Thief

Thief is a stealth title from 2014-02-24, and the combination tested pairs the AMD Ryzen 7 5700G with the NVIDIA GeForce RTX 4070. The CPU is an 8-core, 16-thread Zen 3 part built on TSMC's 7 nm process; it runs at 3.80 GHz base and 4.60 GHz boost, with a 65 W TDP and 16 MB of L3 cache. The GPU is an Ada Lovelace-based card with 12 GB of GDDR6X on a 192-bit bus, a 504.2 GB/s bandwidth, and a 2475 MHz boost clock. The dataset includes measured average FPS (and for Medium, min/max) at three resolutions (1920x1080, 2560x1440, 3840x2160) across four quality presets. In this game's combined ranking, the combo places 1044th out of 3720 tested combinations.

CPU Role — cores, clocks, and how they relate to this game's results

The Ryzen 7 5700G provides 8 Zen 3 cores and 16 threads, with a base of 3.80 GHz and a boost of 4.60 GHz. It is manufactured on TSMC's 7 nm node, which helps keep TDP at 65 W, and it supports dual-channel DDR4 memory with a bandwidth of 51.2 GB/s. In synthetic CPU benchmarks, its 3DMark 16-thread score is 6629, and the single-thread score is 899; the max-thread score is 6627, indicating that the 8-core/16-thread design scales well. Cinebench R23 altmetric returns 12830.5 for multi-thread and 1494 for single-thread, while Geekbench reports 1894 single-core and 9950 multi-core. PassMark's multithread metric reaches 24419, with single-thread at 3283.

Place in the context of the full CPU pool, the 5700G sits at the 79th percentile among all CPUs, which implies it outperforms a solid majority but is not near the very top. Its average benchmark score of 27051 is extremely close to the nearest ranked rivals—the Intel Core i9-9900K (27097, a delta of -0.2%) and the AMD Ryzen AI 7 445 (26936, +0.4%). In other words, the 5700G delivers almost identical compute throughput to treating i9-9900K within a 0.2% margin, and it edges the Ryzen AI 7 445 by 0.4%. The processor is also virtually tied with the Intel Core i5-14400 (27166, -0.4%) and Intel Core i7-1370P (26900, +0.6%). Therefore, from a pure CPU capability standpoint, this part is more than enough for a 2014 stealth title.

How does that translate to Thief? The FPS results show that even at the most demanding resolution, the CPU does not appear to throttle framerates. At 1080p, the average FPS tops at 161 on Low and drops to 130 on Ultra—the difference is width becomes 31 FPS. At 1440p, Low reaches 142 and Ultra hits 96, while at 4K, Low gives 103 and Ultra gives 56. In all cases, the rate is determined primarily by GPU work, as the CPU's single-thread effort (899 on 3DMark) is far higher than the minimum needed to maintain the measured frame rates. The 5700G's 8 cores and 16 threads are unlikely to be a constraint at any of these settings; the substantial FPS gaps between resolutions—e.g., 161→103 on Low 1080 to 4K—signature clearly a graphics-side limit because only the GPU workload scales with pixel count.

GPU Role — VRAM, clocks, and how they relate to this game's results

The NVIDIA GeForce RTX 4070 uses the AD104 chip, built on TSMC 5 nm, with 12 GB GDDR6X memory attached to a 192-bit bus. The GPU runs at a base clock of 1920 MHz, boosts to 2475 MHz, and has a memory clock of 1313 MHz (effective 21 Gbps), producing a bandwidth of 504.2 GB/s. It contains 5888 shading units, 46 RT cores, and 184 tensor cores; its pixel rate is 158.4 GPixel/s and texture rate is 455.4 GTexel/s. The card's TDP is 200 W and it occupies a dual-slot width. It supports PCIe 4.0 x16 and offers 1x HDMI 2.1 and 3x DisplayPort 1.4a outputs.

In synthetic benchmarks, the RTX 4070 achieves a 3DMark Steel Shadow Nomad score of 3854, a Geekbench OpenCL score of 154858, and a Geekbench Vulkan score of 174152. PassMark's G3D yields 26927, and its GPU compute average is 14720. The overall average benchmark score is 37648, placing it at the 81st percentile across all GPUs. Its closest rivals are the NVIDIA Tesla P4 (37628, delta +0.1%), AMD Radeon RX Vega 56 (37507, +0.4%), NVIDIA GeForce RTX 4080 Mobile (38135, -1.3%), and AMD Radeon PRO W6400 (37157, +1.3%). The delta against the RTX 4080 Mobile shows the full-size RTX 4070 lags by only 1.3% in synthetic workloads despite being a smaller desktop variant—a minor difference in performance.

