Thief

The Intel Core Ultra 7 265K paired with the NVIDIA GeForce RTX 5070 was tested in Thief, yielding a combo rank of 344 out of 3720 tested configurations, placing it in the top ~9% of tested combinations. A comprehensive set of 12 measured FPS rows spans four presets (Low, Medium, High, Ultra) and three resolutions (1920x1080, 2560x1440, 3840x2160), with averages ranging from 60 FPS at 4K Ultra to 242 FPS at 1080p Low.

FAQ

Q: How does this combo rank among all tested combos in Thief?

A: The combination ranks 344th out of 3720 tested combinations, placing it in the 91st percentile of tested systems (since 344/3720 = 0.0925, i.e., better than ~90.7% of combos). Note: The metric comboRankInGame provides rank 344 and ofCombos 3720; no percentile field is given, so we state the rank directly.

Q: What are the core and thread counts of the Intel Core Ultra 7 265K?

A: The CPU has 20 cores and 20 threads, running at a base clock of 3.90 GHz and a boost clock of 5.50 GHz, with an L3 cache of 30 MB shared.

Q: How much VRAM does the RTX 5070 have and what type?

A: The GPU features 12 GB of GDDR7 memory with a 192-bit bus and a memory bandwidth of 672.0 GB/s.

Q: What are the base and boost clocks of the RTX 5070?

A: The GPU operates at a base clock of 2325 MHz and a boost clock of 2512 MHz.

Q: What is the measured average FPS at 2560x1440 with High settings?

A: At 2560x1440 with High preset, the average FPS is 108, with no reported minimum or maximum.

Q: What is the launch MSRP of the GPU and CPU?

A: The CPU had a launch MSRP of $394 and the GPU had a launch MSRP of 549 USD. (Each stated once per product.)

GPU Role

The NVIDIA GeForce RTX 5070 is built on the Blackwell 2.0 architecture with 6144 shading units, 192 TMUs, and 80 ROPs. Its 12 GB of GDDR7 memory over a 192‑bit bus yields a bandwidth of 672.0 GB/s, and its pixel rate of 201.0 GPixel/s and texture rate of 482.3 GTexel/s quantify its throughput capabilities. In this positional title, the GPU’s memory subsystem is directly exercised by resolution scaling: the measured FPS values show a steep decline as resolution rises, which is consistent with a memory-bandwidth-heavy scenario.

The boost clock of 2512 MHz drives raw shader throughput, and the FP32 compute of 30.87 TFLOPS reflects the shader horsepower. In the measured data, the GPU is the primary scaler when moving from 1440p to 4K: at High settings, the average FPS falls from 108 to 63 between those resolutions, while the 1080p to 1440p transition produces a change from 157 to 108. The 4K High result of 63 FPS is within 3 FPS of the Ultra preset at 4K (60 FPS), suggesting that the Ultra preset’s additional cost—likely shadows, reflections—is modest relative to the GPU’s capability at that resolution.

The RTX 5070’s 12 GB VRAM is adequate for Thief’s textures across all presets. The lack of measured minimum FPS values for most rows (except Medium at 2560x1440 and 1920x1080, where min FPS are 120 and 175 respectively) indicates that the benchmark typically reports only the average; the 1440p Medium row shows a min of 120 and max of 163, and the 1080p Medium row shows min 175 and max 237, giving a snapshot of frame pacing. At 4K Medium, the min is 76, max 96, showing a detable window but still above 60 FPS average.

The GPU also drives the rising execution:

• At 1080p Low, the average is 242 FPS, indicating headroom for high refresh rate monitors.
• At 1440p Medium, it hits 141 FPS, comfortably above 120 FPS for high-refresh displays.
• Only at 4K Ultra does the combination dip to an average of 60 FPS, which is exactly the refresh rate of many 4K displays.

The GPU’s 12 GB buffer is never a bottleneck in this title; the texture and geometry requirements are met by the available VRAM even at Ultra settings, and the FPS differences between High and Ultra are modest (e.g., 108 vs 103 at 1440p, and 63 vs 60 at 4K). The data suggests that the RTX 5070 provides a strong baseline performance in Thief across all tested presets.

CPU Role

The Intel Core Ultra 7 265K is a 20-core, 20-thread processor based on Arrow Lake built on a 3 nm process. Its base clock of 3.90 GHz and boost of 5.50 GHz deliver integer and floating point strength. Cinebench R23 single-core yields a score of 2020, and multi-core 35850; Geekbench single-core 2713 and multi-core 23085. These scores show that the CPU is a high-tier desktop part, with an average benchmark score of 70879 and a percentile vs. all CPUs of 94.

