Thief

# Thief Performance Analysis: AMD Ryzen 7 9700F + NVIDIA GeForce RTX 5070

The AMD Ryzen 7 9700F paired with the NVIDIA GeForce RTX 5070 delivers a compelling 1080p and 1440p experience in Thief, a 2014 stealth title that remains surprisingly demanding at higher resolutions. The measured FPS data reveals a system that excels at lower resolutions but shows clear signs of GPU saturation as pixel count increases, with the 4K Ultra preset barely maintaining playable frame rates. This analysis examines how the Zen 5 processor and Blackwell 2.0 GPU interact across twelve distinct resolution-settings combinations, providing a detailed picture of where this combo thrives and where it struggles.

Resolution Scaling

The FPS drop from 1080p to 4K is substantial across all settings, but the pattern reveals important insights about the system's bottleneck characteristics. At Low settings, the average frame rate falls from 231 FPS at 1920x1080 to 186 FPS at 2560x1440, then drops sharply to 110 FPS at 3840x2160. This represents a 52.4% reduction from 1080p to 4K at Low settings, indicating that even the least demanding preset becomes increasingly GPU-bound at higher resolutions.

The Medium preset shows a similar trajectory: 206 FPS at 1080p, 141 FPS at 1440p, and 83 FPS at 4K. The percentage drop from 1080p to 4K here is 59.7%, slightly larger than the Low preset's decline. This suggests that as settings increase, the GPU's workload grows disproportionately relative to the CPU's ability to feed frames, a classic sign of the RTX 5070 becoming the limiting component at higher resolutions.

High settings follow the trend with 157 FPS at 1080p, 108 FPS at 1440p, and 63 FPS at 4K—a 59.9% reduction from the lowest to highest resolution. Ultra settings show the most dramatic scaling: 149 FPS at 1080p, 103 FPS at 1440p, and just 60 FPS at 4K, a 59.7% decrease. The consistency of these percentage drops across Medium, High, and Ultra (all hovering near 60%) indicates that the GPU's rendering pipeline is the primary constraint once resolution increases, rather than any CPU-side limitation.

The 4K Ultra result of 60 FPS is particularly noteworthy because it sits exactly at the threshold many consider the minimum for smooth gameplay. The data implies that the RTX 5070's 12 GB of GDDR7 memory and 672.0 GB/s bandwidth are sufficient for 4K in this title, but just barely at maximum settings. The fact that 4K Low achieves 110 FPS while 4K Ultra achieves only 60 FPS—a 45.5% performance penalty—shows that the GPU's compute resources, not memory capacity, are the limiting factor at this resolution.

Interestingly, the scaling from 1440p to 4K is steeper than from 1080p to 1440p at every settings level. For example, at High settings, the drop from 1080p to 1440p is 31.2%, while the drop from 1440p to 4K is 41.7%. This non-linear scaling suggests that the GPU's memory bandwidth or pixel throughput (201.0 GPixel/s) becomes increasingly strained as the pixel count approaches 4K's 8.3 million pixels, reinforcing the GPU-bound conclusion.

Settings Recommendations

The measured data suggests that the optimal experience depends heavily on target resolution. At 1920x1080, even the Ultra preset delivers 149 FPS, which is well above the refresh rate of most standard monitors. The Medium preset offers 206 FPS, a 38.3% improvement over Ultra, while Low reaches 231 FPS. Given that the CPU's PassMark single-thread score of 4691 and multithread score of 36470 indicate strong processing capability, the 1080p results are likely CPU-limited at lower settings, meaning the GPU has headroom that isn't being utilized.

For 2560x1440, the High preset at 108 FPS appears to be the sweet spot. It provides a 4.9% improvement over Ultra's 103 FPS while presumably offering similar visual quality to High settings. The Medium preset at 141 FPS offers a 30.6% boost over High, which could benefit high-refresh-rate 1440p monitors, but the min FPS of 120 suggests occasional dips that might be noticeable. The Low preset at 186 FPS is excessive for most displays and sacrifices too much visual fidelity for the marginal gain.

At 3840x2160, the situation becomes more constrained. The Ultra preset's 60 FPS is technically playable but leaves no headroom for demanding scenes. The High preset at 63 FPS offers only a 5% improvement, suggesting that the Ultra-to-High transition doesn't provide significant relief. The Medium preset at 83 FPS with a min FPS of 71 and max of 96 represents a more comfortable margin, offering a 38.3% improvement over Ultra while maintaining acceptable visual quality. For 4K gaming on this combo, Medium appears to be the recommended preset, as it provides the best balance between frame rate stability and visual fidelity.

The min FPS data for Medium settings across all resolutions is telling: 175 FPS at 1080p, 120 FPS at 1440p, and 71 FPS at 4K. These minimums indicate that the GPU can maintain reasonably consistent frame delivery, with the gap between average and minimum FPS widening at higher resolutions. At 4K Medium, the 12 FPS difference between average (83) and minimum (71) suggests occasional stutters, but nothing that would render the experience unplayable.

