Thief

# Thief with Intel Core i5-10400F + NVIDIA GeForce RTX 5090

The pairing of a 6-core Comet Lake CPU with a 32 GB Blackwell flagship GPU creates a fascinating bottleneck profile in Thief, a 2014 stealth title that predates the CPU by six years. The measured FPS data shows a system that scales predictably with resolution, but the CPU's modest single-thread performance (3DMark single-thread score of 688, PassMark single-thread of 2541) becomes the limiting factor at lower resolutions. The RTX 5090, with its 104.8 TFLOPS FP32 throughput and 1.79 TB/s memory bandwidth, is clearly capable of far more than this CPU can feed at 1080p, yet the game remains playable across all tested configurations.

Resolution Scaling

The FPS delta from 1080p to 4K reveals the classic signature of a CPU-bound scenario at lower resolutions. At Ultra settings, the average frame rate drops from 100 FPS at 1920x1080 to 89 FPS at 2560x1440 (an 11% decline), and then to 74 FPS at 3840x2160 (another 16.9% decline). The total drop from 1080p to 4K is 26% at Ultra, which is modest compared to what a pure GPU-bound title would show with this hardware. A GPU-limited game would typically see a much steeper falloff as pixel count quadruples; here, the 4K-to-1080p ratio is only 0.74, indicating that the CPU is holding back performance more than the GPU at every resolution.

The Low settings preset tells an even more revealing story. At 1080p Low, the system hits 125 FPS, but at 1440p Low it reaches 110 FPS, and at 4K Low it manages 92 FPS. The drop from 1080p to 4K at Low is 26.4% — nearly identical to the Ultra preset's decline. This consistency across settings strongly suggests that the bottleneck is not the GPU, since lowering graphical load would normally produce larger FPS gains if the GPU were the constraint. Instead, the data indicates that the i5-10400F's 6 cores and 12 threads, boosting to 4.30 GHz, are the primary limiter. At 1080p Low, the GPU is likely idling while waiting for the CPU to deliver draw calls, which explains why lowering settings from Ultra to Low only yields a 25% FPS improvement (100 to 125) rather than the massive jump one would expect from such a powerful GPU.

At Medium settings, the pattern holds with slightly more headroom. The 1080p Medium result of 113 FPS (with a minimum of 96 and maximum of 130) drops to 99 FPS at 1440p (min 85, max 114), then to 83 FPS at 4K (min 71, max 96). The minimum frame rates scale proportionally with averages, suggesting no sudden stuttering or frame-time spikes — the CPU's 12 MB of shared L3 cache and 42.7 GB/s memory bandwidth are sufficient to maintain consistent frame pacing even when under load.

The High preset sits between Medium and Ultra, as expected. At 1080p High, the average is 102 FPS; at 1440p it drops to 90 FPS; at 4K it falls to 75 FPS. The 4K-to-1080p ratio at High is 0.735, nearly identical to the other presets. This uniformity across all four settings levels is the strongest evidence that the CPU, not the GPU, dictates performance. The RTX 5090's raw power — 92,200 million transistors on a 750 mm² die with 21760 shading units — is being underutilized in Thief, a game from 2014 that cannot leverage the Blackwell architecture's modern features like 170 RT cores or 680 tensor cores.

FAQ

Q: Why does the FPS drop so little when going from 1080p to 4K?

A: The average FPS at Ultra settings only falls from 100 at 1080p to 74 at 4K — a 26% reduction despite a 4x increase in pixel count. This indicates the Intel Core i5-10400F is the performance ceiling; the RTX 5090 has ample headroom to handle 4K rendering, but the CPU's 6 cores and 12 threads cannot deliver frames faster than roughly 125 FPS even at the lowest settings.

Q: Is the RTX 5090 being fully utilized in Thief?

A: No. The data suggests significant GPU underutilization. At 1080p Low, the system produces 125 FPS, while at 4K Ultra it produces 74 FPS. If the GPU were the bottleneck, the 1080p Low result would be far higher relative to 4K Ultra, given the massive difference in rendering load. The GPU's 32 GB of GDDR7 memory and 512-bit bus are irrelevant to this game's performance ceiling.

Q: How does the i5-10400F compare to its nearest rivals in synthetic benchmarks?

A: The CPU's average benchmark score is 14185, placing it at the 68th percentile of all CPUs. Its nearest rival, the Intel Xeon 6756E, scores 14163 (0.2% lower), while the AMD EPYC 7552 scores 14115 (0.5% lower). On the other side, the Intel Core 7 160UL scores 14232 (0.3% higher) and the AMD Ryzen 3 7320C scores 14277 (0.6% higher). These tiny deltas show the i5-10400F sits in a tightly packed performance cluster.