In Thief, the GPU precisely governs frame output. The severity of the resolution-driven drop reinforces this: at Low preset, the average falls from 164 (1080p) to 142 (1440p) and then to 103 (4K); at Ultra, it falls from 130 to 96 to 58. The progression indicates a GPU limit. The 12 GB VRAM is ample for this game, as even 4K Ultra consumes, but the FPS drops are bound to compute shading (and the indoor-stealth scenes) rather than memory capacity. The RT 4070's score of 26982 PT in 3DMark Nomad is far above the needs of the game, but the FPS data show that 4K output still demands a lot because the engine/pipeline ends up GPU-bound at that higher pixel count.

Settings Recommendations — which preset gives the best experience per the measured rows

The measurements provide average FPS for four presets at each resolution; for Medium only, min and max are also reported. The player should target an average that does not fall below typical comfort. Comfy threshold is generally 60 specifics FPS; we can use that as a practical midpoint.

For 1080p, all presets exceed 130 FPS—Low 161, Medium 143 (min 124, max 165), High 132, Ultra 130. There is no penalty to pick Ultra, as it's 30 FPS below Low but still far above the comfort range. The difference between High and Ultra is a mere 2 FPS (132 vs 130), so Ultra is the sensible choice.

At 1440p, Medium 129 (min 109, max 148) and High 101 are both above 100 FPS; Ultra 96 also satisfies the 60+ marker. Given only 6 FPS separates Medium 129 from Ultra 96, Ultra is playable, but if headroom is sought, High at 101 offers a mix of visuals and smoothness. Ultra would be the recommended pick for users on 144Hz 1440p panels, as 86 minimal falls in the typical range.

At 4K, the picture changes. Low gives 103 FPS (still above some floor), Medium 78 (min 66, max 89), High 59, and Ultra 56. Here, 56 and 59 average are below a 60 target, so High and Ultra are borderline for a V-sync-free experience. Medium 78 makes a comfortable margin with a minimum of 66, making it the best balance between quality and performance. Thus, the data indicates that for 4K, the Medium preset is the reasonable recommend, because it holds a 66 minimum and averages 78, while High and Ultra do not keep up a 60-frame floor. For lower resolutions, choosing Ultra yields the visual peak with no compromise above 130 FPS.

FAQ

Q: What average FPS does this system deliver in Thief at 1440p with High settings?

A: At 2560x1440, High settings give an average of 101 FPS. There is no reported min or max for those rows, but the average precisely 101.

Q: Can the RTX 4070 handle 4K Ultra in Thief?

A: Yes, but only at an average of 56 FPS, with no min or max data. For mainstream expectations of 60+ FPS, 4K Ultra falls short; 4K Medium offers 78 average (min 66, max 89), so Medium is the strongest reachable preset that stays above 60.

Q: What is the minimum FPS recorded across all resolutions and settings?

A: The only preset with a recorded minimum is Medium—124 at 1080p, 109 at 1440p, and 66 at 4K. No min was listed for Low, High, or Ultra, so that datum can't be compared across all settings.

Q: How does the RTX 4070 compare to the NVIDIA GeForce RTX 4080 Mobile in average benchmark score?

A: The RTX 4070's average benchmark, 37648, is 1.3% lower than the RTX 4080 Mobile's 38135 (delta -1.3%). In synthetic tests it sits right next to the Tesla P4 (37628) and Radeon RX 56 (37507), within a 0.4%.

Q: What is the CPU's ranking among all CPUs?

A: The Ryzen 7 5700G is at the 79th percentile for all CPUs, with an average benchmark score of 27051.

Q: What is the combo's rank among all tested hardware combinations in Thief?

A: The duo composed of the 5700G and RTX 4070 is ranked 1044th out of 3720 combos analyzed for this game.

Resolution Scaling — how FPS drops from 1080p to 4K and what this tells about the limiting component

In Thief, the impact of resolution on framerate is linear and steep. Consider the Low preset: at 1920x1080, the average is 161 FPS; rising to 2560x1440 cuts that to 142; and at 3840x2160 it sits at 103. In percentage terms (using raw numbers stored in the pack), the progression of 161→142→103 shows a 37% laugh from 1080 to 4K. For the Medium preset, the numbers are 143 (with min 124 and max 165) at 1080p, 129 (min 109, max 148) at 1440, and 78 (min 66, max 89) at 4K—so exactly the same pattern of loss. High settings deliver 132 FPS at 1080, 101 at 1440, and 59 at 2150. Ultra yields 128, 96, and 58.