CPU Role

The AMD Ryzen 7 9700F brings substantial processing power to this pairing, with 8 cores and 16 threads based on the Zen 5 architecture fabricated on TSMC's 4 nm process. The CPU's boost clock of 5.50 GHz, combined with 32 MB of shared L3 cache, provides the kind of single-thread performance that older titles like Thief often rely upon. The PassMark single-thread score of 4691 places this CPU in the 94th percentile of all CPUs, indicating excellent per-core performance that should handle game logic and draw call submission without becoming a bottleneck.

However, the measured FPS data suggests that the CPU's role diminishes as resolution increases. At 1080p Low, the 231 FPS result is likely near the CPU's maximum frame delivery capability for this game, given that the GPU is barely stressed at this resolution. The fact that 1080p Ultra (149 FPS) is only 35.5% lower than 1080p Low (231 FPS) despite the significantly increased GPU workload suggests that the CPU is keeping pace even at higher settings, with the GPU becoming the primary constraint.

The CPU's multithread score of 36470 and PassMark physics score of 2122 indicate robust multi-core performance, but Thief's age (released in 2014) means it likely doesn't utilize all 16 threads efficiently. The PassMark data compression score of 421988 and integer math score of 120788 suggest strong general-purpose computing, which could translate to faster level loading and AI processing, though these aren't directly measurable from the FPS data.

The CPU's 89.6 GB/s memory bandwidth and DDR5 support provide ample data throughput for the GPU to access. The PCIe Gen 5 interface with 24 lanes ensures that the RTX 5070's data transfers aren't constrained by the bus, though the GPU's 12 GB VRAM likely means most game assets are stored locally on the graphics card after initial loading.

Interestingly, the CPU's nearest rivals in benchmark scores—the AMD Ryzen 9 7940HX (+0.2%), Intel Core i7-14700KF (-0.2%), AMD Ryzen 9 7950X (+0.7%), and Intel Core i7-14700K (+0.9%)—all fall within 1% of the 9700F's average benchmark score of 69996. This indicates that the 9700F is competitively positioned in its performance class, and any frame rate differences in Thief would likely be minimal when swapping among these processors.

FAQ

Q: Is the RTX 5070 sufficient for 4K gaming in Thief?

A: The data shows 60 FPS at 4K Ultra and 63 FPS at 4K High, which are technically playable but leave little headroom. For a more comfortable experience, the Medium preset at 83 FPS with a 71 FPS minimum is recommended.

Q: How does the Ryzen 7 9700F compare to its nearest rivals in CPU performance?

A: The CPU's average benchmark score of 69996 is within 0.9% of all four nearest rivals. It sits 0.2% above the AMD Ryzen 9 7940HX, 0.2% below the Intel Core i7-14700KF, 0.7% above the AMD Ryzen 9 7950X, and 0.9% above the Intel Core i7-14700K.

Q: What's the best settings preset for 1440p gaming?

A: The High preset at 108 FPS offers a good balance, providing a 4.9% improvement over Ultra (103 FPS). The Medium preset at 141 FPS is better for high-refresh displays, but the 120 FPS minimum might cause occasional dips.

Q: Does the CPU bottleneck the GPU at 1080p?

A: The 231 FPS result at 1080p Low suggests the CPU is likely the limiting factor at this resolution, as the GPU should easily handle this workload. The relatively small gap between Low (231 FPS) and Ultra (149 FPS) at 1080p supports this interpretation.

Q: How much VRAM does the RTX 5070 have, and is it enough for Thief?

A: The GPU features 12 GB of GDDR7 memory on a 192-bit bus with 672.0 GB/s bandwidth. The measured 4K results indicate this is sufficient for Thief, as the game runs at playable frame rates even at Ultra settings.

Q: What kind of frame rate stability can I expect at 4K Medium?

A: The min FPS of 71, average of 83, and max of 96 indicate reasonable stability, with a 12 FPS spread between minimum and average. This suggests occasional dips but no severe stuttering.

GPU Role

The NVIDIA GeForce RTX 5070 serves as the primary performance driver in this pairing, and its characteristics directly explain the measured FPS patterns. Built on the Blackwell 2.0 architecture using TSMC's 5 nm process, this GPU packs 31,100 million transistors into a 263 mm² die. The boost clock of 2512 MHz and base clock of 2325 MHz provide the raw compute throughput, while the 6144 shading units deliver 30.87 TFLOPS of FP32 performance.

The GPU's memory subsystem—12 GB of GDDR7 on a 192-bit bus with 672.0 GB/s bandwidth—appears well-matched to Thief's requirements. The 4K benchmark results show that even at Ultra settings, the GPU can maintain 60 FPS, suggesting that the 12 GB capacity is sufficient for this game's textures and assets. The memory clock of 1750 MHz (28 Gbps effective) provides the bandwidth needed to feed the 201.0 GPixel/s pixel rate and 482.3 GTexel/s texture rate.

The GPU's PassMark G3D score of 29137 and 3DMark Steel Nomad DX12 score of 5077 place it in the 82nd percentile of all GPUs. However, its nearest rivals in benchmark scores are somewhat unexpected: the AMD Radeon Pro 580 (+0.1%), AMD Radeon Pro WX 7100 (+0.8%), AMD Radeon Pro 5300 (-1.2%), and NVIDIA RTX A500 Mobile (+2%). These are professional and mobile GPUs, suggesting that the RTX 5070's average benchmark score of 40377 is in a performance tier that includes various workstation-class parts.