Q: What is the best resolution for this combo in Thief?

A: Based on the measured data, 1440p Medium offers the best balance. At 99 FPS average with an 85 FPS minimum, it provides smooth gameplay while the 4K Medium result of 83 FPS (71 FPS minimum) is also viable. The 1080p results exceed 100 FPS at all settings but waste the RTX 5090's resolution capabilities.

Q: How does this combo rank among tested configurations?

A: The combo ranks 1684 out of 3720 tested combinations, placing it in the 54.7th percentile. This middling rank reflects the CPU's inability to fully drive the flagship GPU in this title, despite the GPU's 92nd percentile standing among all GPUs.

Q: Does the CPU's memory bandwidth affect Thief performance?

A: The i5-10400F supports dual-channel DDR4 with 42.7 GB/s bandwidth, which is modest by modern standards. The benchmark results show consistent frame times (e.g., 96-130 FPS range at 1080p Medium), suggesting memory bandwidth isn't causing stutters, but the 12 MB shared L3 cache and 2.90 GHz base clock limit the CPU's ability to feed the GPU faster.

How This Combo Ranks

Among 3720 tested combos, this pairing sits at rank 1684 — almost exactly the median. This is a disappointing result given the RTX 5090's individual standing: the GPU ranks in the 92nd percentile of all GPUs with an average benchmark score of 79842, while the CPU ranks in the 68th percentile with an average score of 14185. The gap between the two components' capabilities explains why this combo doesn't crack the top half of tested systems in Thief.

The CPU's nearest rivals in synthetic testing — the Intel Xeon 6756E (0.2% higher score), Intel Core 7 160UL (0.3% lower), AMD EPYC 7552 (0.5% higher), and AMD Ryzen 3 7320C (0.6% lower) — all fall within a narrow 1.1% performance band. This clustering suggests that the i5-10400F's 6 cores and 12 threads deliver comparable throughput to enterprise Xeon and EPYC parts in general workloads, but Thief's reliance on single-thread performance (3DMark single-thread score of 688) means these synthetic similarities don't translate to gaming equivalence.

The RTX 5090's nearest rivals include the Tesla P100 PCIe 16 GB (0.3% higher), Tesla P100 PCIe 12 GB (0.6% higher), AMD Radeon RX 6850M XT (1.1% higher), and AMD Radeon Pro Vega 64X (1.4% lower). These deltas are all under 1.5%, showing the 5090 is in a performance tier with professional-grade GPUs, yet the Thief results don't reflect this dominance because the game's 2014 engine cannot scale with modern graphics hardware.

The rank of 1684 out of 3720 means 45.3% of tested combos perform worse and 54.7% perform better in Thief specifically. For a system with the most powerful consumer GPU available, this is a clear indication that the CPU is the limiting factor. The 10th-gen Core i5, released in 2020 (six years after Thief's launch), simply lacks the single-core muscle to push frame rates higher in this title, regardless of the GPU's capabilities.

Measured FPS Breakdown

At 1920x1080, the full settings sweep shows a tight band of performance. Low settings produce 125 FPS average, Medium produces 113 FPS (with a 96-130 FPS range), High produces 102 FPS, and Ultra produces 100 FPS. The delta between Low and Ultra is just 25 FPS, or 20%, which is remarkably small for a settings range that typically doubles or triples frame rates in GPU-bound scenarios. The minimum FPS at 1080p Medium is 96, which means even the worst-case frame times stay above 10ms — no noticeable hitches.

Moving to 2560x1440, the pattern repeats with a slight downward shift. Low settings deliver 110 FPS, Medium delivers 99 FPS (85-114 FPS range), High delivers 90 FPS, and Ultra delivers 89 FPS. The Low-to-Ultra gap narrows to 21 FPS (19.1%). The 1440p Medium minimum of 85 FPS is still comfortably above 60 FPS, making this resolution entirely playable even in demanding scenes.

At 3840x2160, the results compress further. Low settings produce 92 FPS, Medium produces 83 FPS (71-96 FPS range), High produces 75 FPS, and Ultra produces 74 FPS. The Low-to-Ultra delta shrinks to 18 FPS (19.6%). The 4K Medium minimum of 71 FPS is the lowest recorded minimum across all tests, but it remains above the 60 FPS threshold for smooth gameplay. The 4K Ultra result of 74 FPS average is respectable for a 2014 title, though it underutilizes the RTX 5090's potential.

