Thief

The Intel Core i5-10400F paired with the NVIDIA GeForce RTX 5070 Ti lands at rank 752 out of 1642 tested combos in Thief, placing it in the upper-middle tier of the database. This position reflects a substantial CPU bottleneck at lower resolutions, while the GPU’s raw power becomes the dominant factor as resolution climbs. The combo’s overall standing is respectable but not elite, with the data showing that the 6-core, 12-thread Comet Lake processor holds back the Blackwell GPU in lighter-load scenarios.

How This Combo Ranks

The combo’s rank of 752 out of 1642 tested configurations means it sits just ahead of the median, specifically in the 54th percentile of all tested pairings. This is a middling result for a system pairing a 10th-gen Core i5 with a flagship-tier RTX 50-series card. The CPU’s average benchmark score of 14168 places it in the 73rd percentile against all CPUs, but its nearest rivals show how tight the competition is: the AMD EPYC 7552 scores 14115 (0.4% higher), the AMD Ryzen 3 7320C scores 14277 (0.8% higher), the Intel Core i5-8500 scores 14332 (1.1% higher), and the AMD Ryzen Embedded V2546 scores 14336 (1.2% higher). The i5-10400F is effectively mid-pack among these comparables, trailing the fastest by roughly 1.2% while edging out the EPYC by a hair.

This CPU performance profile explains why the combo doesn’t crack the top half of the ranking. In a 2014 stealth title like Thief, which is not heavily multithreaded, the i5-10400F’s single-thread score of 688 in 3DMark and 1453 in Cinebench R23 single-core are adequate but unremarkable. The GPU, meanwhile, sits in the 87th percentile of all GPUs with an average score of 52086, making it the clear star of this pairing. The rank discrepancy—GPU in the top 13% versus a combo in the 54th percentile—quantifies the bottleneck: the CPU is the limiting factor in many scenes, dragging the overall combo down from what the GPU alone could achieve.

GPU Role

The RTX 5070 Ti brings 16 GB of GDDR7 memory on a 256-bit bus, delivering 896.0 GB/s of bandwidth. Its clocks are modest relative to its memory speed: a 2295 MHz base and 2452 MHz boost, with memory running at 1750 MHz (28 Gbps effective). These specifications are more than sufficient for Thief, a game from 2014 that predates modern VRAM demands. The GPU’s 8960 shading units and 280 TMUs provide 43.94 TFLOPS of FP32 compute and a texture rate of 686.6 GTexel/s, all of which go largely underutilized at lower resolutions where the CPU becomes the constraint.

The data confirms this dynamic. At 1080p Low, the combo produces 125.2 FPS, but this is likely CPU-limited given the i5-10400F’s modest single-thread capabilities. Moving to 1440p Low, the frame rate drops to 112.7 FPS, a smaller relative decline than what pure GPU scaling would suggest, indicating the CPU is still partially holding things back. Only at 4K does the GPU’s role become fully apparent: 4K Low yields 94.5 FPS, which is a 25% drop from 1080p Low—a scaling pattern that shows the GPU now being the primary bottleneck. The GPU’s 70 RT cores and 280 tensor cores are irrelevant for Thief, as the game does not use ray tracing or DLSS, but the raw rasterization power ensures playable frame rates even at maximum settings in 4K.

FAQ

Q: Is the RTX 5070 Ti overkill for a 2014 game like Thief?

A: The data shows the GPU is severely underutilized at lower resolutions. At 1080p Ultra, the combo produces 99 FPS, which is close to the 1080p High result of 99.8 FPS, suggesting the CPU is capping performance. Only at 4K Ultra does the GPU’s workload become significant, yielding 68.8 FPS.

Q: How much does the CPU bottleneck the GPU in this pairing?

A: The bottleneck is most visible at 1080p. The frame rate difference between 1080p Low (125.2 FPS) and 1440p Low (112.7 FPS) is only 10%, which is far less than the 44% increase in pixel count, indicating the CPU cannot feed the GPU fast enough at the lower resolution. At 4K, the scaling becomes more linear, showing the GPU taking over as the limiting factor.

Q: What is the best settings/resolution combination for this combo?

A: The data suggests 1440p Ultra offers the best balance, producing 90 FPS. This is only 9 FPS below the 1080p Ultra result (99 FPS), meaning the resolution upgrade costs little performance while the CPU bottleneck is less severe. 4K High is also viable at 75.9 FPS.

