops from 161 at 1080p to 142 at 1440p (a 12% decrease) and then to 119 at 4K (a 16% decrease from 1440p). The smaller percentage drop at 4K Low compared to High suggests that the GPU is more capable of handling the lower settings, but the resolution still has a significant impact. At Medium settings, the FPS drops from 146 at 1080p to 129 at 1440p (a 12% decrease) and then to 108 at 4K (a 16% decrease from 1440p). The minimum FPS at Medium also drops from 124 at 1080p to 109 at 1440p and 92 at 4K, showing that the frame pacing becomes less consistent at higher resolutions.

The scaling from 1080p to 4K at each settings level shows a total FPS reduction of approximately 30% at High, 26% at Low, and 26% at Medium. This indicates that the GPU is the primary limiting component, as the CPU's workload does not change significantly with resolution. The 5700G's performance is sufficient to feed the GPU at all resolutions, but the RTX 5090 D's raw power is the main driver of FPS. The data implies that if the CPU were the bottleneck, the FPS would remain constant across resolutions, but the clear downward trend proves that the GPU is doing the heavy lifting. This is a well-balanced pairing for this game, with the GPU being the primary performance determinant and the CPU providing adequate support.