IMPORTANTLY, at 1080p the CPU is almost never in the picture—the Low 161 and Medium 143 (on a 2014 gaming title) shows the per-pixel work is far from the GPU's coverage. The spread between 1080 and 1440 (e.g. 161→142 for Low) is moderate, while from 1440 to 4K (142→103) it is more than 27 FPS; low-FPS falloff increases with the extra 4.5 million pixels. This divider strongly indicates a GPU-bounded stage. A CPU-bound program would show very similar frame rates across different resolution hashes meaning numbers near constant. Instead the FPS decreases by almost 60% from 8 to 4K (161 to 103 at Low), a pure graphic scaling curve that is typical of a raster-limited scenario.

Therefore, among the components, the NVIDIA GPU is the primary performance restrictor, especially at 1440p and 4K. The Ryzen 7 5700G's compute is never seen as a limiting effect in this title; its single-thread score of 899 (in 3DMark) and 16 threads are far beyond the need for a stealth game from 2014.

Measured FPS Breakdown — resolution by resolution, settings by settings, exact numbers

The table below provides the measured data from the FACT PACK. "—" indicates that the min and max are not reported.

| Resolution | Preset | Average | Minimum | Maximum |

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

| 1920x1080 with Low | 1920x1080 | Low | 161 | — | — |

| 1920x1080 | Medium | 143 | 124 | 165 |

| 1920x1080 | High | 132 | — | — |

| 1920x1080 | Ultra | 130 | — | — |

| 2560x1440 | Low | 142 | — | — |

| 2560x1440 | Medium| 126 | 109 | 148 |

| 2560x1440 | High | 101 | — | — |

| 2560x1440 | Ultra | 96 | — | — |

| 3840x2160 | Low | 103 | — | — |

| 3840x2160 | Medium| 78 | 66 | 89 |

| 3840x2160 | High | 59 | — | — |

| 3840x2160 | Ultra | 56 | — | — |

These numbers show three discrete performance tiers. At 1080p all settings deliver between 130 and 161 as typical. At 1440p the gap narrows further where Low (142) and Medium (129) are 13 apart, but High at 101 and Ultra at 96. At 4K, even Low is above 100 at 103, Medium is 78, High 59, Ultra 56; only Medium clearly crosses the 60+ threshold.

The Minimum and Maximum values (recorded only for Medium in each resolution) reveal stability: the 4K row has min 66, indicating that the frame times never dip below 66 FPS even if the averages afford 59 for High. That tells the pacing is smoother on the Medium preset than the numbers alone might suggest; for High Gamma, the absence of min may leave the actual discomfort unmeasured.

How This Combo Ranks

Among the 3720 world of tested platform/platform/sonde combinations in Thief, the combo AMD Ryzen 7 5700G + RTX 4070 achieves rank 1044. Placed in context, this is within the upper‑intermediate tier—nearly top 30% but not the peak. The CPU's percentile is 79 and the GPU’s is 81, both slightly above the average baseline must take into account that for this specific game, the ranking of 1044 includes also many lower-end CPUs paired with high-end GPUs, and vice versa.

The rank is an explicit head‑to‑head comparing the combination, so it is unlikely to align with the sum of each part's popularity. For example, the CPU sits near the Intel Core i9‑9900K (‑0.2% delta) in synthetic load, while the GPU is on par closest to Tesla P4 (0.1%) and Radeon RX 56 (0.4%). FPS numbers position this pairing as a 1080p‑Ultra cattle beast and a 1440p‑Medium Strong. But in Thief specifically, the engine stalls less on the CPU. Consequently, the 1044 rank primarily reflects that the RTX 4070 is finite at 4K workloads, hence the middle‑table standing.

If a user looks for the strongest combos in again, the difference suggests that a few pairings with faster GPUs (e.g., likely those with RTX 4080/7900) or slightly faster CPUs occupy the top ranks. In contrast, this combination remains completely removed from entry‑level territory; its FPS data are consistently above 56 at the hardest 4K preset, and the 1080p low‑end 161 remains in the top echelon of playable rates. So the measured data place it in the 1044‑th spot out of 3720, aligning with its CPU and GPU percentiles and the fact that 4K is near‑bounds by the video card.