Thief
This combination provides smooth gameplay with an average of 90 FPS, suitable for most gaming scenarios.
Average FPS: resolution vs quality settings
| Ultra | High | Medium | Low | |
|---|---|---|---|---|
| 4K Ultra HD | 40 | 42 | 56 | 73 |
| 1440p QHD | 68 | 72 | 94 | 124 |
| 1080p Full HD | 100 | 105 | 138 | 171 |
Cell color: green is 120+ FPS, teal is 60+, amber is 30+, red is below 30.
Thief with Intel Core Ultra 7 165H + NVIDIA GeForce RTX 5070, In-Depth Analysis
# Thief with Intel Core Ultra 7 165H + NVIDIA GeForce RTX 5070
Thief, the 2014 stealth classic, presents an interesting testbed for this mobile CPU and desktop GPU pairing. The benchmark data reveals a system that scales predictably across resolutions, with the RTX 5070 clearly dominating the workload at 1080p but ceding ground to the CPU at 4K. The measured frames per second show a combo ranking in the 1968th position out of 3720 tested combinations, placing it in the upper-middle tier of all tested systems. This analysis digs into what those numbers mean for actual gameplay, how the components interact, and where the bottlenecks emerge as resolution increases.
Resolution Scaling
The FPS drop from 1080p to 4K is substantial but not uniform across quality presets, which tells a revealing story about where the limiting component shifts. At Low settings, the average FPS falls from 171 at 1920x1080 to 73 at 3840x2160 — a 57% reduction. At Ultra settings, the decline is from 100 at 1080p to 40 at 4K, a 60% drop. These similar percentage losses across presets suggest the GPU is the primary constraint at every quality level, though the absolute numbers paint a more nuanced picture.
Consider the scaling between 1440p and 4K. At High settings, FPS drops from 72 to 42, a 42% decrease. At Medium, it goes from 94 to 56, a 40% reduction. This consistency implies the RTX 5070's rendering workload scales almost linearly with pixel count, which is typical for a GPU-bound scenario. However, the jump from 1080p to 1440p tells a different story. At Low settings, FPS goes from 171 to 124, only a 27% drop, while at Ultra it goes from 100 to 68, a 32% decrease. The smaller proportional penalty at lower resolutions hints that the CPU begins to matter more when the GPU has less work to do.
The data suggests that at 1080p, the Intel Core Ultra 7 165H's 16 cores and 22 threads can keep feeding the RTX 5070 adequately, but the GPU still dominates frame production. At 4K, however, the GPU's raw throughput becomes the hard ceiling. The 73 FPS average at 4K Low versus 171 FPS at 1080p Low — a 98 FPS gap — demonstrates that resolution scaling exacts a heavy toll on this combo. For gamers targeting 4K, the RTX 5070's 12 GB of GDDR7 memory and 672.0 GB/s bandwidth are sufficient, but the frame rates indicate that High and Ultra presets will struggle to maintain smoothness.
How This Combo Ranks
With a combo rank of 1968 out of 3720 tested combinations, this pairing sits just past the midpoint of the database. That rank might seem modest given the RTX 5070's individual performance percentile of 82, but several factors explain the position. The Intel Core Ultra 7 165H ranks in the 84th percentile among all CPUs, yet its average benchmark score of 34083 places it nearly equal to the AMD Ryzen 7 3700X, which scores 34260 with a deltaPct of -0.5%. This near-parity with a desktop CPU from an earlier generation suggests the mobile processor is not the bottleneck in most games.
The GPU's nearest rivals further contextualize the ranking. The RTX 5070's average benchmark score of 40377 is within 0.1% of the AMD Radeon Pro 580 (40318) and 0.8% ahead of the AMD Radeon Pro WX 7100 (40063). Interestingly, the RTX 5070 trails the AMD Radeon Pro 5300 by 1.2% (40870) and leads the NVIDIA RTX A500 Mobile by 2% (39568). These deltas are small, suggesting the GPU's raw compute power is competitive with professional workstation cards, but the rank of 1968 in Thief specifically indicates that game optimization and CPU interplay matter more than raw benchmark scores.
The rank implies that while this combo is capable, many other pairings deliver higher frame rates in this title. The data does not specify which combos rank above, but the position suggests that systems with stronger CPUs or more optimized pairings pull ahead. For a game released in 2014, the CPU's single-thread performance — a Cinebench R23 single-core score of 1759 — may be more relevant than its multi-core prowess, since older titles often rely on fewer threads. The rank of 1968 out of 3720 is a reminder that component strength in isolation does not guarantee top-tier game performance.