The GPU's role becomes increasingly dominant as resolution rises. At 1080p, the CPU likely provides some limitation, particularly at Low settings where 231 FPS may approach the CPU's maximum frame rate. But at 4K, the GPU is clearly the bottleneck, as evidenced by the consistent ~60% FPS drop from 1080p to 4K across Medium, High, and Ultra settings. The 80 ROPs and 192 TMUs are sufficient for Thief's rendering requirements, but the pixel throughput at 4K (8.3 million pixels) stresses the GPU's fill rate capabilities.

The RTX 5070's support for DirectX 12 Ultimate (12_2), Vulkan 1.4, and OpenGL 4.6 means it can handle Thief's rendering features, though the game's 2014 release date suggests it likely uses DirectX 11 or earlier APIs. The GPU's 48 RT cores and 192 tensor cores provide future-proofing for newer titles, but they don't directly impact Thief's performance since the game doesn't use ray tracing.

Measured FPS Breakdown

At 1920x1080, the system delivers exceptional performance across all settings. Low settings produce 231 FPS average, Medium achieves 206 FPS (min 175, max 237), High reaches 157 FPS, and Ultra maintains 149 FPS. The gap between Low and Ultra is 82 FPS (35.5%), indicating that even at maximum settings, the system operates well above the 144 Hz refresh rate common on modern monitors. The Medium preset's min FPS of 175 ensures that even demanding scenes stay above 144 FPS.

At 2560x1440, performance remains strong but shows the GPU's increasing workload. Low settings drop to 186 FPS, Medium averages 141 FPS (min 120, max 163), High achieves 108 FPS, and Ultra delivers 103 FPS. The difference between High and Ultra is just 5 FPS (4.9%), suggesting that Ultra adds minimal visual benefit relative to its performance cost. The Medium preset's 120 FPS minimum is adequate for most displays, though high-refresh-rate panels might notice occasional dips.

At 3840x2160, the system's limitations become apparent. Low settings produce 110 FPS, Medium averages 83 FPS (min 71, max 96), High reaches 63 FPS, and Ultra drops to 60 FPS. This is the first resolution where no preset exceeds 120 FPS, and Ultra's 60 FPS result barely meets the minimum threshold for smooth gameplay. The Medium preset offers the best balance with 83 FPS average and 71 FPS minimum, providing a 38.3% improvement over Ultra while maintaining playable frame rates. The 4K results clearly indicate that the RTX 5070, while capable, is not over-provisioned for this game at maximum settings.

The complete dataset shows a consistent pattern: each settings increase costs roughly 20-30% performance at any given resolution, and each resolution increase costs roughly 30-60% performance depending on settings. The most efficient configuration in terms of frames per settings increment appears to be the Medium preset, which offers the smallest average-to-minimum FPS gap and the best balance of performance and visual quality across all resolutions.

How This Combo Ranks

The AMD Ryzen 7 9700F + NVIDIA GeForce RTX 5070 combination ranks 267th out of 2851 tested combos in Thief, placing it in the top 9.4% of all configurations. This ranking reflects the system's strong overall performance, particularly at 1080p and 1440p where it delivers frame rates that exceed the requirements of most displays. The top 10% placement indicates that this combo outperforms the vast majority of tested systems, likely due to the combination of the CPU's high single-thread performance and the GPU's robust compute capabilities.

The ranking is somewhat surprising given that the GPU's benchmark percentile (82nd) is lower than the CPU's (94th), suggesting that the GPU is the limiting factor in this pairing. The fact that the combo still ranks in the top 9.4% of all tested combinations indicates that Thief is either more CPU-sensitive than the FPS data suggests, or that many other tested combos have even weaker GPUs or CPUs. The rank of 267 out of 2851 also implies that roughly 2584 tested combos perform worse than this pairing, which reflects the balanced nature of the 9700F and RTX 5070 combination.

Compared to the CPU's nearest rivals (all within 0.9% in benchmark scores), this combo's performance in Thief would likely be very similar if paired with the same GPU, since the CPU is not the primary bottleneck at higher resolutions. However, at 1080p Low settings where the CPU is more likely to be the limiting factor, the 9700F's strong single-thread performance (PassMark single-thread score of 4691) might provide a slight edge over rivals with lower single-thread scores, though the delta would be small given the narrow benchmark differences.

The GPU's nearest rivals in benchmark scores are primarily professional or mobile parts, which suggests that the RTX 5070's performance class in Thief might not align perfectly with its synthetic benchmark positioning. The combo's strong ranking indicates that the RTX 5070's real-world gaming performance in Thief exceeds what its 82nd percentile benchmark score might suggest, possibly due to the game's age and the GPU's modern architecture handling older workloads efficiently. Overall, this combo delivers a high-quality Thief experience, with the data showing it excels at 1080p and 1440p while remaining playable at 4K with appropriate settings adjustments.