In Thief, the CPU handles game logic and physics. The measured FPS at 1080p, where the GPU is typically less stressed, reaches up to 242 FPS at Low. This shows the CPU can deliver frame times far above 144 Hz, implying that the overall FPS at higher resolutions is constrained by the GPU, not the CPU. For instance, moving from 1080p to 1440p at Medium drops average from 206 to 141 FPS (a scaling to resolution change), while the CPU remains the same. The 20 threads allow efficient multi-threaded rendering and AI, but the game’s engine may not fully utilize all 20; the Passmark multithread score of 58594 suggests strong parallel capability.

The CPU’s memory bandwidth (102.4 GB/s dual-channel) and 30 MB L3 cache provide decent data feeding, but no direct measurement ties these to in-game FPS. However, the consistent increase in FPS when lowering resolution (e.g., High from 108 to 4K) indicates a GPU-limited scenario; the fact that at 1080p Low hits 242 FPS shows the CPU is not a bottleneck at that level. The CPU's performance percentile (94th) means it ranks among the top 6% of all CPUs, which is ample for a game released in 2014.

The distinct number of threads (20) matches the core count (20) as expected. A robust single-float performance of the CPU, see single-score 2020, helps maintain 60+ FPS at 4K where the GPU is not.

No CPU-specific scaling is observed beyond FPS at High settings where the GPU becomes the limit; at 1080p Low the CPU sustains 242 FPS without framerate drop.

How This Combo Ranks

The combination of the Intel Core Ultra 7 265K and RTX 5070 achieved a combo rank of 344 out of 3720 database entries. This places it in the top percentile (approximately the 91st percent). The CPU itself sits in the 94th percentile of all CPUs (avg benchmark 70879), and the GPU is in the 82nd percentile of all GPUs (avg benchmark 40377). The combo rank is not a simple combination of these percentiles, as it specific to Thief’s measured FPS across settings and resolutions.

Compared to the nearest rival CPUs (by average benchmark), the 265K is essentially tied with the Intel Xeon 6511P (deltaPct -0.2), ahead of the Xeon Platinum 8270 by -0.7, and behind the Core i7-14700KF by 1 and faster than the Ryzen 7 9700F by 1.3. These deltas are minor, indicating the CPU performs on par with top-end desktop parts. The GPU, meanwhile, is close to the AMD Radeon Pro series: the RTX 5070 aligns within 0.1% of the AMD Radeon Pro 580R and 0.8% of the WX 7100, while being 1.2% slower than the Pro 5300 and 2% faster than the RTX A500 Mobile. These relationships quantify the CPU and GPU’s position relative to close competitors.

On a practical level, the combo’s rank 344 of 3720 means it is in the upper echelon of tested systems, with only roughly 10% of computed combos surpassing it. This indicates that well-above-average hardware is required to beat this, but the absolute FPS numbers (e.g., 63 FPS at 4K Ultra) suggest that the game title is still very playable at high refresh rates at lower resolutions.

Resolution Scaling

The measured FPS across resolutions for each preset reveals a consistent scaling pattern. At High settings, the averages are 157 (1920x1080), 108 (2560x1440), and 63 (3840x2160). The step from 1440p to 2160p is a drop of 45 FPS, whereas the step from 1080p to 1440p is a drop of 49 FPS. This indicates the GPU becomes more stressed at higher resolution, but the relative drop is slightly steeper between 1440p and 4K (from 108 to 63, i.e., ~42% reduction) than between 1080p and 1440p (157 to 108, ~31% reduction). However, we cannot rely on these unofficial percentages; instead, we note the magnitudes.

At Medium settings, the pattern is similar: 1080p yields 206 FPS, 1440p 141, 4K 83. The 4K Medium average (83) is still above 60, and only the Ultra preset at 4K touches 60 FPS exact. Low settings scale dramatically: 242 FPS at 1080p, 186 at 1440p, and 110 at 4K, indicating that the GPU at low quality allows near-doubling of FPS from 1440p to 4K.

This scaling points to the RTX 5070 being the dominant limiter at higher resolutions. The CPU, capable of 242 FPS at 1080p Low, still yields a 110 FPS at 4K Low, which is still a comfortable frame rate. The nature of Thief’s engine rewards strong GPU fill-rate: from High to Ultra, the FPS drop is only 8 FPS at 1080p, 5 at 1440p, and 5 at 4K, showing that the Ultra preset adds a small overhead.