GPU Role
The NVIDIA GeForce RTX 5070 brings substantial hardware to Thief, and the measured frames reflect its capabilities. The GPU's 12 GB of GDDR7 memory on a 192-bit bus delivers 672.0 GB/s of bandwidth, which is ample for Thief's textures at any resolution tested. The base clock of 2325 MHz and boost clock of 2512 MHz are the operative frequencies during gameplay, and the benchmark data shows the GPU scaling its output as resolution and settings change.
At 1080p Ultra, the GPU produces 100 FPS average, but at 4K Ultra that drops to 40 FPS. This 60% reduction highlights the GPU's fill-rate limits at high pixel counts. The pixel rate of 201.0 GPixel/s and texture rate of 482.3 GTexel/s are the theoretical ceilings, and the measured results suggest Thief's rendering pipeline is hitting those limits at 4K. The 80 ROPs and 192 TMUs are the hardware units responsible for these operations, and the data shows they are fully utilized at 4K, given the steep FPS drop.
The GPU's FP32 performance of 30.87 TFLOPS is impressive on paper, but Thief's DirectX 11-era rendering likely does not leverage the full compute capability. The PassMark DirectX 11 score of 277 versus the DirectX 12 score of 108 indicates this GPU performs better in older API workloads, which aligns with Thief's 2014 release date. The RTX 5070's architecture supports DirectX 12 Ultimate (12_2), but the game's engine was built for earlier API versions. The result is that the GPU's modern features like ray tracing cores (48) and tensor cores (192) are largely idle, yet the raw rasterization power still drives playable frame rates at most settings.
Measured FPS Breakdown
At 1920x1080, the combo shows its strongest performance. Low settings deliver 171 FPS average, which drops to 138 FPS at Medium, 105 FPS at High, and 100 FPS at Ultra. The Medium preset includes a minimum FPS of 117 and maximum of 158, indicating smooth frame pacing with no catastrophic dips. The gap between High and Ultra is only 5 FPS, suggesting diminishing returns from the highest preset. The 71 FPS difference between Low and Ultra shows the GPU has headroom but the settings substantially affect output.
Moving to 2560x1440, the performance scales down proportionally. Low settings yield 124 FPS average, Medium produces 94 FPS (with min 80, max 109), High delivers 72 FPS, and Ultra drops to 68 FPS. The High-to-Ultra gap narrows to just 4 FPS, reinforcing that Ultra offers marginal visual gains for significant performance cost. The 56 FPS difference between Low and Ultra at 1440p is smaller than at 1080p, indicating the GPU is becoming more of a limiting factor.
At 3840x2160, the numbers become more challenging. Low settings still manage 73 FPS average, which is playable. Medium drops to 56 FPS (min 47, max 64), High falls to 42 FPS, and Ultra bottoms out at 40 FPS. The Medium preset's minimum of 47 FPS suggests occasional stutter, while High and Ultra will likely feel sluggish during busy scenes. The 33 FPS difference between Low and Ultra at 4K is the smallest across all resolutions, confirming that the GPU is the dominant constraint. The step from 1080p Low to 4K Low — a 98 FPS drop — is the largest single-resolution penalty in the dataset.
CPU Role
The Intel Core Ultra 7 165H is a 16-core, 22-thread processor based on the Meteor Lake architecture, built on Intel's 7 nm process. Its base clock of 3.80 GHz and boost clock of 5.00 GHz provide strong single-thread performance, as evidenced by the Cinebench R23 single-core score of 1759. The multi-core score of 14551 in the same test shows the processor's capability with threaded workloads, but Thief's age means it likely uses only a handful of cores.
The CPU's 24 MB of shared L3 cache and 2 MB per-core L2 cache are more than sufficient for a 2014 title, and the dual-channel DDR5 memory support with 89.6 GB/s bandwidth ensures data flows to the GPU without stalling. The PassMark single-thread score of 3509 and multithread score of 25933 demonstrate that the CPU performs well above average in both metrics, but the nearest rival data shows the competition is close. The AMD Ryzen 7 3700X, a desktop part, scores 34260 on average, just 0.5% higher than the Ultra 7 165H's 34083. This parity suggests the mobile CPU is not significantly holding back the RTX 5070 in Thief.
At 1080p Low, the 171 FPS average likely approaches the CPU's frame-delivery limit, since the GPU has relatively little work at such low settings. The 124 FPS at 1440p Low further supports this, as the CPU must still generate draw calls and physics updates. However, the fact that FPS scales nearly linearly with resolution at higher settings indicates the CPU is not the primary bottleneck in most scenarios. The Cinebench R15 single-core score of 257 and multi-core score of 2641 are modest compared to newer desktop chips, but for a game from 2014, the Ultra 7 165H's architecture provides ample headroom.