Across all resolutions, the Medium preset provides the only complete dataset with min/max values. The 1080p Medium range spans 96-130 FPS (a 34 FPS spread), 1440p Medium spans 85-114 FPS (29 FPS spread), and 4K Medium spans 71-96 FPS (25 FPS spread). The narrowing range at higher resolutions suggests that GPU load becomes more consistent as pixel count increases, reducing CPU-induced frame time variability.

Settings Recommendations

For 4K displays, the Medium preset at 83 FPS average (71 FPS minimum) is the optimal choice. The jump from Medium to High costs 8 FPS (83 to 75), while High to Ultra costs only 1 FPS (75 to 74). The diminishing returns above Medium indicate that the CPU bottleneck caps performance around 74-75 FPS at Ultra/High, so the 9 FPS gain from dropping to Medium (83 vs 74) is essentially free performance headroom without sacrificing much visual quality. The 4K Low preset at 92 FPS is available for players prioritizing frame rate over visuals, but the 4K Medium result already exceeds 60 FPS with comfortable margins.

At 1440p, the Medium preset delivers 99 FPS with an 85 FPS minimum, making it the best balance of smoothness and fidelity. The High preset at 90 FPS and Ultra at 89 FPS are nearly identical, suggesting that the CPU saturates around 90 FPS at this resolution. Pushing to Low at 110 FPS provides a 11% improvement over Medium, but the visual downgrade is likely not worth the modest FPS gain. The 89-90 FPS range at High/Ultra is still well above the 60 FPS threshold, so players with 144Hz monitors can choose either preset based on visual preference.

For 1080p, the High preset at 102 FPS or Ultra at 100 FPS are both excellent choices. The 1080p Medium result of 113 FPS (96 FPS minimum) offers the best frame pacing, but the jump to High only costs 11 FPS and provides better image quality. The 1080p Low result of 125 FPS is the maximum achievable with this CPU, but it offers no meaningful advantage over High for most players since 102 FPS is already smooth. The recommendation across all resolutions is to start at Medium, verify the frame rate is acceptable, then adjust to High or Ultra only if the FPS remains above your target.

CPU Role

The Intel Core i5-10400F is a 6-core, 12-thread processor based on Comet Lake architecture, built on Intel's 14 nm process. It features a base clock of 2.90 GHz and a boost clock of 4.30 GHz, with 64 KB L1 cache per core, 256 KB L2 per core, and 12 MB of shared L3 cache. The CPU supports dual-channel DDR4 memory with 42.7 GB/s bandwidth and connects via PCIe Gen 3 with 16 lanes. Its 65W TDP makes it a modest power consumer, but its single-thread performance — a 3DMark single-thread score of 688 and PassMark single-thread of 2541 — is what matters most for Thief.

The benchmark data shows the CPU's multithreaded capabilities are solid for its class: Cinebench R23 multicore scores 10283, PassMark multithread scores 12115, and Geekbench multicore scores 6257. The 3DMark thread scaling from 2 threads (1350) to 4 threads (2560) to 8 threads (3929) to max threads (4735) shows reasonable scaling, but the single-thread score of 688 is the limiting factor in a game like Thief that relies on one primary game thread.

The CPU's 68th percentile ranking among all CPUs (average benchmark score 14185) places it above average but not exceptional. Its nearest rivals — the Xeon 6756E, Core 7 160UL, EPYC 7552, and Ryzen 3 7320C — all score within 0.6% of the i5-10400F, indicating that this chip sits in a densely populated performance tier. However, the RTX 5090's 92nd percentile ranking creates a massive capability gap between the two components.

In Thief, the CPU's role is to feed the GPU with draw calls and game state updates. The measured FPS ceiling of approximately 125 FPS at 1080p Low suggests that the CPU cannot process frames faster than this, regardless of the GPU's ability to render them. The 4.30 GHz boost clock helps, but the 14 nm process and Comet Lake architecture limit how much single-thread performance can be extracted. The 12 MB L3 cache is sufficient for the game's working set, as evidenced by the consistent frame times, but the 42.7 GB/s memory bandwidth may contribute to the plateau at higher FPS.

For a game released in 2014, this CPU delivers playable performance at all tested settings and resolutions. The bottleneck is not a stuttering or hitching issue — the minimum FPS values (96 at 1080p Medium, 85 at 1440p Medium, 71 at 4K Medium) all remain above playable thresholds. Instead, the CPU simply caps the maximum achievable frame rate, leaving the RTX 5090 with unused headroom. Players with high refresh rate monitors (165Hz or 240Hz) will not reach those frame rates in Thief with this combo, but those targeting 60-120 FPS will find the experience smooth and consistent.