Q: How does this combo compare to similar CPU-GPU pairings?

A: The GPU’s nearest rival, the NVIDIA GeForce RTX 4080, scores 52085 (0% deltaPct), meaning the 5070 Ti is statistically identical in average benchmarks. The CPU’s rivals are all within 1.2% of its average score, with the AMD Ryzen 3 7320C being the closest at 14277 (0.8% higher).

Q: Does the 16 GB VRAM matter for Thief?

A: No. Thief, released in 2014, does not require anywhere near 16 GB of VRAM. The 896.0 GB/s of bandwidth and 16 GB capacity are far beyond what this game can utilize, which is why the GPU’s percentile (87th) does not translate into a top-tier combo rank.

Q: Is the 5070 Ti’s performance consistent with its benchmark scores?

A: The GPU’s average benchmark score of 52086 places it in the 87th percentile, and its measured performance in Thief reflects this—it delivers over 90 FPS at 1440p High and over 75 FPS at 4K High. However, the combo’s 752nd rank out of 1642 shows that the CPU prevents the GPU from reaching its full potential in this title.

Measured FPS Breakdown

The measured data covers four settings levels (Low, Medium, High, Ultra) across three resolutions, providing a complete picture of scaling behavior.

At 1920x1080, the combo delivers 125.2 FPS on Low, 110.9 FPS on Medium, 99.8 FPS on High, and 99 FPS on Ultra. The tight spread between High and Ultra (0.8 FPS) indicates a hard CPU ceiling around 99-100 FPS, as the GPU can clearly render more frames but is not being fed fast enough. The jump from Low to Medium costs 14.3 FPS, while the jump from Medium to High costs 11.1 FPS, but the High-to-Ultra step is negligible—a classic sign of processor limitation.

At 2560x1440, the results are 112.7 FPS on Low, 100.9 FPS on Medium, 92.6 FPS on High, and 90 FPS on Ultra. The CPU ceiling has lifted somewhat because the GPU is taking on more work per frame, but the pattern persists: Ultra is only 2.6 FPS slower than High, while Low to Medium drops 11.8 FPS. The 1440p Ultra result of 90 FPS is only 9 FPS below 1080p Ultra, confirming that the system is still partially CPU-bound even at this higher resolution.

At 3840x2160, the GPU finally becomes the dominant constraint. The results are 94.5 FPS on Low, 86 FPS on Medium, 75.9 FPS on High, and 68.8 FPS on Ultra. Here, the scaling is more gradual and linear: Low to Medium costs 8.5 FPS, Medium to High costs 10.1 FPS, and High to Ultra costs 7.1 FPS. This pattern indicates the GPU is now the bottleneck, as each settings increase has a consistent impact on frame rate.

Resolution Scaling

The frame rate drop from 1080p to 4K reveals the shifting bottleneck. At Low settings, performance goes from 125.2 FPS at 1080p to 112.7 FPS at 1440p (a 10% drop) and then to 94.5 FPS at 4K (a 25% drop from 1080p). The 1440p result is disproportionately high relative to the pixel count increase, confirming CPU limitation. However, the 4K result drops more steeply, showing the GPU beginning to struggle with the 8.3 million pixel workload.

At Ultra settings, the pattern is different. 1080p Ultra yields 99 FPS, 1440p Ultra yields 90 FPS (a 9% drop), and 4K Ultra yields 68.8 FPS (a 31% drop from 1080p). The 1440p-to-4K transition is the steepest, costing 21.2 FPS, while the 1080p-to-1440p transition costs only 9 FPS. This inverted scaling curve—where the high-resolution drop is larger than the mid-resolution drop—is the signature of a system transitioning from CPU-bound to GPU-bound as resolution increases.

The limiting component is clear: at 1080p and 1440p, the i5-10400F’s single-thread performance (688 in 3DMark single-thread, 1453 in Cinebench R23 single-core) caps frame rates, particularly at High and Ultra settings. At 4K, the RTX 5070 Ti’s rendering workload becomes heavy enough that its 43.94 TFLOPS of compute and 686.6 GTexel/s texture rate are the constraining factors. The data shows that this combo is best suited for 1440p gaming, where the CPU bottleneck is moderate and the GPU still has headroom, or 4K gaming where the GPU is fully utilized but frame rates remain playable. The 1080p results, while high in absolute terms, do not scale with settings changes as expected, indicating wasted GPU potential.