Settings Recommendations
Based solely on the measured FPS data, the optimal experience depends heavily on the target resolution. At 1920x1080, Ultra settings deliver 100 FPS average, which is smooth for most displays. The 5 FPS difference between High and Ultra at this resolution makes Ultra the recommended choice, as the visual quality improvement comes at minimal performance cost. Even the Medium preset's 138 FPS would be overkill for a standard 60 Hz monitor, so players can push to the highest settings without concern.
At 2560x1440, the recommendation shifts. High settings at 72 FPS are playable, but Ultra at 68 FPS is nearly identical and offers better visuals. The 4 FPS difference between these presets makes Ultra the sensible pick for 1440p gamers. The Medium preset's 94 FPS provides more headroom for high-refresh-rate displays, but the visual trade-off may not be worth it. Low settings at 124 FPS are unnecessary unless chasing maximum frame rates.
At 3840x2160, the choices become more constrained. Low settings at 73 FPS is the only preset that comfortably exceeds 60 FPS, making it the default recommendation for 4K. Medium at 56 FPS is borderline, and its minimum of 47 FPS could cause noticeable stutter during action sequences. High at 42 FPS and Ultra at 40 FPS are both below smooth thresholds, so they are only viable on displays with variable refresh rate technology. The data suggests that 4K players should accept Low settings or consider lowering the resolution to 1440p for a better balance of quality and performance.
FAQ
Q: Is this combo capable of running Thief at 4K with playable frame rates?
A: The data shows that at 4K Low settings, the combo averages 73 FPS, which is playable. Medium drops to 56 FPS with a minimum of 47 FPS, while High and Ultra fall to 42 and 40 FPS respectively. Only Low settings reliably exceed 60 FPS at 4K.
Q: How does the RTX 5070's performance in Thief compare to its nearest GPU rivals?
A: The RTX 5070 has an average benchmark score of 40377, which is 0.1% ahead of the AMD Radeon Pro 580 (40318) and 0.8% ahead of the AMD Radeon Pro WX 7100 (40063). It trails the AMD Radeon Pro 5300 by 1.2% (40870) and leads the NVIDIA RTX A500 Mobile by 2% (39568).
Q: What is the best settings preset for 1440p gaming?
A: At 2560x1440, Ultra settings deliver 68 FPS average, while High produces 72 FPS. The 4 FPS difference makes Ultra the recommended choice, as it offers the best visual quality with negligible performance loss. Medium at 94 FPS is better for high-refresh-rate displays.
Q: Does the Intel Core Ultra 7 165H bottleneck the RTX 5070 in Thief?
A: The CPU's average benchmark score of 34083 is within 0.5% of the AMD Ryzen 7 3700X (34260), indicating strong performance. The FPS scaling across resolutions suggests the GPU is the primary bottleneck at 4K, while the CPU may become more relevant at 1080p Low where it reaches 171 FPS.
Q: How large is the performance gap between 1080p and 4K at Ultra settings?
A: At Ultra settings, the average FPS drops from 100 at 1920x1080 to 40 at 3840x2160, a 60% reduction. This indicates that the RTX 5070's rendering pipeline is heavily taxed by the 4K pixel count, regardless of the CPU's capabilities.
Q: What does the combo rank of 1968 out of 3720 signify?
A: The rank places this pairing in the upper-middle tier of all tested combinations. Despite the GPU's 82nd percentile and CPU's 84th percentile performance, the combo does not reach top-tier status in Thief, suggesting other system configurations achieve higher frame rates in this specific title.
Hardware Specifications
Intel Core Ultra 7 165H
NVIDIA GeForce RTX 5070
FPS Benchmarks by Resolution & Settings
Performance data for Intel Core Ultra 7 165H + NVIDIA GeForce RTX 5070 in Thief
| Resolution | Settings | Avg FPS |
|---|---|---|
| 3840x2160 | Low | 73.0 |
| 3840x2160 | Medium | 56.0 |
| 3840x2160 | High | 42.0 |
| 3840x2160 | Ultra | 40.0 |
| 2560x1440 | Low | 124.0 |
| 2560x1440 | Medium | 94.0 |
| 2560x1440 | High | 72.0 |
| 2560x1440 | Ultra | 68.0 |
| 1920x1080 | Low | 171.0 |
| 1920x1080 | Medium | 138.0 |
| 1920x1080 | High | 105.0 |
| 1920x1080 | Ultra | 100.0 |
Thief with iUltra 7 165H + Other GPUs
See how Thief performs with the same CPU but different graphics cards.
Thief with RTX 5070 + Other CPUs
See how Thief performs with the same GPU but different processors.
Other Games with iUltra 7 165H + RTX 5070
Explore FPS benchmarks for other games using this same hardware combination.