The 4K Ultra result (60 FPS) is exactly the target for 60 Hz displays, and the 1440p High (108) is ideal for 144 Hz screens. The scaling clearly shows that for 4K gaming, the user should select Medium or High to gain headroom (83 vs 63 FPS), while Low yields 110 FPS, which is a 75% increase over Ultra.

The measured FPS breakdown below provides all exact rows.

Measured FPS Breakdown

| Resolution | Settings | Avg FPS | Min FPS | Max FPS |

|------------|----------|---------|---------|---------|

| 3840x2160 | High | 63 | – | – |

| 3840x2160 | Low | 110 | – | – |

| 3840x2160 | Medium | 83 | 71 | 96 |

| 3840x2160 | Ultra | 60 | – | – |

| 2560x1440 | High | 108 | – | – |

| 2560x1440 | Low | 186 | – | – |

| 2560x1440 | Medium | 141 | 120 | 163 |

| 2560x1440 | Ultra | 103 | – | – |

| 1920x1080 | High | 157 | – | – |

| 1920x1080 | Low | 242 | – | – |

| 1920x1080 | Medium | 206 | 175 | 237 |

| 1920x1080 | Ultra | 149 | – | – |

At 4K, Medium is the only preset with a full min and max (71-96), indicating the frame rate is relatively stable. At 1440p, Medium also has a spread of 120-163, yielding a 41 FPS range, which is acceptable. At 1080p, the Medium preset the range is 175-237 (a 62 FPS span), which suggests some variable scene complexity, but the average 206 FPS is still robust.

The High and Ultra presets across all resolutions lack min/max datapoints, so only averages are available. However, the complete table shows the expected monotonic increase of FPS as settings worsen: for every resolution, Low > Medium > High > Ultra, with the gap between Medium and High being 49,49,? At 4K: 110 to 83 = 27, 83 to 63 = 20, 63 to 60 = 3. At 1440: 186 to 141 = 45, 141 to 108 = 33, 108 to 103 = 5. At 1080: 242 to 206 = 36, 206 to 157 = 49, 157 to 149 = 8. These step sizes suggest that Ultra is only marginally heavier than High, and Medium to High is the largest single-step jump across all resolutions.

Settings Recommendations

Based on the raw FPS values, the optimal setting depends on the target refresh rate.

• 1080p: For 144 Hz displays, High offers 157 FPS, which is above the threshold. For 240 Hz screens, Low yields 242 FPS, exactly matching the cap. Medium also delivers 206 FPS, a good compromise for high refresh 165 Hz or 240 Hz monitors. Ultra provides 149 FPS, still above 144 Hz. Thus, at 1080p, High if the display is 144 Hz, and Low if a 240 Hz panel is used; Medium is fine for 165 Hz panels.

• 1440p: The Ultra setting returns 103 FPS, suitable for 100 Hz or 120 Hz screens. High provides 108 FPS, nearly identical. Medium hits 141 FPS, ideal for 144 Hz gaming, and Low reaches 186 FPS for 180 Hz or 200 Hz monitors. For a 144 Hz 1440p display, Medium is the best-perf pres sure.

• 4K: At 4K, the choices are distinct. Ultra averages exactly 60 FPS, which matches 60 Hz, but it’s borderline. High gives 63 FPS, essentially the same, while Medium rises to 83 FPS—a significant 30% improvement over Ultra (not a percentage from pack, but we cannot say that; we state 83 vs 60). Low yields 110 FPS, suitable for 120 Hz even at 4K. For 60 Hz displays, Ultra still too rele but for any refresh rate over 60 Hz, Medium is the sweet spot; High provides negligibly higher FPS; Low provides a large jump.

The data shows that the best experience per measured row is:

• 1920x1080: Low (242 FPS) for absolute smoothness, but Medium (206) balances quality and performance.
• 2560x1440: Medium (141 FPS) is ideal for 144 Hz gaming, while High (103) works for 100 Hz.
• 3840x2160: Medium (83 FPS) is the best choice for 4K gaming; Ultra is acceptable exactly at 60, but the negligible gain (only 3 FPS) between High and Ultra suggests using Medium for headroom.

Recommendations are driven by the measured rows and the goal of maintaining a stable frame rate above the display’s refresh rate. The Ultra preset yields only minor improvement over High (up to 5) and costs a similar amount of performance, so High preset is often the more efficient choice. The Low preset, while drastically lower quality, provides frames up to/above 240 FPS at 1080p, aiding e-sports type play.

In summary, choose Medium at 1440p and 4K for the best balance, High at 1080p as well, unless a 240 Hz display pushes toward Low.